Datasets:

alon-albalak

/

licensed_pubmed

like 0

PMC3053314

Introduction {#s1} ============ Ion channels are the fundamental elements underlying neuronal excitability and information transfer, inter- and intracellularly. These protein pores, found also in other excitable cell types, undergo fast conformational modifications (hereafter referred to as *channel gating*) induced by a change in the electric field or by the binding of ligand molecules. By doing so, channels selectively affect the ionic conductances of the membrane and enable ions to flow according to their electrochemical potentials [@pcbi.1001102-Johnston1]. The impact of the first quantitative deterministic description of conductance gating [@pcbi.1001102-Hodgkin1] was extremely significant, as testified by its wide use up to today [@pcbi.1001102-DeSchutter1]. Since the 1970s however, the stochastic nature of the single ion channels gating has been fully recognised. The resulting random fluctuations in the membrane conductances (which are known as *channel noise*) have been the subject of intense theoretical and experimental research [@pcbi.1001102-Chen1]--[@pcbi.1001102-Mino3]. Nevertheless, only recently *channel noise* was emphasised to have a significant impact on neuronal signals generation, propagation and integration, and it was suggested for consideration in realistic models of single neurons [@pcbi.1001102-Mainen1]--[@pcbi.1001102-Jacobson1]. In some parts of the peripheral nervous system, *channel noise* has been demonstrated to be prominent for information transfer and perception (e.g., see [@pcbi.1001102-Bruce1] and references therein). However, in the central nervous system whether or not *channel noise* plays a role at the level of large networks of interacting neurons, how heterogeneous ion channel types contribute to spontaneous network firing, and whether *channel noise* combines or interferes with other sources of noise (synaptic, for instance) remain open questions. Towards addressing these questions, the increasing availability of cheap parallel computing resources and improved algorithms [@pcbi.1001102-Migliore1], [@pcbi.1001102-Hines1] allow one to approach *in silico* the study of networks of thousands of morphologically detailed multi-compartmental model neurons [@pcbi.1001102-Markram1]. In addition, a diversity of voltage- and ligand-gated ion channel types can be included in these large models with biophysical realism [@pcbi.1001102-Druckmann1]. Unfortunately, *channel noise* is rarely considered for large network simulations or detailed multi-compartmental models [@pcbi.1001102-Cannon1], due to its heavy computational load. Implementing single-channel stochastic models explicitly, for each of the thousands of channels per ion conductance type and per neuron, requires Montecarlo simulation techniques [@pcbi.1001102-Clay1], [@pcbi.1001102-Chow1], [@pcbi.1001102-Schneidman1], [@pcbi.1001102-Skaugen1] that are computationally intensive even for single compartmental neurons, regardless of excellent speed-up techniques [@pcbi.1001102-Chow1]. Throughout this paper, we refer to such explicit and *exact* simulation methods by the term *microscopic*, regardless of the details of their actual numerical implementation [@pcbi.1001102-Mino4]. For the specific case of the Hodgkin-Huxley (HH) equations, Fox and collaborators proposed an alternative approximate method to mimic *channel noise*, avoiding a microscopic description of the individual channels [@pcbi.1001102-Fox1], [@pcbi.1001102-Fox2]. This method relies on the use of stochastic differential equations to *macroscopically* account for the fluctuations in the overall conductance of sodium and potassium channels, with formal analogies to the Langevin equation [@pcbi.1001102-Chow1], [@pcbi.1001102-Cox1]. Although this approach is very attractive and was employed widely in the literature (see references in [@pcbi.1001102-Bruce2]), its accuracy was recently challenged and debated by several authors [@pcbi.1001102-Mino4], [@pcbi.1001102-Bruce2]--[@pcbi.1001102-Sengupta1]. These authors compared numerical simulations of the exact microscopic descriptions of the HH model with those obtained by Fox\'s, finding some inconsistencies. It was however only with the work by Bruce (2009), that a straightforward test and framework were proposed to quantify the accuracy of Fox\'s algorithm. Simulating a *voltage-clamp* experiment, while recording ion currents, clearly shows that Fox\'s approximation does not capture correctly the microscopic statistical properties, regardless of how large the number of single ion channels to be approximated is. An *ad hoc* partial correction of Fox\'s algorithm - based on the simultaneous Montecarlo simulations of single channels - was also proposed for some activity regimes [@pcbi.1001102-Bruce2], but it cannot be generalised to arbitrary simulation conditions. In this paper we introduce and operatively define a general method, based on the diffusion approximation [@pcbi.1001102-Cox1], to transform any deterministic model neuron into its *effective* stochastic version, for an arbitrary set of ion conductances. As in previous studies, we focus on discrete Markov processes [@pcbi.1001102-Hille1], [@pcbi.1001102-Colquhoun1], routinely employed in the experimental identification of voltage-gated channels and synaptic receptors. Our purpose is to reintroduce channel noise in deterministic conductance-based models with limited computational overhead. We also aim at accurately replicating the statistical properties of ion conductances, as predicted by the exact microscopic description, while avoiding the use of any *ad hoc* correction or heuristics in the choice of the parameters [@pcbi.1001102-Saarinen1]. Our approach is related to previous Langevin-based formulations, although with a significant difference in the way channel fluctuations are reintroduced in model equations. It can be considered as an accurate and systematic generalisation of Fox\'s algorithm, to the case of voltage-, ion-, and ligand-gated channels with arbitrary complexity. We numerically compare our approach to that by Fox and we provide, as a [Supporting Information](#s5){ref-type="sec"}, some analytical results showing where it fails. We validate our approach for single-compartmental neuronal simulations, incorporating HH fast inactivating sodium channels and delayed rectifier potassium channels, analogously to previous works. By comparing our effective method to the exact simulations of the stochastic channel kinetic schemes, we obtain satisfying agreement. Materials and Methods {#s2} ===================== In this section, we briefly review the deterministic HH model and then introduce our algorithm. We present our method for ion channels whose microscopic correlate is represented by a population of identical 2-state channels. Only in this specific case, our method coincides with Fox\'s approach. We then generalise the method to channels characterised by -state kinetics and show that, for the special case of multiple independent subunits, each composed by 2-state gating mechanisms as in HH-like currents, the mathematical expressions underlying our algorithm greatly simplify. Neuron model {#s2a} ------------ We consider a single-compartmental conductance-based neuron model [@pcbi.1001102-Dayan1]. For this class of models, the membrane potential obeys the following current balance equation [@pcbi.1001102-Johnston1] where is the specific membrane capacitance and is an externally applied current density (expressed in ). These models comprise a leak current and a number of intrinsic (as well as synaptic) currents that can be similarly expressed by an *ohmic* relationship , which links the current to the membrane potential. Each ionic conductance is completely determined by the fraction of corresponding channels in the *open* state (see [Fig. 1A--D](#pcbi-1001102-g001){ref-type="fig"}). ::: {#pcbi-1001102-g001 .fig} 10.1371/journal.pcbi.1001102.g001 Figure 1 ::: {.caption} ###### Markov kinetic schemes. In the simplest 2-state kinetics (**A**), a single channel can be in one of two configurations with only one of them associated to a non-zero conductance (filled grey circle). The kinetic parameters and are rates, as they represent the transition probabilities between states, expressed per time unit. In a more general case, single-channel kinetics is described by an -state scheme. Voltage-gated fast-inactivating sodium (**B**) and delayed-rectifier potassium channels (**C**) are two examples, where only one state corresponds to a non-zero channel conductance (filled grey circle). An alternative model for sodium channels (**D**) (Vandenberg and Bezanilla, 1991) is also shown for comparison. We point out that our method can be applied to any kind of kinetic schemes, where the transition rates are known. For (**B--C**), each state is identified by an arbitrary name convention (, , , etc.), referring to the underlying mapping of these 8- and 5-state channels into multiple 2-state gated subunits (panel **E**). Indeed, some -state kinetic schemes may be mapped into, or experimentally identified as, a set of independent 2-state gates: the open state of the full scheme corresponds to all the elementary gates in the open states, simultaneously. For instance, the kinetic scheme (**B**) could be mapped into a set of four independent 2-state gates (**E**) (i.e., the familiar *activation* gates and the *inactivation* gate of sodium fast-inactivating currents), three of whom are identical. :::  ::: For reference to previously published papers [@pcbi.1001102-Mino1], [@pcbi.1001102-Mino2], [@pcbi.1001102-Mino3], [@pcbi.1001102-Schneidman1], [@pcbi.1001102-Steinmetz1], [@pcbi.1001102-Fox2], we consider here the HH voltage-gated currents and with standard parameters [@pcbi.1001102-Hodgkin1]. Therefore, we consider and . In the deterministic model, and are expressed phenomenologically as a product of activation and inactivation deterministic variables [@pcbi.1001102-Fleidervish1]--[@pcbi.1001102-Vandenberg1]:Each of these variables obeys a first-order ordinary differential equation of the form where and , are kinetic parameters. All the model parameters are summarised in [Table 1](#pcbi-1001102-t001){ref-type="table"}. ::: {#pcbi-1001102-t001 .table-wrap} 10.1371/journal.pcbi.1001102.t001 Table 1 ::: {.caption} ###### Parameters employed for the deterministic simulations. ::: {#pcbi-1001102-t001-1} Symbol Description Value -------- ------------------------------ ------- Membrane capacitance Leak conductance Leak reversal potential Max sodium conductance Sodium reversal potential Max potassium conductance Potassium reversal potential Kinetic parameter of gates Kinetic parameter of gates Kinetic parameter of gates Kinetic parameter of gates Kinetic parameter of gates Kinetic parameter of gates ::: Exact simulation of the microscopic models {#s2b} ------------------------------------------ Montecarlo methods represent the most commonly adopted way to simulate the random temporal evolution of ion conductances in a membrane patch, populated by a set of identical independent channels. Due to spatial proximity, channels are assumed to be coupled together by a common gating variable, such as the membrane potential or the local neurotransmitter concentration. Then, the full knowledge of the Markov kinetic scheme (see [Fig. 1A--D](#pcbi-1001102-g001){ref-type="fig"}) describing the distinct conformational states of each ion channel, as well as the transition probabilities across states, are needed [@pcbi.1001102-Liebovitch1], [@pcbi.1001102-Liebovitch2]. The kinetic scheme is employed to simulate the random transitions of the state of each individual ion channel, by repeated pseudo-random number generation (see [@pcbi.1001102-Clay1], [@pcbi.1001102-Chow1], [@pcbi.1001102-Schneidman1], [@pcbi.1001102-Skaugen1] and references therein). Although refined fast-computation techniques have been proposed [@pcbi.1001102-Chow1], we employ here a basic numerical implementation. Briefly, instead of tracking the state of each channel, the number of channels in a given state is tracked and updated at each time step (), with conditional probabilities that depend on the transition rates of the Markov scheme, as exemplified in [Fig. 1A](#pcbi-1001102-g001){ref-type="fig"}. We recall that simulating the occurrence of a random event with probability can be achieved by generating a pseudo-random number , uniformly distributed between and , and testing whether or not [@pcbi.1001102-Press1]. In the simulations reported here, we set the single-channel conductance for both sodium and potassium channels to , unless specified otherwise, and we consider a fixed channel density of and for sodium and potassium currents, respectively. In all simulations, a cylindrical single compartment was used with length and diameter equal to , unless otherwise noted. Albeit conceptually simple, these algorithms require a great amount of computational power, which increases with the number of channels that are to be simulated and with the probability of their activation. Simulation code and analysis scripts, developed in C++ and in NEURON [@pcbi.1001102-Carnevale1], are available from ModelDB [@pcbi.1001102-Hines2] at <http://senselab.med.yale.edu/modeldb> via accession number 127992. Population of two-state channels {#s2c} -------------------------------- We examine the case of a ion current whose microscopic correlate is represented by a population of ion channels. The single-channel kinetics is a 2-state scheme: *open* and *closed*, as shown in [Fig. 1A](#pcbi-1001102-g001){ref-type="fig"}. This is the simplest kinetic scheme and is often employed, for instance, for the minimal description of ionotropic AMPA-receptors [@pcbi.1001102-Destexhe1]. The symbols and in [Fig. 1A](#pcbi-1001102-g001){ref-type="fig"} represent the transition probabilities between states, expressed per time unit (i.e., as rates). They are functions of the channel gating variable(s) -- such as membrane voltage, intracellular calcium concentration, extracellular magnesium concentration, extracellular glutamate concentration, etc. [@pcbi.1001102-Hille1] -- and are experimentally identified by routine electrophysiological techniques [@pcbi.1001102-Sakmann1] and optimisation methods [@pcbi.1001102-Colquhoun1]. For the definition of our effective simulation technique for *channel noise*, we consider five realistic assumptions: (i) the channels are identical and statistically independent; (ii) for simplicity, only one conformational state is associated to a non-zero ion conductance ; (iii) is large and is known; (iv) the single-channel kinetics is described by a Markov process, where transition probabilities depend only on the current state and on the gating variable(s), and not on the previous occupancy history; and (v) the gating variables (e.g., ) change slowly, compared to the channel kinetics, with time constant [@pcbi.1001102-Johnston1]. Because of (i)--(ii), the maximal ion conductance associated to the channels can be expressed as . Then, the time-varying conductance depends only on , the fraction of channels in the *open* state:Since individual channels undergo random transitions between states [@pcbi.1001102-Sakmann1], is a non-stationary random variable, whose instantaneous value is distributed according to a binomial probability function: the number of open channels, (with constant), is a binomial random variable. As a consequence, the statistical properties of are fully specified by , the probability of occupancy of the *open* state [@pcbi.1001102-Conti1]. By assumption (iii), the binomial distribution of can be approximated by a Gauss distribution, invoking the de Moivre-Laplace (or central limit) theorem, valid when [@pcbi.1001102-Papoulis1]. By (iv), can be numerically computed as the solution of the following linear differential equation [@pcbi.1001102-Conti1], formally equivalent to the deterministic kinetic Eq. 1 [@pcbi.1001102-Destexhe2]:with and . Finally, under assumption (v), Eq. 3 can be solved analytically and is expressed as an exponential function. Under these approximations, is Gauss-distributed and completely described by its mean and by its (auto)covariance function , which at the steady-state has an exponentially decaying profile: [@pcbi.1001102-Conti1], [@pcbi.1001102-Tuckwell1]. In the theory of stochastic processes, is called a *diffusion* process, with and its steady-state variance and autocorrelation time constant, respectively [@pcbi.1001102-Cox1]. By these considerations, it follows that can be approximated and computer-simulated by an efficient method, alternative to the exact Montecarlo simulation of the discrete kinetic scheme [@pcbi.1001102-Chow1]. This method consists in generating a realisation of an Ornstein-Uhlenbeck\'s process [@pcbi.1001102-Cox1], with time-varying mean , steady-state variance , and autocorrelation time constant : where is a -correlated Gauss-process with zero mean [@pcbi.1001102-Papoulis1] (see also Eq. 20). Since [@pcbi.1001102-Conti1], [@pcbi.1001102-Tuckwell1], the deterministic component of evolves as Eq. 3, which is the familiar equation one expects by the mass-action law (i.e., Eq. 1), while interpreting as deterministic the scheme of [Fig. 1A](#pcbi-1001102-g001){ref-type="fig"} [@pcbi.1001102-Hodgkin1], [@pcbi.1001102-Yamada1]. For clarity, we rewrite such an equation as with , and . As opposed to the deterministic HH formalism however, the stochastic nature of is now explicitly captured by , algorithmically generated as a pseudo-random process by iterating the discrete-time version of Eq. 5 [@pcbi.1001102-Gillespie1], reported for the sake of completeness in Eqs. 23--24. Thus, by setting , Eqs. 4, 5, and 6 reproduce both the time-varying mean and the steady-state covariance of . More precisely, and the covariance of the term relax to the same analytical expression , after a transient of the order of . Finally, the clipping of negative conductance values for may be necessary but, if lacking, it will not affect by accumulation the numerical integration of in the present form of Eq. 4. We remark that we do not (heuristically) add a noise term in the right-hand-side of Eq. 6, as in previous Langevin-based algorithms. Instead, a precise approximation procedure is employed to statistically mimic the effect of channel noise fluctuations in . Although for 2-state channels Eqs. 4--6 are indeed equivalent to Fox\'s formulation (see the [Text S1](#pcbi.1001102.s001){ref-type="supplementary-material"}), our approach differs considerably from that by Fox as soon as multiple-state channels are considered, e.g., the sodium fast-inactivating and the potassium delayed-rectifier channels. Population of *M*-state channels {#s2d} -------------------------------- We now generalise the diffusion approximation (Eqs. 4--6) to the more general case of a large population of identical independent channels, whose single-channel dynamics is described by an *M*-state Markov scheme. Under the same assumptions (i)--(v), the probability of occupancy of the open state fully describes the fraction of open channels (see Eq. 2). However, now is a particular (say, the *k-*th) element of the probability vector of state occupancy, and each element of corresponds to a distinct state of the kinetic scheme. By assumption (iv), satisfies a system of *M* linear ordinary differential equations, which can be written in compact form asThe transition matrix is filled with the appropriate combinations of the individual transition rates between all the possible states [@pcbi.1001102-Gantmacher1]. is a vector with only one (the *k-*th) non-zero element, set to . Under assumption (v), can be computed analytically as a linear combination of a steady-state value and of *M-1* exponentials with time constants , each being the inverse of the absolute value of a non-zero eigenvalue of [@pcbi.1001102-Gantmacher1]. As for the 2-state kinetics, the statistical properties of the fraction of open channels are fully specified by and by the binomial distribution [@pcbi.1001102-Conti1]. By assumption (iii), the distribution of can be approximated by a Gauss-distribution [@pcbi.1001102-Papoulis1], and can be numerically simulated by an equivalent diffusion process. However, differently from the previous case, the steady-state covariance contains a weighted sum of *M-1* exponentials [@pcbi.1001102-Conti1], [@pcbi.1001102-Tuckwell1] and not a single term:Therefore, Eq. 4 no longer approximates , and it must be extended to a linear combination of *M-1* Ornstein-Uhlenbeck\'s independent processes , with appropriate coefficients and time constants: As for the 2-state model, . Then, one always recovers the deterministic description of the -state channels, formally coincident with Eq. 7. The derivation of the analytical expressions for and is necessary, as they depend on the values of the gating variable(s) (e.g., ), and requires the full expression of [@pcbi.1001102-Conti1], [@pcbi.1001102-Tuckwell1],which can be obtained by Laplace-transforms or linear algebraic methods [@pcbi.1001102-Brogan1]. We remark that, for our purposes, the derivation of Eq. 11 is important mainly to introduce Eqs. 8--10. Indeed, Eq. 11 considerably simplifies in the case of ion channels whose -state kinetics can be mapped into, or have been experimentally identified as, the composition of several 2-state subunits. For instance, the scheme of [Fig. 1B](#pcbi-1001102-g001){ref-type="fig"} can be mapped into the equivalent kinetic scheme shown in [Fig. 1E](#pcbi-1001102-g001){ref-type="fig"}. This is very common in the computational neuroscience literature for voltage- and ligand-gated ion channels, whose single-channel open state corresponds to the simultaneous active state of a multiple number of independent subunit types. To illustrate how Eq. 11 simplifies, we discuss a specific example where three different subunit types are present [@pcbi.1001102-Fleidervish1], [@pcbi.1001102-Yamada1], although our considerations hold for any number of different subunit types. We name these three subunit types as *m*, *h*, and *s*, and for each of them we compute the steady-state probabilities of the active state and the gating time constants, following from the solution of Eq. 3:We further assume that the overall single-channel conductance results from the composition of a given number of elements of each subunit type: say, *q*, *r*, and *w* subunits of the type *m*, *h*, and *s*, respectively. For instance, in the kinetic scheme of [Fig. 1E](#pcbi-1001102-g001){ref-type="fig"}, we have , , and . Since each subunit is described by 2-state kinetics, the total number *M* of states is . By this definition, the process is binomial and described by the joint probability that all subunits are simultaneously in their open state. Because of the statistical independence of each subunit, the joint probability is the product of elementary probabilities [@pcbi.1001102-Conti1]. Under the same assumptions of previous section, can be approximated by a diffusion stochastic process, combining deterministic and stochastic terms, as in Eq. 4. Being , we can rewrite Eq. 9 as follows: Since in this case the covariance of a product is the product of covariances, Eq. 11 reduces to [@pcbi.1001102-Conti1], [@pcbi.1001102-Tuckwell1] with , and . Expanding the powers and products of Eq. 15 and obtaining the expressions for the distinct coefficients and time constants , needed for Eqs. 9 and 10, is easier than manipulating the matrix exponential of Eq. 11. In the specific case of HH fast-inactivating sodium (i.e., , , and ) and delayed rectifier potassium channels (i.e., , and ) ([Fig. 1B--C](#pcbi-1001102-g001){ref-type="fig"}), and take the expressions reported in [Table 2](#pcbi-1001102-t002){ref-type="table"}. ::: {#pcbi-1001102-t002 .table-wrap} 10.1371/journal.pcbi.1001102.t002 Table 2 ::: {.caption} ###### Values of the coefficients and of the time constants for fast-inactivating sodium and delayed-rectifier potassium channels to be used in Eqs. 9--10. ::: {#pcbi-1001102-t002-2} Coefficient Sodium Potassium Time constant Sodium Potassium ------------- -------- ----------- --------------- -------- ----------- -- -- -- -- -- -- The steady-state symbol () was omitted for the sake of notation, from all occurrences of and . ::: Approximate reduction to a single noise term {#s2e} -------------------------------------------- In order to further gain in computational efficiency, while numerically implementing our diffusion approximation of *channel noise* (Eqs. 9--10), it is possible to reduce to one the number of required independent Ornstein-Uhlenbeck\'s stochastic processes. This additional approximation consists in interpolating the covariance of by a single decaying exponential, by replacing Eq. 9 with Eq. 4. Indeed, since Eq. 8 is the weighted sum of exponentials, one should not privilege any of those terms *a priori* and appropriately choose (in Eq. 4) and (in Eq. 5) as best-fit parameters for each value of the gating variable(s), so that Alternatively, by expanding both sides of Eq. 16 by the Taylor series, extended to the first-order (or higher), the dominant term around can be approximated by setting In investigating the impact of *channel noise* on the computational properties of single-neurons and networks, such a systematic and controlled reduction procedure should replace heuristic methods and may be extremely useful to dissect whether or not each of the terms is needed in accounting for a particular observation. The complete effective model {#s2f} ---------------------------- Following Eqs. 9--10 and [Table 2](#pcbi-1001102-t002){ref-type="table"}, we now formulate the effective stochastic model, corresponding to the deterministic HH model introduced earlier: The deterministic gating variables still obey Eq. 1, while each of the new stochastic variables ( and ) is described by Eqs. 9 and 10:where , , , and are the coefficients given in [Table 2](#pcbi-1001102-t002){ref-type="table"}, while are independent, identical, -correlated, Gauss-distributed processes with zero means and unitary variances (see Eqs. 23--24). We emphasise that the procedure leading to Eq. 18 is general and can be easily applied to more complex (single- and multi-compartmental) neuron models, which incorporate arbitrary ionic currents. The Ornstein-Uhlenbeck\'s stochastic process {#s2g} -------------------------------------------- Since the Ornstein-Uhlenbeck\'s stochastic process has been referred to repeatedly in the previous sections, we concisely review its definition and its practical numerical simulation. A realisation of this process, say , can be operatively defined as the exponential filtering of a Gauss-distributed white noise. Abusing the notation of ordinary differential equations, is the solution ofThe term represents a stationary Gauss-distributed stochastic process, which is a white-noise, fully specified by its mean and covariance . By linearity, is also Gauss-distributed [@pcbi.1001102-Papoulis1] and characterised by non-stationary mean and covariance : These quantities converge to stationary values after a time of the order of , so that at the steady-state has mean and variance equal to zero and , respectively, and an exponentially-decaying autocorrelation function, with time constant . For the purpose of obtaining independent realisations of in computer simulations, a discrete-time equivalent of Eq. 20 must be employed to generate a sequence of values . A simple iterative update formula is available,which requires the generation of a Gauss-distributed pseudo-random number at each iteration, with zero mean and unitary variance [@pcbi.1001102-Press1]. Such an iterative expression is exact, in the sense that neither needs to be uniform nor infinitesimal for to approximate the statistical properties of [@pcbi.1001102-Gillespie1]. For very small compared to , Eq. 23 can be also approximated by a first-order Taylor expansion, leading to Results {#s3} ======= In the [Materials and Methods](#s2){ref-type="sec"} section, we have motivated and operatively defined a procedure to derive an effective stochastic version for each ion current composing a conductance-based model neuron. This approximation is entirely based on probability calculus and on analytical expressions derived earlier for experimental channel-noise analysis [@pcbi.1001102-Conti1], and it does not require the Fokker-Planck formalism [@pcbi.1001102-Fox1], [@pcbi.1001102-Fox2]. We have applied here these expressions for synthetic purposes, based on the *a priori* knowledge of the Markov kinetic scheme underlying each voltage- and ligand-gated membrane conductance. The overall conductance associated to each current is modified to include the very same deterministic variables and additive noise term(s), as opposed to previous Langevin-based approaches to *channel noise* macroscopic simulation, where noise terms are (heuristically) applied to the differential equations describing activation and inactivation variables. In addition, the variance and the spectral properties of the extra noise terms are chosen accurately to reproduce the statistical properties of the corresponding microscopic model [@pcbi.1001102-Conti1]. In order to assess the validity and accuracy of our approximation procedure, we choose a single-compartmental model neuron and the fast-inactivating and delayed-rectifier sodium and potassium HH currents. We perform Montecarlo microscopic simulations of the exact full Markov model associated to each current, and compare the results to those obtained by its effective macroscopic description. First we test individual ion currents separately as in *voltage-clamp* experiments, upon clamping their gating variable , and then we study some passive and active membrane properties, as in *current-clamp* experiments. Statistical properties under *voltage-clamp* {#s3a} -------------------------------------------- We keep the membrane voltage fixed in time, while numerically simulating Eqs. 18, 19. We then study the dependence of the fraction of open channels on at the steady-state, computing mean, variance and autocorrelation time length of , . The results confirm that our effective reduction allows one to match accurately the statistical features of the microscopic models, obtained by Montecarlo simulations of the Markov-schemes. [Fig. 2](#pcbi-1001102-g002){ref-type="fig"} summarises these results for a range of clamped membrane potentials and different total numbers of ion channels. Panels A--C refer to the steady-state properties of HH potassium currents and panels D--F refer to sodium currents. In each panel, black and red markers refer to the actual numerical simulation of the microscopic and the effective models, respectively, whereas solid lines represent the theoretical steady-state values. The mean of the fraction of open channels accurately matches the theoretical predictions ( and for panels A, D - see Eqs. 13--14) and, as expected, it is independent of the number of channels . The variance inversely depends on and no difference is evident by comparing microscopic and effective simulations. The solid lines of panels B,E are obtained by plotting and (see [Table 2](#pcbi-1001102-t002){ref-type="table"}). ::: {#pcbi-1001102-g002 .fig} 10.1371/journal.pcbi.1001102.g002 Figure 2 ::: {.caption} ###### Steady-state statistical properties of the fraction of open channels , under voltage-clamp. Panels **A--C** refer to delayed-rectifier potassium channels (see [Fig. 1B](#pcbi-1001102-g001){ref-type="fig"} and [Table 2](#pcbi-1001102-t002){ref-type="table"}), whereas panels **D--F** refer to fast-inactivating sodium channels (see [Fig. 1A](#pcbi-1001102-g001){ref-type="fig"} and [Table 2](#pcbi-1001102-t002){ref-type="table"}). Black and red dots result from the simulations of the exact kinetic schemes and from our diffusion approximation, respectively. The continuous traces in **A,B,D,E** are drawn by the analytical expressions derived in the text, and refer to an increasing number of simulated channels (namely, 360, 1800, 3600). The dependence on the membrane-patch voltage is studied for the mean of (**A,D**) and for its variance (**B,E**). For an increasing number of channels, the variance decreases, as expected. Panels **C,F** show the time constant of the best-fit single-exponential, which approximates the covariance of (see Eq. 17). The mismatch between actual best-fit values and the characteristic subunit gating time-constants (, , , shown for comparison), clearly indicates that great care should be taken in deriving accurate Langevin-kind formulations. Panels **G--L** repeat the very same comparisons presented in panels **A--F**, for the Langevin-approximation introduced by Fox and coworkers (Fox, 1997; Fox and Lu, 1994): the variance of potassium currents is overestimated (**H**), whereas the variance of sodium currents is underestimated (**K**). In addition, the autocorrelation properties are not reproduced correctly (**I,L**). :::  ::: For each value of , the covariance has a decaying profile characterised by multiple time constants (see Eq. 8 and [Table 2](#pcbi-1001102-t002){ref-type="table"}). In order to represent concisely how such a decaying profile changes for distinct values of , panels C and F show (magenta curves) the values obtained by best fitting with a single exponential function the autocorrelation function of . The agreement between microscopic and effective simulations is satisfying and demonstrates that, when predicting and mimicking the autocorrelation properties of channel-noise fluctuations, the kinetic terms , , and , emerging in previous Langevin-based approaches as effective autocorrelation time constants, fail significantly. When a single Ornstein-Uhlenbeck process is used to increase the computational efficiency, the single noise term approximation given in Eqs. 16--17 turns out to be more accurate than the heuristics based on the kinetic time constants , , and or the submultiples , and (see also [Text S1](#pcbi.1001102.s001){ref-type="supplementary-material"}). In the lower part of [Fig. 2](#pcbi-1001102-g002){ref-type="fig"} (panels G--L), the same analysis is repeated, comparing the microscopic Markov-scheme simulations and the results obtained by the Langevin-based approximation proposed by Fox and coworkers [@pcbi.1001102-Fox1], [@pcbi.1001102-Fox2]. According to the mathematical expressions reported in the [Supporting Information](#s5){ref-type="sec"}, numerical simulations of the Fox\'s model show that, regardless of the number of channels, the variance of potassium currents is overestimated (panel H), whereas the variance of sodium currents is underestimated (panel K). Because of the inherent limitations of the Langevin-based approach, where a single noise term is added to the differential equations describing activation and inactivation variables, the autocorrelation properties of channel noise fluctuations (panels I,L) are mismatched. Finally, [Fig. 3](#pcbi-1001102-g003){ref-type="fig"} illustrates for the agreement between the microscopic model and our effective approximation (panels A--F), as well as the mismatch of Fox\'s algorithm (panels G--L), displaying sample time series of channel noise. Both histograms of fluctuations amplitude (panels B,E,H,K) and autocorrelation functions (panels C,F,I,L) confirm and further support the results of [Fig. 2](#pcbi-1001102-g002){ref-type="fig"}. ::: {#pcbi-1001102-g003 .fig} 10.1371/journal.pcbi.1001102.g003 Figure 3 ::: {.caption} ###### Sample time-series of the fraction of open channels , under voltage-clamp (). Panels **A--C** refer to delayed-rectifier potassium channels (see [Fig. 1B](#pcbi-1001102-g001){ref-type="fig"} and [Table 2](#pcbi-1001102-t002){ref-type="table"}), and panels **D--F** to fast-inactivating sodium channels (see [Fig. 1A](#pcbi-1001102-g001){ref-type="fig"} and [Table 2](#pcbi-1001102-t002){ref-type="table"}). Black and red traces and dots result from the simulations of the exact kinetic schemes and from our diffusion approximation, respectively. The continuous traces in **A,D** are steady-state realisations of the fraction of open potassium and open sodium channels, respectively, while panels **B,E** display the amplitude histogram. Under the conditions considered here (360 potassium and 1200 sodium channels), the Gauss-distributed effective stochastic process approximates well the microscopic model. Panels **C,F** report the autocorrelation function of channel noise fluctuations, demonstrating an excellent agreement of the effective and microscopic simulations (see also [Fig. 2C,F](#pcbi-1001102-g002){ref-type="fig"}). Panels **G--L** repeat the same comparisons presented in panels **A--F**, for the Langevin-approximation introduced by Fox and coworkers (Fox, 1997; Fox and Lu, 1994). As in [Fig. 2H,K](#pcbi-1001102-g002){ref-type="fig"} the variance of potassium currents is overestimated (**G--H**) while the variance of sodium currents is underestimated (**J--K**). In addition, the autocorrelation properties are not reproduced correctly (**I,L**). Additional simulations, for distinct values of the holding membrane potential, are provided as Supporting Information ([Figures 5](#pcbi-1001102-g005){ref-type="fig"}--10 in [Text S1](#pcbi.1001102.s001){ref-type="supplementary-material"}). :::  ::: Spontaneous action potential generation {#s3b} --------------------------------------- As the steady-state properties of the fractions of open channels are equivalent in the microscopic and effective models, we tested the full model as in a *current-clamp* experimental protocol. In this case, the gating variable is not clamped to a fixed value and both passive and active membrane properties arise by the interplay between ion currents. Once injected with a weak depolarising DC current , both the microscopic and the effective model neurons fire irregular action potentials [@pcbi.1001102-Chow1], as shown in [Fig. 4A](#pcbi-1001102-g004){ref-type="fig"}. In the absence of *channel noise* (i.e., for and ), is not strong enough to elicit spiking activity as it is below threshold for (deterministic) excitability. ::: {#pcbi-1001102-g004 .fig} 10.1371/journal.pcbi.1001102.g004 Figure 4 ::: {.caption} ###### Spontaneous firing in the microscopic and effective models. When weakly depolarising DC currents (**A**, ) are applied to both the microscopic (black sample trace) and the effective models (red sample trace), the increase in channel noise variances (see [Fig. 2C,F](#pcbi-1001102-g002){ref-type="fig"}) induces a highly irregular spontaneous emission of action potentials, with qualitatively very similar properties. In these simulations, both length and diameter of the neuron are set to , and the single channel conductance for both sodium and potassium channels is . Panels **B,C** show respectively the CV of the ISI distribution and the mean firing rate as a function of cell diameter: results are reported for the microscopic, effective and Fox\'s models (black, red and blue traces, respectively). The results of panels **B,C** refer to spontaneous activity (i.e., no injected current) with neuron length held fixed at the value . :::  ::: In order to quantify more accurately this phenomenon, we show in [Fig. 4B](#pcbi-1001102-g004){ref-type="fig"} the coefficient of variation (CV) of the interspike interval distribution obtained simulating the microscopic, effective and Fox\'s models (black, red and blue traces, respectively), for increasing values of the membrane patch area (i.e., of the number of ion channels). Note that Fox\'s model exhibits no spontaneous activity for larger cell sizes. On the other hand, the CV of the microscopic and effective models are very close. [Fig. 4C](#pcbi-1001102-g004){ref-type="fig"} shows the corresponding spontaneous mean firing rates: the presence of an "offset" in the results obtained by the effective model is evident, which is greatly reduced as the membrane area increases. This is due to the small number of channels in the membrane patch when the area is very small, against assumption (iii). Firing efficacy, latency and jitter in response to monophasic and preconditioned stimuli {#s3c} ---------------------------------------------------------------------------------------- In order to perform a direct comparison with the analysis carried out in [@pcbi.1001102-Mino4], a monophasic current pulse of fixed duration and increasing amplitude was applied 10000 times to probe the impact of channel noise on neuronal evoked responses. In [Fig. 5](#pcbi-1001102-g005){ref-type="fig"}, panel A displays the firing efficacy (i.e., the fraction of trials where a spike was elicited), panel B shows the average latency of the evoked action potential with respect to the stimulation time, and panel C displays the standard-deviation (i.e., the jitter) of the firing latency. Black and red traces and dots result from the simulations of the exact kinetic schemes and from our diffusion approximation, respectively, while in blue we indicate the results from the simulation of the Langevin-approximation introduced by Fox. The satisfactory agreement between microscopic and effective models is apparent, whereas simulations according to Fox\'s algorithm differ considerably. Panel D shows the distribution of spike occurrence times, evoked by a biphasic stimulus over 10000 trials. The distributions of spike times obtained by the microscopic and effective models almost overlap, while Fox\'s distribution has a significantly different shape. ::: {#pcbi-1001102-g005 .fig} 10.1371/journal.pcbi.1001102.g005 Figure 5 ::: {.caption} ###### Comparison of firing efficacy, latency and jitter of a sharp current pulse. Panels **A**, **B** and **C** display the firing efficacy, the average latency and the jitter of the evoked responses, respectively, after the application of a monophasic stimulus of duration repeated for 10000 trials. Black and red traces and dots result from the simulations of the exact kinetic schemes and from our diffusion approximation, and in blue we indicate the results from the simulation of the Langevin-approximation introduced by Fox. Panel **D** shows the distribution of spike occurrence times, evoked by a biphasic stimulus over 10000 trials: the duration and amplitude of the preconditioning part are and , respectively, the duration and amplitude of the second part are and . In all panels, the neuron is simulated as a single cylindrical compartment of length and diameter equal to and single channel conductances equal to , for both sodium and potassium channels. The integration time step was set to . :::  ::: The results we present here for the microscopic and Fox\'s models are in close agreement with those discussed in greater detail in [@pcbi.1001102-Mino4]. Reliability of evoked spike timing and response latency {#s3d} ------------------------------------------------------- The results shown in [Fig. 5](#pcbi-1001102-g005){ref-type="fig"} refer to the application of either a mono- or biphasic stimulus of *short* duration, in the order of milliseconds. Here, we extend the previous analysis to the case of significantly longer stimulations: our objective is to study the so-called *reliability* of spike timing along the lines of the experimental protocol defined in [@pcbi.1001102-Mainen1]. It is well known that, as a consequence of *channel noise*, the reliability of evoked spike timing is higher for current stimuli fluctuating in time than for DC current pulses [@pcbi.1001102-Mainen1], [@pcbi.1001102-Schneidman1], [@pcbi.1001102-White1]. Indeed, larger fluctuations induced in the membrane potential by the driving stimulus transiently hyperpolarise the cell, thus reducing the variance of *channel noise* (see [Fig. 2B,E](#pcbi-1001102-g002){ref-type="fig"}). A similar phenomenon has been described in the case of inhibitory autapses in the cerebral cortex [@pcbi.1001102-Bacci1] and it could also be represented at microcircuit-level by the role of disynaptic inhibition [@pcbi.1001102-Silberberg1]. A single-compartmental model simulation incorporating *channel noise* can replicate this effect [@pcbi.1001102-Schneidman1] and constitutes a further benchmark to compare microscopic and effective models. We note that for this analysis, we have chosen the neuron parameters in order to reproduce the results presented in [@pcbi.1001102-Mainen1]. The agreement between models is very good as shown in [Fig. 6](#pcbi-1001102-g006){ref-type="fig"}, where black (red) traces and markers refer to the microscopic (effective) model. The spike responses to two repeated identical stimuli were considered: a DC pulse (panel A) and a realisation of an exponentially-filtered white noise (panel B). The raster diagrams of the spike times (upper plots), as well as the corresponding time histograms (lower plots), demonstrate that the two models perform in the same way as the spread and latency of the spike times, in response to the repeated identical stimulation, are practically identical. Finally, a quantitative measure of both precision and reliability (computed according to [@pcbi.1001102-Mainen1]) provides values similar to those measured in *in vitro* experiments (see figure caption). ::: {#pcbi-1001102-g006 .fig} 10.1371/journal.pcbi.1001102.g006 Figure 6 ::: {.caption} ###### Raster plots and peristimulus time histograms (PSTH) for the timing of spiking responses to repeated identical DC pulses (A) and fluctuating currents (B). Red traces and markers refer to Montecarlo microscopic simulations of the full model, while black traces and markers refer to the effective model. The values of reliability () and precision () are in accordance with those measured in *in vitro* experiments. In particular, in panel A: , for the microscopic model, , for the effective model. Panel B: , for the microscopic model, , for the effective model. The DC pulse has an amplitude of , whereas the noisy stimulus is the same realisation of an Ornstein-Uhlenbeck\'s process, with mean and standard deviation set to , and with autocorrelation time length set to . :::  ::: Frequency-current () response curves {#s3e} ------------------------------------ For stronger depolarising DC currents , the firing of both the microscopic and the effective models becomes more regular. The mean firing rate, as a function of was studied to test the agreement between their evoked response properties. [Fig. 7](#pcbi-1001102-g007){ref-type="fig"} shows the curves computed over -long evoked spike-trains. For each current amplitude, the simulation was repeated 10 times, and firing rates obtained in each repetition were averaged. Error bars indicate the standard deviation of the firing rate across repetitions. Responses of both the microscopic and the effective models result in almost identical variability across repetitions and in both cases the type-II behaviour, typical of the deterministic HH model, fades away. This is a known consequence of the presence of *channel noise*, which smooths what would be an abrupt transition from a quiescent to a spiking regime. These irregular transitions occur for both models in the very same range of input currents (green-shaded region in the figure), where the membrane potential repeatedly switches between a resting equilibrium point and a spiking limit cycle (see [@pcbi.1001102-Schneidman1] for an extended discussion). ::: {#pcbi-1001102-g007 .fig} 10.1371/journal.pcbi.1001102.g007 Figure 7 ::: {.caption} ###### Frequency-current () response curves. Mean firing rate, in response to a DC current injection, studied for increasing stimulus intensities in both Montecarlo microscopic (black trace) and effective model (red trace) simulations. Single-channel conductance for both sodium and potassium channels set to . :::  ::: Power-spectral density of membrane voltage fluctuations {#s3f} ------------------------------------------------------- We finally compare the power-spectral densities of subthreshold membrane potential trajectories, obtained in simulations of the microscopic and effective models. We followed closely the numerical analysis of [@pcbi.1001102-Steinmetz1], where a comparison between the microscopic model and a quasi-active linearised model with phenomenological inductances was instead presented. Once more, the agreement between the two models is satisfactory: in [Fig. 8](#pcbi-1001102-g008){ref-type="fig"} we show the results, indicating by thick shaded curves the power spectra computed from the microscopic model, and by thin solid lines the power spectra computed from the effective model. The agreement is good over the entire frequency domain, reproducing some of the features that have been experimentally measured in cortical neurons and related to *channel noise* [@pcbi.1001102-Jacobson1]. ::: {#pcbi-1001102-g008 .fig} 10.1371/journal.pcbi.1001102.g008 Figure 8 ::: {.caption} ###### Voltage power spectral densities of subthreshold membrane potential trajectories. Comparison between the microscopic (thick shaded lines) and the effective (thin solid lines) models. of simulated recordings of the membrane potential were obtained under weak holding currents ({), resulting in membrane potential traces fluctuating around an offset ({). Rare spontaneous spikes were removed from the analysis, excluding the preceding and the following each spike. The spectra have been obtained by applying the Welch method, on moving windows of duration and overlapping by , and subsequently averaging the results. :::  ::: Discussion {#s4} ========== In this paper, we introduced the systematic generalisation and improvement of previous Langevin-based channel-noise effective simulation techniques. By the diffusion approximation of ion channels population dynamics, we aimed at efficient and accurate computer simulation of *channel noise*. Our method approximates correctly the statistical properties of individual ion conductances (their mean and autocovariance function), matching those emerging from the Montecarlo simulation of their corresponding Markov schemes. In addition, both passive and active properties of neuron model simulations are replicated with satisfying accuracy. While simulating of model time by a conventional Montecarlo algorithm takes about 22 hours for completion, the same simulation with very similar statistical features is replicated by the effective model in only 124 seconds, on a machine equipped with a Intel Core i7, with of RAM, running Ubuntu Linux 9.10. When relating the computation times to the benchmarking provided by [@pcbi.1001102-Mino4], our diffusion approximation is only 1.5 times slower than Fox\'s algorithm and therefore more than 4.5 times faster than the fastest available algorithm for exact microscopic simulations [@pcbi.1001102-Chow1]. Our results have been obtained by custom C++ and NEURON model simulations (see the [Materials and Methods](#s2){ref-type="sec"} section), but the implementation of the method in other languages (MATLAB, Python) or other simulation environments (Genesis, NEST, Brian) is straightforward. Besides the speed increase, the value of our contribution is threefold: i) mean, variance and spectral properties of fluctuations induced by the stochasticity of individual ion currents are correctly approximated, regardless of the number of channels; ii) our method is presented operatively, allowing any deterministic neuron model, whose ion conductance kinetics is described by a Markov scheme, to be quickly converted into an equivalent stochastic version without involving any heuristics on the choice of the parameters for extra noise sources; iii) the underlying assumptions for the validity of our approximation are also indicated with full details. The earlier proposals of [@pcbi.1001102-Fox1], [@pcbi.1001102-Fox2], recently challenged for their accuracy, are indeed very similar to our method, although focused only on the HH model. In these papers, the equations that state variables , , and obey to are modified by adding a single noise term , as follows:where and is a Gauss-distributed noise term with zero mean and covariance given by By direct inspection and comparison of Eqs. 4, 5, 6, and 20, it is possible to show that Eq. 25 and Eqs. 4--5 are equivalent (see [Text S1](#pcbi.1001102.s001){ref-type="supplementary-material"}). In other words, for 2-state kinetics the approximation given by Eq. 25 is correct but fails when the powers , , and are computed and when they are combined in the product . Under these circumstances, mean, variance and covariance function indeed deviate considerably from the correct dependence on , emerging from the microscopic simulations or computed analytically (see [Text S1](#pcbi.1001102.s001){ref-type="supplementary-material"}). Briefly, the potassium current simulated by the fourth power overestimates the correct variance, does not share the correct mean and has qualitatively different autocorrelation properties. The sodium current simulated by the third power and the product by instead underestimates the correct variance, does not share the correct mean and has quantitatively different autocorrelation properties. The interested reader can find all the details in the [Supporting Information](#s5){ref-type="sec"}. We believe that the reason for the success of our approximation, compared to Fox\'s approach, lies not only in the correct agreement of fluctuations mean and variance, validated by direct comparison with the theoretical and numerical results of the *microscopic* description [@pcbi.1001102-Bruce2], but also in the fact that the covariance function of those fluctuations must be precisely matched and should be approximated by a sum of white-noise terms and not by adding noisy terms to the deterministic kinetic equations for activation and inactivation variables. However, we note that under current-clamp condition, there is no *a priori* guarantee that any Langevin-based approach, including our diffusion approximation, works faithfully [@pcbi.1001102-White2]. In fact, our assumption (v), that the gating variable (e.g., ) changes slowly compared to channel kinetics, may not be *instantaneously* satisfied during very fast transients. Although the same condition is anyway employed for obtaining numerical speed-up in deterministic conductance-based models [@pcbi.1001102-Johnston1], [@pcbi.1001102-Moore1], [@pcbi.1001102-Mascagni1], the instantaneous channel noise fluctuations might lag behind what predicted by microscopic exact models (see Figs. 11--12 in [Text S1](#pcbi.1001102.s001){ref-type="supplementary-material"}). Nevertheless, owing to the satisfying results we obtained in terms of firing-rate properties, firing time reliability, precision, efficacy, latency, jitter as well as subthreshold membrane fluctuations, we speculate that inaccuracies during very fast transients might still be compatible with accurate model performances (perhaps due to the low-pass properties of the membrane), provided that first- and second-order voltage-clamp statistics are correctly matched. A very similar reduction procedure is implicitly mentioned in [@pcbi.1001102-Steinmetz1], where the authors developed a quasi-active membrane potential equation employed only for the spectral analysis of subthreshold voltage noise, but not for its actual numerical simulations. The authors state clearly that their approximation can be viewed as a linearised approximation of the Fokker-Planck master equation [@pcbi.1001102-Fox2]. As opposed to our method, which requires adding multiplicative noise terms to the membrane potential equation, their quasi-active model includes only additive noise, upon linearisation, resulting in the definition of electrical circuit analogs (capacitances and inductors) useful for the intuitive understanding of *channel noise* for subthreshold passive membrane properties, and for the analytical prediction of the spectral properties of membrane potential fluctuations. The authors, however, do not explicitly provide any derivation of their approach and do not test it for the excitable response neuronal properties as a replacement of microscopic simulations. One further approach to *channel noise* modelling has been proposed in [@pcbi.1001102-Saarinen1]. We share the motivation of performing accurate and fast simulation by a Langevin-based approach, but we use stochastic processes with precise and defined statistical properties, coincident with those emerging from the microscopic description of the stochastic behaviour of channels. In the proposal by [@pcbi.1001102-Saarinen1], the effective stochastic term is modelled as Brownian motion, i.e., as a Gauss-distributed process with independent increments and heuristically fixed constant variance, ignoring its voltage-dependence and the variety of autocorrelation time constants. Since the analytical derivation of the accurate statistical properties of *channel noise* is possible, and its implementation straightforward as we showed here, there is no need to use arbitrary parameters for simulating the stochastic components of ion currents gating. It is worth mentioning that *population density* approaches, proposed for integrate-and-fire as well as conductance-based models [@pcbi.1001102-Knight1]--[@pcbi.1001102-Chizhov1], share to some extent the motivations of our work: exploring the impact of endogenous or exogenous noise sources while developing tools to capture or effectively simulate population-level dynamics [@pcbi.1001102-Nykamp1], [@pcbi.1001102-Fourcaud1]. Those works also aim at correctly mimicking actual network interactions in terms of an equivalent stochastic additive input to a generic unit of the network [@pcbi.1001102-LaCamera1], as in the *mean-field* approximation of synaptic interactions [@pcbi.1001102-Treves1]. Since our work provides an accurate effective description of an intrinsic (multiplicative) noise source, our formulation could be very relevant for those approaches, in extending population density descriptions to incorporate endogenous channel noise. In conclusion, we believe that our method could open new possibilities for the investigations of *channel noise* impact in morphologically detailed conductance-based model neurons, as well as in large networks models, where realism cannot be compromised by computational parsimony. Spike timing computation in neural networks [@pcbi.1001102-Karmarkar1] with specific microcircuit architectures [@pcbi.1001102-Silberberg1] might be for instance easily complemented by stochastic components of neural excitability, employing detailed neuron models. Finally, the possibility of further increasing the level of approximation, involving only a modification of the spectral properties of *channel noise* without affecting the accuracy of its variance, may lead to an in depth understanding of what temporal correlation properties are relevant for specific computational neuronal properties and how *channel noise* interacts with other noise sources. Supporting Information {#s5} ====================== Text S1 ::: {.caption} ###### This supporting information reviews a few results of the theory of stochastic processes, useful for supporting our discussion and for the comparison between Fox\'s and our method. It also contains Figures where extended numerical comparisons between Fox\'s and our method are presented. (2.06 MB PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: We are grateful to Dr. I. Segev, Dr. E. Vasilaki, and V. Delattre for discussions, to Dr. S. Martinoia for comments on an earlier version of this manuscript, and to Dr. H. Mino for sharing his code for Montecarlo simulations. The authors have declared that no competing interests exist. This work was supported by the University of Genoa (<http://www.unige.it>), the Fondation Francqui (<http://www.francquifoundation.be>), the Belgian Interuniversity Attraction Pole (grant n. IUAP P6/29, <http://www.belspo.be>), the University of Antwerp (Nieuw Onderzoeks Initiatief, Bijzonder Onderzoeks Fonds, NOI-BOF2009, <http://www.ua.ac.be>), the Flanders Research Foundation (grants n. G.0836.09 and n.G.0244.08, <http://www.fwo.be>), and the Royal Society (JP091330-2009/R4, <http://royalsociety.org>). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: DL MG. Performed the experiments: DL. Analyzed the data: DL MG. Wrote the paper: DL MS MG.

PubMed Central

2024-06-05T04:04:19.658587

2011-3-10

PMC3053315

Introduction {#s1} ============ Rheumatoid arthritis (RA) is a common autoimmune disease that is characterized by inflammation and destruction of the joints [@pcbi.1001105-Scott1]. Inflammation is caused by the infiltration of inflammatory cells, including neutrophils, macrophages, B- and T-cells into the normally acellular synovial tissue that lines the junction of bones. A key feature of the inflammatory process is the release by the infiltrating cells of proinflammatory cytokines, including TNF-α, interleukin-1 (IL-1), interleukin-6 (IL-6), and others. These cytokines promote further inflammation and joint destruction by activating infiltrating immune cells as well as resident bone cells and promoting the release of degradative enzymes such as matrix metalloproteases and cathepsins. Another key feature of the disease is the presence of autoantibodies directed against citrullinated proteins and other targets that contribute to joint damage as well as systemic manifestations of RA [@pcbi.1001105-Firestein1]. Like other autoimmune diseases, RA is caused by complex interactions between genes and environment [@pcbi.1001105-MullerLadner1]. RA is treated with small molecule, disease-modifying anti-rheumatic drugs (DMARDs) including Methotrexate, Sulfasalazine, Leflunomide and others [@pcbi.1001105-Matsumoto1]. Biologic agents include several tumor necrosis factor alpha (TNF-α) blockers such as Etanercept, Infliximab and Adalimumab, co-stimulation blockers (abatacept or CTLA4-Ig) and B-cell depleters (rituximab). DMARDs are often combined with tumor necrosis factor alpha (TNF-α) blockade. For many patients, TNF-α blockade effectively relieves arthritis symptoms as measured by American College of Rheumatology (ACR) or Disease Activity Score (DAS) scoring systems that measure numbers of tender and swollen joints as well as other clinical parameters. Typically 39% of patients score better than ACR 50 in etanercept trials when dosed at 10 mg or 25 mg twice weekly, and 64% of patients scored better than ACR 20 [@pcbi.1001105-Bathon1]. However, for particular subsets of rheumatoid arthritis patients, TNF-α blockade does not appear to relieve the symptoms of RA. A genetic component to RA has been established from twin and family studies. The estimated heritability of RA is about 60%, and the genetic basis is complex with consistent association of *HLA, PTPN22, TRAF1-C5*, and several other loci [@pcbi.1001105-Raychaudhuri1]. Some of these genetic variants have also been associated with differential response to treatment with TNF-α blockers [@pcbi.1001105-Cui1]. However, establishing the causal molecular mechanisms by which genetic variants affect RA phenotypes or differential response to TNF-α blocker to inform rational selection of molecular intervention targets for RA remains a challenging problem. Establishing causal mechanisms, particularly in clinical data, is a difficult exercise and is often assessed using two popular approaches [@pcbi.1001105-Russo1]. One approach uses probabilistic evidence from cross-sectional population studies and discovers new stable statistical associations between the measurements. However, discovering statistical links between the measurements alone cannot determine the causal direction. Another uses mechanistic evidence arising from knowledge of an existing physical property and establishes a predictable dependency over time. However, discovering new mechanistic knowledge demands experimentation and leads directly to the first approach of statistical, probabilistic analysis of the collected data. Probabilistic or mechanistic causality models alone appear to be at once insufficient and irreconcilable. Scientific inferences from approaches that are unable to effectively reconcile these two notions of causality typically experience delays in their acceptance. There have been many instances of strong probabilistic links in clinical data that have not been accepted until the mechanism had been discovered. Rigorous clinical science correctly requires that both probabilistic and mechanistic arguments need to be met simultaneously before a particular claim can be accepted as causal [@pcbi.1001105-Russo1]. The scientific method provides a framework to address both aspects of causality simultaneously [@pcbi.1001105-Russo1]. First, existing evidence is used to propose a constrained *rational mechanism*. Second, controlled experiments or *perturbations* are designed to test the mechanism, and relevant data are collected on the proposed mechanistic molecules. Finally, an appropriate systematization of the collected data *deduces probabilistic reflections* of known mechanisms and *infers* new aspects of the mechanism that can be directly tested. If the inferred aspects are confirmed, the proposed mechanism is accepted as causal. To develop a causal data analysis approach to rational drug target and biomarker discovery in RA, we used published data from the Autoimmune Biomarkers Collaborative Network (ABCoN) [@pcbi.1001105-Liu1]. The ABCoN recruited more than 100 patients naïve to anti-TNF treatment for systematic clinical and molecular analysis. Clinical data for calculating Disease Activity Score 28 (DAS28), and blood samples for genetic (SNP) and expression profiling analysis, were collected at baseline (pretreatment), as well as 6 weeks and 14 weeks after starting therapy for one of three anti-TNF molecules (Etanercept, Infliximab or Adalimumab). In our analysis of ABCoN data, we assume a rational mechanism that genetic variation arising from meiosis in this study population together with drug therapy are systematic perturbations that impact RA through multiple molecular and physiological interactions that are probabilistically reflected in the blood transcription profile data. Systematization of the gene expression and clinical data enable us to distinguish transcripts that play a role in modulating RA phenotypes from those that are either simply correlated with or secondary to the phenotypes. Bayesian networks provide a convenient framework for systematizing data to deduce probabilistic orderings and modeling large systems of interacting variables [@pcbi.1001105-Friedman1], [@pcbi.1001105-Pearl1], [@pcbi.1001105-Sachs1], [@pcbi.1001105-Madigan1], [@pcbi.1001105-Friedman2]. Most previous studies have concentrated on either estimating the structural connections in the system under study or on the identification of disease associated genetic polymorphisms. Simulation of Bayesian networks can be used to predict the effect of specific interventions. Whereas some previous studies make predictions based on a single network topology [@pcbi.1001105-ChaibubNeto1], [@pcbi.1001105-Chen1], [@pcbi.1001105-Schadt1], [@pcbi.1001105-Zhu1], [@pcbi.1001105-Zhu2], our approach adds to these in two important ways. First it generates a statistical sample, or ensemble, of network structures that are consistent with data collected [@pcbi.1001105-Peer1], [@pcbi.1001105-Peer2]. Second, it enables quantitative prediction of the effects of perturbations [@pcbi.1001105-Penny1] that account for uncertainty about network topology. With the ability to predict the impact of specific interventions that were not part of the collected data in defined study subjects, the ensemble organizes the data into a rational model and also predicts unseen events. In this regard, these simulation-ready integrative genomic ensembles capture the essence of the scientific method described previously. Here we concentrate on the prediction of gene expression levels that are critical to explaining the number of swollen joints (SJ), the number of tender joints (TJ), the amount of pain and the plasma concentration of C-reactive protein (CRP) in the subjects enrolled in the ABCoN trial with and without TNF-α blocker treatment. Sets of biomarkers predictive of TNF-α blocker response have been identified [@pcbi.1001105-Bienkowska1], [@pcbi.1001105-Julia1]. This should serve as an important diagnostic tool to help determine who would or would not benefit from a TNF-α blocker therapy. While there are treatment options for those who do not respond to TNF-α blocker therapies, identifying those patients before they begin biologics or early in their treatment would help rationalize their treatment. Additionally, uncovering molecular alterations underlying TNF-α blocker response will be critical to discovering new and effective therapies for RA. Results {#s2} ======= Using ABCoN study data to explore mechanisms of differential response to TNF-α blockade {#s2a} --------------------------------------------------------------------------------------- While DMARDs, TNF-α blockers and other treatments are available for RA patients [@pcbi.1001105-Scott1], [@pcbi.1001105-Buch1], there is still a need to identify new drug targets for patients whose disease does not respond to available therapies. If the molecular mechanism for TNF-α blocker failure were understood, the knowledge would allow drug researchers to more effectively select molecules that could target this particular subgroup of patients. Simulation-capable ensemble network models of cause and effect are theoretically capable of providing clues to the reasons some patients do not respond. The ideal strategy to learn the probabilistic mechanisms of non-response would be to integrate circulating drug concentrations (pharmacokinetic data), measures of the effectiveness of the drug (pharmacodynamic data), genetic variation, relevant molecular measures from the disease-affected tissues and finally, components of DAS28 into a single network ensemble model. Differences in all these measures would lead to an ensemble model that could effectively recover TNF-α blockade response in genetically defined subjects. To recover non-response targets, molecular measures for non-responding subjects would be set in the ensemble and systematic *in silico* perturbations for each non-responding subject would be completed to give a set of intervention points for these particular subjects. For alternative targets to TNF-α in non-responding subjects, the pharmacokinetic variables would be set to mimic a TNF-α blocker concentration of zero. If the focus were to find combination therapies, the simulations would be completed where the pharmacokinetic variables mimicked realistic circulating doses. For the ABCoN study, there are no pharmacokinetic or pharmacodynamic measures, such as phospho-proteomic data available to enable the modeling strategy outlined. In addition, and as seen in this data set and other studies [@pcbi.1001105-Ebert1], TNF-α transcript levels vary so little in whole blood that they cannot be considered as a surrogate for pharmacokinetic or pharmacodynamic measures. This means that the data cannot be used to effectively understand the probabilistic mechanisms of response, or non-response, to TNF-α blocker therapy. However, while a mechanistic analysis is difficult within the current study design, we can investigate transcription variation with and without an undefined amount of circulating TNF-α blocker activity with the expectation that models built using data collected before TNF-α blocker therapy will be informative for untreated RA and mainly reflect TNF-α dependent mechanisms. Additionally, models built using data collected from treated subjects are expected to reveal aspects of RA that are still important even after TNF-α signaling has been interdicted. Critical transcripts identified from the network ensemble built from data collected after treatment could represent both starting points for drug target identification for subjects who will not respond and combination therapy targets that are only important when TNF-α blocker is circulating in the patient. Reverse-Engineer and Forward-Simulate (REFS) predictive framework {#s2b} ----------------------------------------------------------------- Our method works by considering all possible associations between DNA variation, gene expression, and RA clinical data, resulting in a collection of network fragments derived from these measurements and reflecting not only associations between the traits, but how the different variables are causally associated as well. Each network fragment defines a quantitative, continuous relationship among all possible sets of 3 or fewer measured molecular variables. Because of the experimental design, where DNA variation is leveraged as the randomization mechanism needed to make causal inferences, these fragments approximate stable, probabilistic cause and effect relationships [@pcbi.1001105-Pearl1]. The relationship is supported by a Bayesian probabilistic score that deduces how likely the candidate relationship is given the measurements and also penalizes the relationship for its mathematical complexity ([Figure 1B](#pcbi-1001105-g001){ref-type="fig"}). By exhaustively scoring all of the possible pairwise and three-way relationships inferred from the DNA, expression, and RA clinical data, the most likely fragments can be identified and held aside preferentially in the collection for later use. Network fragments that include SNPs are constrained such that the genotypes are always upstream of gene expression or clinical outcome data, reflecting the assumption that genotypes are systematic perturbations to the disease. However, network fragments that are comprised of gene expression and clinical outcome data alone could affect each other through multiple causal mechanisms, and therefore we considered all possible orderings of three or fewer variables. In addition to assessing the probability of a particular relationship, the quantitative parameters of the relationship are computed and stored, an important departure from previous methods that have primarily focused on structure. Rather than discarding this quantitative information, we store it so that it can be queried to draw more complex inferences later on in our process. In the second phase, we estimate an ensemble of network models based on data from the integrative genomics experiment. A statistical definition of the ensemble is a sample of networks drawn from all possible networks consistent with the data, and whose properties are robust, regardless of either the actual structures contained within it or the algorithmic starting conditions that generated it. Even though a normally distributed random variable can take on infinitely many values, sampling even as few as 30 numbers from the distribution can provide reliable estimates of the mean and standard deviation of the distribution, thereby perfectly characterizing the behavior of the random variable. In much the same way, our approach samples from the space of all possible networks to approximate the distribution of that space, enabling an effective characterization of that space. In the third phase we use forward simulation of networks in the ensemble to generate predictions of the effects of perturbations. The entire process is summarized in [Figure 1](#pcbi-1001105-g001){ref-type="fig"}. The entire process is repeated for data collected before and 14 weeks after TNF-α blocker therapy, giving two distinct network ensembles that capture untreated and treated aspects of RA mechanism. ::: {#pcbi-1001105-g001 .fig} 10.1371/journal.pcbi.1001105.g001 Figure 1 ::: {.caption} ###### Schematic of Bayesian network reverse engineering and Monte Carlo simulation. A. Genetic, gene expression and phenotypic data are prepared for modeling by formatting followed by investigation to select the appropriate data transformation for the particular data type. Confounding factors and other explanatory variables are considered and modeled if appropriate. B. Likely fragments for network reconstruction are identified by scoring all 2-, 3- and 4-variable combinations with the constraint that SNPs are causally upstream. There are too many scored combinations to consider all during network reconstruction. The fragments that had the most likely Bayesian scores for each individual variable were identified and retained for network reconstruction. C. Parallel global network sampling constructs an ensemble of 1024 network structures that explain the data. The probabilistic directionality computed by the Bayesian framework allows inferences to be made about what lies upstream and downstream of particular phenotypic variables D. Diversity in network structures identified during network reconstruction captures uncertainty in the model. Hypotheses are extracted from the network ensemble by completing Monte Carlo simulations of "what-if" scenarios. Down-regulating the blue transcript would be expected to impact both TJ and SJ, while leaving CRP unchanged. The change in these phenotypic parameters would further predict *reactive* transcription changes in the liver. The separation of upstream and downstream components can identify potential intervention points and markers. :::  ::: Summaries of data processing, diagrams of the consensus structure of the two models with respect to the DAS28 components ([Figure 2](#pcbi-1001105-g002){ref-type="fig"}) and the mathematical assessment of the quality of the models generated by the framework are provided ([Table 1](#pcbi-1001105-t001){ref-type="table"} in [Text S1](#pcbi.1001105.s007){ref-type="supplementary-material"} and [Figures S1](#pcbi.1001105.s001){ref-type="supplementary-material"}--[S3](#pcbi.1001105.s003){ref-type="supplementary-material"}). Briefly, the untreated network ensemble was built using 6,075 SNPs, which included imputed genotypes for previously identified SNPs associated with RA, 4,794 gene expression values and 4 DAS28 component scores. The treated network ensemble was built using 6,076 SNPs, 4,512 gene expression values and 4 DAS28 component scores. The network sample properties were assessed by completing three independent network sampling procedures using different, random starting conditions. There were no significant differences in the sample properties regardless of the starting conditions demonstrating that the sampling procedure had converged. The predictive properties of the network sample were assessed using simulations designed to measure how accurate the network ensemble recovered the actual observed data ([Text S1](#pcbi.1001105.s007){ref-type="supplementary-material"}). ::: {#pcbi-1001105-g002 .fig} 10.1371/journal.pcbi.1001105.g002 Figure 2 ::: {.caption} ###### Consensus topology of network ensembles for pre-treated and TNF-α blocker treated data. A. Snapshot of the network ensemble at 2.5% consensus topology generated from pre-treated subjects. Pain does not appear to be controlled by measures extracted from whole blood mRNA profiling in pre-treated subjects. B. Snapshot of the network consensus topology at 2.5% consensus generated from TNF-α blocker treated data. :::  ::: ::: {#pcbi-1001105-t001 .table-wrap} 10.1371/journal.pcbi.1001105.t001 Table 1 ::: {.caption} ###### Category 1 Transcripts from the post-treatment model. ::: {#pcbi-1001105-t001-1} Gene Symbol Name TJ SJ DAS28 ------------- --------------------------------------------------------------------------------------------------- ---- ---- ------- RAP2C RAP2C, member of RAS oncogene family X X X ANXA1 Annexin A1 X X X GON4L gon-4-like (C. elegans) X X X CPVL carboxypeptidase, vitellogenic-like X X X SMARCD2 SWI/SNF related, matrix associated, actin dependent regulator of chromatin, subfamily d, member 2 X X X SLC6A6 solute carrier family 6 (neurotransmitter transporter, taurine), member 6 X X X EID1 EP300 interacting inhibitor of differentiation 1 X X X MALAT1 metastasis associated lung adenocarcinoma transcript 1 (non-coding RNA) X X X EIF3D eukaryotic translation initiation factor 3, subunit D X X CHCHD2 coiled-coil-helix-coiled-coil-helix domain containing 2 X X NDE1 nudE nuclear distribution gene E homolog 1 (A. nidulans) X X NCOA1 nuclear receptor coactivator 1 X X ARHGAP25 Rho GTPase activating protein 25 X X MBP Myelin basic protein X X JMJD3 jumonji domain containing 3 X X In [Tables 1](#pcbi-1001105-t001){ref-type="table"} and [2](#pcbi-1001105-t002){ref-type="table"}, these transcripts have a predicted novel role in either untreated or treated RA, are not directly related to TNFα biology and have statistically significant impacts on DAS28 or joint health (denoted by X, *p*\<0.05). ::: Model intervention simulations -- Virtual clinical trials {#s2c} --------------------------------------------------------- Systematic *in silico* simulations of 10-fold knockdown of all 9,306 transcripts were completed to provide quantitative predictions of how modulation of a particular gene expression measure would affect the DAS28 score in every particular subject given their own, individualized genotype and gene expression values before and after treatment with TNF-α blocker. Simulated distributions of SJ, TJ, Pain and CRP were compiled from 30 replicate gene expression simulations for each subject and for each gene. A single predicted DAS28 component score for each individual patient was then estimated as the median of the 30 replicate simulations to provide a robust point estimate of the range of predicted values. The simulated and summarized SJ, TJ, Pain and CRP scores were used to estimate a simulated DAS28 score using the standard equation supplied to clinicians. Gene expression perturbations were ranked by their ability to sufficiently modulate RA clinical measures in a significant number of patients using either a χ^2^ test for TJ and SJ scores or a Student\'s t-test for Pain and CRP with respect to simulated untreated gene expression DAS28 scores. Identification of three classes of causal transcripts {#s2d} ----------------------------------------------------- 78 transcripts from the untreated network ensemble (see [Table S1](#pcbi.1001105.s005){ref-type="supplementary-material"}) and 97 transcripts from the TNF-α blocker treated network ensemble (see [Table S2](#pcbi.1001105.s006){ref-type="supplementary-material"}) were predicted to significantly modulate any of the DAS28 component scores (*p*\<0.05). The transcripts identified can be assigned to one of three broad categories based on their predicted efficacies and druggability, which was assessed using Ingenuity Pathway Analysis (Ingenuity Systems, [www.ingenuity.com](http://www.ingenuity.com)). Category 1 transcripts are defined as transcripts that have a predicted novel role in RA ([Tables 1](#pcbi-1001105-t001){ref-type="table"} and [2](#pcbi-1001105-t002){ref-type="table"}), are not directly related to TNF-α biology and have statistically significant impacts on DAS28 and joint health. Category 2 transcripts are potential alternatives to TNF-α therapies and have known dependencies on TNF-α or are proteins that could be targeted with a small molecule and are predicted to modulate DAS28 and joint health ([Tables 3](#pcbi-1001105-t003){ref-type="table"} and [4](#pcbi-1001105-t004){ref-type="table"}). Category 3 transcripts are those that are predicted to impact joint health but not DAS28 across the subjects simulated ([Tables 5](#pcbi-1001105-t005){ref-type="table"} and [6](#pcbi-1001105-t006){ref-type="table"}). ::: {#pcbi-1001105-t002 .table-wrap} 10.1371/journal.pcbi.1001105.t002 Table 2 ::: {.caption} ###### Category 1 Transcripts from the pre-treatment model. ::: {#pcbi-1001105-t002-2} Gene Symbol Name TJ SJ DAS28 ------------- ------------------------------------------- ---- ---- ------- DOK3 Docking protein 3 X X C15orf39 Chromosome 15 orf 39 X X FLOT2 Flotillin 2 X X RHOG Ras homolog gene family, member G (rhoG) X X NBPF1 neuroblastoma breakpoint family, member 1 X X GLT1D1 glycosyltransferase 1 domain containing 1 X X FLJ43663 hypothetical protein FLJ43663 X X ::: ::: {#pcbi-1001105-t003 .table-wrap} 10.1371/journal.pcbi.1001105.t003 Table 3 ::: {.caption} ###### Category 2 Transcripts from the post-treatment model. ::: {#pcbi-1001105-t003-3} Gene Symbol Name TJ SJ DAS28 ------------- ------------------------------------------------------------ ---- ---- ------- TRAF3IP3 TRAF3 interacting protein 3 X X PSMD4 proteasome (prosome, macropain) 26S subunit, non-ATPase, 4 X X CD86 CD86 molecule X X X In [Tables 3](#pcbi-1001105-t003){ref-type="table"} and [4](#pcbi-1001105-t004){ref-type="table"}, transcripts are potential alternatives to TNF-α therapies and have known dependencies on TNF-α or are proteins that could be targeted with a small molecule and their predicted modulation might be expected to modulate DAS28 or joint health (*p*\<0.05). ::: ::: {#pcbi-1001105-t004 .table-wrap} 10.1371/journal.pcbi.1001105.t004 Table 4 ::: {.caption} ###### Category 2 Transcripts from the pre-treatment model. ::: {#pcbi-1001105-t004-4} Gene Symbol Name TJ SJ DAS28 ------------- ----------------------------------- ---- ---- ------- WARS tryptophanyl-tRNA synthetase X X LASS5 LAG1 homolog, ceramide synthase 5 X X CTSC Cathepsin C X X ::: ::: {#pcbi-1001105-t005 .table-wrap} 10.1371/journal.pcbi.1001105.t005 Table 5 ::: {.caption} ###### Category 3 Transcripts from the post-treatment model. ::: {#pcbi-1001105-t005-5} Gene Symbol Name TJ SJ DAS28 ---------------- ------------------------------------------------------------------------ ---- ---- ------- 1558906\_a\_at NA X X MGST3 microsomal glutathione S-transferase 3 X FLJ22662 hypothetical protein FLJ22662 X X LANCL1 LanC lantibiotic synthetase component C-like 1 (bacterial) X SFRS18 splicing factor, arginine/serine-rich 18 X X KPNB1 karyopherin (importin) beta 1 X X EXOSC6 exosome component 6 X RAD23A RAD23 homolog A (S. cerevisiae) X DPYD dihydropyrimidine dehydrogenase X X FUSIP1 FUS interacting protein (serine/arginine-rich) 1 X X WIPF2 WAS/WASL interacting protein family, member 2 X X CFLAR CASP8 and FADD-like apoptosis regulator X HNRPA3P1 heterogeneous nuclear ribonucleoprotein A3 pseudogene 1 X INHBC inhibin, beta C X X COX18 COX18 cytochrome c oxidase assembly homolog (S. cerevisiae) X FGL2 fibrinogen-like 2 X GSTO1 glutathione S-transferase omega 1 X FLJ43663 hypothetical protein FLJ43663 X TNRC6B trinucleotide repeat containing 6B X EIF3F eukaryotic translation initiation factor 3, subunit F X MBNL1 muscleblind-like (Drosophila) X SUB1 SUB1 homolog (S. cerevisiae) X PRDM2 PR domain containing 2, with ZNF domain X C10orf46 chromosome 10 open reading frame 46 X CENTB2 centaurin, beta 2 X LRRFIP2 leucine rich repeat (in FLII) interacting protein 2 X DOCK8 dedicator of cytokinesis 8 X TMEM14C transmembrane protein 14C X OSGEP O-sialoglycoprotein endopeptidase X ATP6AP2 ATPase, H+ transporting, lysosomal accessory protein 2 X LOC727918 hypothetical protein LOC727918 X RPS2 ribosomal protein S2 X VAPA VAMP (vesicle-associated membrane protein)-associated protein A, 33kDa X RNF6 ring finger protein (C3H2C3 type) 6 X HLA-DPB1 major histocompatibility complex, class II, DP beta 1 X EXT1 exostoses (multiple) 1 X In [Tables 5](#pcbi-1001105-t005){ref-type="table"} and [6](#pcbi-1001105-t006){ref-type="table"}, transcripts are predicted to impact joint health (*p*\<0.05) but not DAS28 (*p*\>0.05). ::: ::: {#pcbi-1001105-t006 .table-wrap} 10.1371/journal.pcbi.1001105.t006 Table 6 ::: {.caption} ###### Category 3 Transcripts from the pre-treatment model. ::: {#pcbi-1001105-t006-6} Gene Symbol Name TJ SJ DAS28 ------------- ------------------------------------------------------------------------------------------- ---- ---- ------- PITPNA phosphatidylinositol transfer protein, alpha X N4BP1 Nedd4 binding protein 1 X XPO6 exportin 6 X IMPDH1 IMP (inosine monophosphate) dehydrogenase 1 X HLA-E major histocompatibility complex, class I, E X DPYSL2 dihydropyrimidinase-like 2 X LRP10 low density lipoprotein receptor-related protein 10 X EIF4G2 eukaryotic translation initiation factor 4 gamma, 2 X RALY RNA binding protein, autoantigenic (hnRNP-associated with lethal yellow homolog (mouse)) X NT5C2 5′-nucleotidase, cytosolic II X PSMB9 proteasome (prosome, macropain) subunit, beta type, 9 (large multifunctional peptidase 2) X STAT3 signal transducer and activator of transcription 3 (acute-phase response factor) X FNBP1 formin binding protein 1 X IL32 interleukin 32 X TRBC1 T-cell receptor beta constant 1 X PCSK7 proprotein convertase subtilisin/kexin type 7 X SSH2 slingshot homolog 2 (Drosophila) X MAPKAP1 mitogen-activated protein kinase associated protein 1 X LAMP2 lysosomal-associated membrane protein 2 X PCGF3 polycomb group ring finger 3 X RUNX1 runt-related transcription factor 1 (acute myeloid leukemia 1; aml1 oncogene) X NFIL3 nuclear factor, interleukin 3 regulated X CFLAR CASP8 and FADD-like apoptosis regulator X Transcripts predicted to impact joint health (*p*\<0.05) but not DAS28 (*p*\>0.05) ::: Literature analysis of REFS identified intervention points {#s2e} ---------------------------------------------------------- Intervention points identified by REFS were analyzed to identify significantly enriched biological process using the GOstats package from R/Bioconductor, an open source software for bioinformatics [@pcbi.1001105-Falcon1], [@pcbi.1001105-Gentleman1]. Molecular processes associated with nuclear factor kappa-B (NF-kappa-B) were identified from the pretreatment gene list ([Table 7](#pcbi-1001105-t007){ref-type="table"}). NF-kappa-B is a family of transcription factors that are induced by TNF-α and other stimuli, and are critically important for inflammatory processes [@pcbi.1001105-Brown1]. Genes associated with myeloid cell differentiation are also enriched in the untreated samples. Myeloid cells, including neutrophils, monocytes and macrophages, are producers of TNF-α. These cells also express TNF receptors, and on stimulation with TNF-α can produce even more of this cytokine as well as other pro-inflammatory molecules [@pcbi.1001105-Smiljanovic1]. Terms relating to T helper cell differentiation and IL-4 biosynthesis were also among the significantly enriched molecular processes for transcripts identified as important for RA under TNF-α blocker therapy ([Table 8](#pcbi-1001105-t008){ref-type="table"}), perhaps reflecting T-cell related pathways as well as B-cell activations and affecting RA while TNF-α signaling is therapeutically suppressed. ::: {#pcbi-1001105-t007 .table-wrap} 10.1371/journal.pcbi.1001105.t007 Table 7 ::: {.caption} ###### Significantly enriched immune response related GO terms for transcripts from untreated model. ::: {#pcbi-1001105-t007-7} GOBPID Pvalue Count Size Term ------------ ------------- ------- ------ ---------------------------------------------------------- GO:0045639 0.000201651 2 5 positive regulation of myeloid cell differentiation GO:0006955 0.002364924 10 778 immune response GO:0045637 0.002648076 2 17 regulation of myeloid cell differentiation GO:0030097 0.008668954 3 93 hemopoiesis GO:0006436 0.009086186 1 2 tryptophanyl-tRNA aminoacylation GO:0042386 0.035861484 1 8 hemocyte differentiation (sensu Arthropoda) GO:0000086 0.044627838 1 10 G2/M transition of mitotic cell cycle GO:0042345 0.048981641 1 11 regulation of NF-kappaB import into nucleus GO:0043123 0.049036663 2 78 positive regulation of I-kappaB kinase/NF-kappaB cascade ::: ::: {#pcbi-1001105-t008 .table-wrap} 10.1371/journal.pcbi.1001105.t008 Table 8 ::: {.caption} ###### Significantly enriched immune response related GO terms for transcripts from anti-TNF-α treated model. ::: {#pcbi-1001105-t008-8} GOBPID Pvalue Count Size Term ------------ ------------- ------- ------ ------------------------------------------------------ GO:0045064 0.010795911 1 2 T-helper 2 cell differentiation GO:0042097 0.010795911 1 2 interleukin-4 biosynthesis GO:0042109 0.010795911 1 2 tumor necrosis factor-beta biosynthesis GO:0045624 0.016150771 1 3 positive regulation of T-helper cell differentiation GO:0030154 0.025370559 7 537 cell differentiation GO:0042092 0.02677505 1 5 T-helper 2 type immune response GO:0042093 0.02677505 1 5 T-helper cell differentiation GO:0045086 0.032044769 1 6 positive regulation of interleukin-2 biosynthesis GO:0045621 0.037286403 1 7 positive regulation of lymphocyte differentiation ::: Network analysis predicts *LASS5* and *IL32* as TNF-α dependent causal factors for swollen joint count {#s2f} ------------------------------------------------------------------------------------------------------ *LASS5*, longevity assurance homolog 5, is identified as a modulator for number of swollen joints. It is a ceramide synthase that synthesizes ceramide *de novo*. It has been shown that overgrowth of rheumatoid synoviocytes leads to joint destruction, and ceramide regulates cell growth by inhibiting pro-survival signals such as those from *AKT*, *MEK* and *ERK* [@pcbi.1001105-Migita1]. Ceramide is also known as a pro-inflammatory signaling molecule and may regulate several matrix metalloproteases that can degrade cartilage tissue [@pcbi.1001105-Bauer1], [@pcbi.1001105-Kapila1]. Simulation results suggest that *LASS5* expression interacts with the expression of interleukin 32 (*IL32*) to modulate the number of swollen joints ([Figure 3](#pcbi-1001105-g003){ref-type="fig"}--[4](#pcbi-1001105-g004){ref-type="fig"}). *In silico* perturbations of *IL32* are predicted to affect number of swollen joints. *IL32* is a cytokine induced by TNF-α and may play a critical role in rheumatoid arthritis. Injection of human IL-32 protein into knee joints of mice leads to joint swelling as well as inflammation and cartilage damage and this effect is dependent on the presence of TNF-α [@pcbi.1001105-Joosten1]. The predicted effect of *IL32* is not detected in the network ensemble built from data from TNF-α blocker treated subjects, agreeing with experimental data and further suggesting that the treated network ensemble is capturing TNF-α independent mechanisms in RA. ::: {#pcbi-1001105-g003 .fig} 10.1371/journal.pcbi.1001105.g003 Figure 3 ::: {.caption} ###### 10-fold knockdown of *IL32* modulates number of swollen joints. Plots of simulated number of tender joints, swollen joints, pain and C-reactive protein concentrations in response to a 10-fold knockdown in gene expression of cytokine *IL32* in A. pretreated subjects and B. TNF-α blocker treated subjects. The effects are only predicted in pre-treated patients, suggesting a dependence on TNF-α signaling. The largest predicted effect is to modulate the number of swollen joints. :::  ::: ::: {#pcbi-1001105-g004 .fig} 10.1371/journal.pcbi.1001105.g004 Figure 4 ::: {.caption} ###### Swollen joints predicted to be modulated by *LASS5* in pre-treated patients only. Plots of simulated number of tender joints, swollen joints, pain and C-reactive protein concentrations in response to a 10-fold knockdown in gene expression of ceramide synthase, *LASS5*, in A. pretreated subjects and B. TNF-α blocker treated subjects. The modulation of swollen joints is only predicted in pre-treated patients, suggesting a dependence on TNF-α signaling. :::  ::: Network analysis predicts *WARS* as a TNF-α dependent causal factor for tender joint count {#s2g} ------------------------------------------------------------------------------------------ Persistence of autoimmune activated T-cells is a feature of RA [@pcbi.1001105-Londei1]. Typically, T-cells can be suppressed by indoleamine 2,3-doxygenase (IDO) signaling which triggers the catabolism of tryptophan and subsequent suppression of T-cell response [@pcbi.1001105-Terness1]. *WARS* is a tryptophanyl-tRNA synthetase whose gene expression is significantly elevated in T-cells derived from the synovial fluid of RA patients and leads directly to the sequestration of intra-cellular tryptophan in the form of tryptophanyl-tRNA. With intra-cellular stores of free tryptophan lowered by over-expression of *WARS*, IDO signaling is now muted and can explain the persistence of activated T-cells in RA patients and their resistance to IDO [@pcbi.1001105-Zhu3]. Simulations suggest that knockdown of *WARS* gene expression can significantly affect the number of tender joints sufficiently and strongly enough to modulate DAS28. *WARS* gene expression is predicted to be a modulator of RA only in pretreated subjects, suggesting that *WARS* mechanism is TNF-α dependent ([Figure 5](#pcbi-1001105-g005){ref-type="fig"}). This assertion is confirmed because enhanced gene expression of *WARS* that leads to tryptophan sequestration is dependent on TNF-α in experimental systems [@pcbi.1001105-Zhu3]. ::: {#pcbi-1001105-g005 .fig} 10.1371/journal.pcbi.1001105.g005 Figure 5 ::: {.caption} ###### Tender joints predicted to be modulated by *WARS* in pre-treated patients only. Plots of simulated number of tender joints, swollen joints, pain and C-reactive protein concentrations in response to a 10-fold knockdown in gene expression of tryptophanyl-tRNA synthetase, *WARS*, in A. pretreated subjects and B. TNF-α blocker treated subjects. The modulation of tender joints is only predicted in pre-treated patients, suggesting a dependence on TNF-α signaling. :::  ::: *CD86* is a predicted to be a TNF-α independent causal factor for joint health {#s2h} ------------------------------------------------------------------------------ *CD86* (B7-2) is identified as a strong modulator of TJ, SJ and DAS28 in the network analysis of TNF-α blocker treated patients and should represent a particular TNF-α independent mechanism that is active in RA patients ([Figure 6](#pcbi-1001105-g006){ref-type="fig"}). ::: {#pcbi-1001105-g006 .fig} 10.1371/journal.pcbi.1001105.g006 Figure 6 ::: {.caption} ###### Modulation *CD86* predicted to affect both tender and swollen joint counts in TNF-α treated patients. Plots of simulated number of tender joints, swollen joints, pain and C-reactive protein concentrations in response to a 10-fold knockdown in gene expression of *CD86*, the target of abatacept (CTLA4-Ig), in A. pretreated subjects and B. TNF-α blocker treated subjects. The modulation of tender joints is only predicted in TNF-α treated patients, suggesting both a mechanism that is independent of TNF-α signaling that could be exploited for subjects that do not respond well to TNF-α blocker therapies. :::  ::: *CD86* is a type I membrane protein expressed by antigen-presenting cells (APC) and is the ligand for *CD28* and *CTLA4* on the surface of T-cells. Binding of this ligand to *CD28* or *CTLA4* provides co-stimulatory signals that can either positively or negatively regulate T-cell activation that are independent of TNF-α. Abatacept (CTLA4-Ig) is an approved biologic drug that exploits this mechanism by blocking the co-stimulatory signal from CD80/CD86 thus preventing the full activation of T-cells [@pcbi.1001105-Lagana1]. The drug is approved for patients that have shown unsatisfactory response to TNF-α blocker drugs, clearly demonstrating that network reconstruction and simulation of data collected under TNF-α blocker treatment recovers TNF-α independent mechanisms that could be used to identify drug targets for the segment of the population that does not respond to TNF-α blocker therapy. *RAP2C* and *GON4L* are novel predicted TNF-α independent causal factors for TJ and SJ {#s2i} -------------------------------------------------------------------------------------- *In silico* perturbations further identified *RAP2C* ([Figure 7](#pcbi-1001105-g007){ref-type="fig"}) and *GON4L* ([Figure 8](#pcbi-1001105-g008){ref-type="fig"}) as causal factors for TJ, SJ and DAS28 score. Very little is currently known about either of these transcripts and this is the first analysis that suggests a role for these transcripts as TNF-α independent modulators of joint health in RA patients. ::: {#pcbi-1001105-g007 .fig} 10.1371/journal.pcbi.1001105.g007 Figure 7 ::: {.caption} ###### *RAP2C* predicted to modulate both tender and swollen joint counts in TNF-α treated patients. Plots of simulated number of tender joints, swollen joints, pain and C-reactive protein concentrations in response to a 10-fold knockdown in gene expression of *RAP2C*, a recently described ras G-protein, in A. pretreated subjects and B. TNF-α blocker treated subjects. The simulations suggest that this novel gene can modulate both tender and swollen joint count in TNF-α treated subjects. *RAP2C* may provide insight into novel TNF-α independent signaling pathways in RA. :::  ::: ::: {#pcbi-1001105-g008 .fig} 10.1371/journal.pcbi.1001105.g008 Figure 8 ::: {.caption} ###### *GON4L* predicted modulate both tender and swollen joint count in TNF-α treated patients. Plots of simulated number of tender joints, swollen joints, pain and C-reactive protein concentrations in response to a 10-fold knockdown in gene expression of *GON4L*, recently described as a novel factor in B-cell differentiation, in A. pretreated subjects and B. TNF-α blocker treated subjects. The simulations suggest that this novel gene can modulate both tender and swollen joint count in TNF-α treated subjects. *GON4L* may provide insight into novel TNF-α independent signaling pathways in RA. :::  ::: *RAP2C* is a novel member of Ras G-protein family and the model predicts that perturbation of *RAP2C* leads to significant changes of TJ, SJ and DAS28 score. *GON4L* is a human ortholog of the *Caenorhabditis elegans* cell lineage regulator of gonadogenesis *GON4L*. This gene is a putative transcription factor that may regulate cell cycle control. *GON4L* has been shown to be an essential regulator of B-cell development [@pcbi.1001105-Lu1]. Notably, B-cells are the target for rituximab, a monoclonal antibody against CD20 that is approved for the treatment of RA patients with insufficient response to TNF-α blockade [@pcbi.1001105-Cohen1]. B cells produce IL-6 upon activation, and IL-6 blocking molecules have also been approved for RA treatment [@pcbi.1001105-Smolen1]. Our analysis suggests that *GON4L* may be a novel target for RA by modulating B-cell differentiation. Discussion {#s3} ========== It is important to recognize that the network ensembles are built from gene expression profiles of whole blood and are not directly measured from diseased tissue in the joints, which complicates the assessment of the transcripts identified as outright targets in RA. Additionally, mRNA from whole blood had to be depleted of hemoglobin mRNA and the resulting subtracted mRNA sample was amplified using specialized PCR fluorescence labeling protocols that may introduce bias in the measurements. The unavoidable technical details meant that collected data were noisy and there was a large amount of network structure diversity recovered in the ensemble sample after network reconstruction. However, the predicted magnitude of the effects of some of the transcripts identified was large enough to infer significant effects despite the uncertainty in the network ensemble. Furthermore, when interpreting inferences from the network ensemble it is important to recognize that there could be hidden common or intermediate causes that were not part of the measured dataset that could modify the logic of the causal inferences encoded in the ensemble. The transcripts identified using network simulations represent quantitative hypotheses about control points in untreated and TNF-α blocker treated RA patients. The promise of this approach is exemplified by the recognition of the T cell co-stimulatory molecule CD86 and the B-cell restricted molecule GON4L in the TNF-α blocker treated samples. The network simulations were based purely on the data and did not include any up-front literature information or supervision. The models suggest that modulation of these molecules would impact disease scores, and these molecules represent previously validated pathways for treating patients with insufficient response to TNF-α blockade. Not every intervention point identified by the models will be as clear as CD86, and the genes we identify should be considered as starting points in the investigation of a wider range of pathway and signaling mechanisms that have not been widely examined for therapeutic benefit. For example, the therapeutic potential of a systematic *WARS* inhibitor is questionable but other molecular components of IDO signaling might present better therapeutic potential. Rheumatoid Arthritis drugs recommended by The American College of Rheumatology (ACR) fall into several categories: small molecule DMARDs including methotrexate, lefluonamide and others; anti-TNF-α agents; T-cell activation modulators; IL-6 antagonists; IL-1 antagonists; and B-cell directed therapy. The analysis conducted in this study is relevant for all of these therapeutic approaches, but ideally suited for identifying new pathways to target in patients with insufficient response to first-line DMARDS and biologics such as anti-TNFs. Whereas the DMARDS have relatively broad mechanisms of action impacting several inflammatory pathways, the biologics all have very targeted mechanisms and are generally used only after DMARDS have failed. While the pleiotropic effects of the biologics overlap, each class of biologics targets a specific feature of inflammation. The anti-TNFs are by far the most common, and these target inflammatory processes driven by macrophages that are close to the top of the inflammatory cascade. In current practice, other biologics are used primarily after DMARDS and anti-TNFs fail. This partly reflects the culture of new drug adoption -- anti-TNFs have a track record and there are several to choose from -- but may also reflect underlying disease heterogeneity. The network analysis supports the notion that underlying disease heterogeneity is important. Different biological pathways are active with and without anti-TNF-α treatment. Furthermore, T-cell and B-cell pathways targeted by approved RA drugs clearly emerge from the model as reflected by the impact of modulating CD86, part of the abatacept (CTLA4-Ig) target, and GON4L, expressed in B cells that are the target of rituximab (anti-CD20). Candidate targets identified by network analysis can be put into specific inflammatory pathways. While the pathways have been previously identified, some of the specific transcripts have not been recognized as being associated with RA before. These newly identified transcripts highlight pathways for small-molecule or biologic treatments of RA. In addition to the few genes mentioned above, network analysis predicts additional transcripts such as *STAT3*. STAT3 has been identified as a pro-survival molecule for RA synoviocytes [@pcbi.1001105-Krause1], and has also been shown to mediate some of the pro-inflammatory signaling of IL-6 {Hirano, \#88} -- yet another target of approved RA treatments. Additionally, *FLJ43663* is predicted to be a secreted protein and represents a novel type of molecule identified through network simulation. Of the eighteen genes significantly affecting counts of Swollen Joints (SJ), correlative statistical models identified only two transcripts, *CTSC* and *IL32*. REFS identified the importance of *RUNX1*, a transcription factor that regulates genes such as *BLK*, *TCR*, *CD3* and *GM*-*CSF* in lymphoid cells [@pcbi.1001105-AlarconRiquelme1] and which may play a role in autoimmune disease such as rheumatoid arthritis [@pcbi.1001105-Yamada1]. REFS ranked *RUNX1* as the fifth most important transcript modulating SJ while correlative statistical models ranked *RUNX1* as the 1280^th^ most important transcript (See [Figure S4A](#pcbi.1001105.s004){ref-type="supplementary-material"}). Further investigation of the REFS network model suggests that *RUNX1* affects SJ by modulating hypothetical gene *FLJ43663*. In this manner, REFS identified a completely novel intervention point and provided insight into its potential mechanism (See [Figure S4B](#pcbi.1001105.s004){ref-type="supplementary-material"}). Network models that incorporate genetic, molecular, and clinical data collected from longitudinal samples represent a powerful complement to classical statistical models for identifying genes and pathways important for disease processes. We developed ensemble models for a small cohort of RA patients sampled prior to and during treatment with TNF-α blockers. These ensemble models accurately predict the involvement of known RA drug targets including T cell co-stimulation and B cell regulation. These observations suggest that other genes and pathways identified in the network ensembles represent promising targets for further investigation. Methods {#s4} ======= Processing and imputation of genotyping data {#s4a} -------------------------------------------- The Illumina HAP300 chip was used to profile the genotypes of patients. The most recent genomic coordinates for human SNPs were downloaded from the Ensembl website. The updated genomic positions were used together with MACH [@pcbi.1001105-Li1], a Markov-chain haplotyper to impute missing genotypes from the data as well as previously identified SNPs associated with RA. HLA types were mapped to SNP positions based on HLA-SNP map [@pcbi.1001105-deBakker1]. All genotypes from the X and Y chromosome were removed from consideration because the analysis of hemizygous genotypes produces false positive associations. SNPs that failed Hardy-Weinberg equilibrium test, or had a minor allele frequency (MAF) (*p*\<0.05), or had a call-rate\<95% were also removed. The SNP QC process resulted in 279,557 SNPs selected for further analysis and the process is detailed in [Table 2](#pcbi-1001105-t002){ref-type="table"} in [Text S1](#pcbi.1001105.s007){ref-type="supplementary-material"}. EIGENSTRAT was used to detect and correct for population stratification on a genome-wide scale that was detected in the full data set using smartPCA. The associated Armitage chi-squared statistic is computed for each SNP and used to rank SNPs associated with phenotypes to produce a ranked list of SNPs that were potentially important in explaining RA treated and untreated phenotypes. 6,075 SNPs and 6,076 SNPs were chosen from the ranked lists to model untreated and treated subjects respectively. Assessment and processing of gene expression data {#s4b} ------------------------------------------------- Simpleaffy was used to assess the quality of microarray hybridization [@pcbi.1001105-Wilson1]. The majority of samples lay within the tolerances suggested by Affymetrix for amplified RNA samples. The entire microarray data set was normalized using FARMS [@pcbi.1001105-Hochreiter1]. This normalization technique has outperformed previously developed methods in the Affycomp II competition [@pcbi.1001105-Irizarry1] in detecting differential gene expression and is used in conjunction with an associated package called I/NI (Informative/Non-Informative) that uses variance across the dataset to identify informative genes for further analysis [@pcbi.1001105-Talloen1]. Assessment and processing of DAS28 data {#s4c} --------------------------------------- All subjects in the data had DAS28 scores for all visits and the components of the DAS28 score were also available. Pair plots of the components of the DAS28 scores show that the components themselves are orthogonal to each other and capture different aspects of RA. Visual analogue scale for overall health assessment scores and the health assessment questionnaire showed a high degree of correlation with the pain score component of the DAS28 and were considered redundant. Tender joint count (TJ), Swollen joint count (SJ) and Pain scores were logit transformed with an additional discrete to continuous continuity correction applied prior to REFS modeling. C-reactive protein concentrations (CRP) were log transformed prior to analysis to ensure valid, non-zero simulation results in response to intervention queries and to stabilize the variance across the data set. Learning probabilistic models from data {#s4d} --------------------------------------- A multivariate system with random variables *X = (X~1~, ..., X~n~)* where each variable may take on values from a discrete (genetic markers) or continuous domain (gene expression and phenotypic data) may be characterized probabilistically by a joint multivariate probability distribution function *P(X~1~, ..., X~n~; Θ).* However, full specification of such joint probability distributions requires a large number of parameters *Θ*. Such a global joint probability distribution admits the following factorization into a product of *local* conditional probability distributions:where each variable *X~i~* is independent of its nondescendants given its *K~i~* parents *Y~j1~, ..., Y~jKi~* (local Markov condition) and *Θ~i~* are parameters for *P~i~*. The *Y* variables are simply a subset of the *X*\'s; we use the *Y* notation to indicate they are inputs to the conditional probability. This approach yields a framework where each particular factorization and choice of parameters is a distinct probabilistic model *M* of the structure of the process that created the observed data [@pcbi.1001105-Pearl1]. Learning these models *M* from a data set *D* is simply determining which factorizations of *P(X~1~, ..., X~n~;Θ)* are most likely given the observation of *D*, and given a factorization, what are the likely values for its parameters *Θ = (Θ~1~, ..., Θ~n~).* Each factorization of *P(X~1~, ..., X~n~)* into model *M* (as in Eq. 1) is represented by a unique Directed Acyclic Graph (DAG) *G* with a vertex for each *Xi* and directed edges between vertices to represent the dependencies between variables embedded in the local conditional distributions, *P~i~(X~i~\|Y~j1~, ..., Y~jKi~)*. In addition to the graph *G*, *M* also specifies distributions for all *Θ~i~* the parameters of local conditional distributions *P~i~*. Subgraphs of *G*, consisting of a vertex and a set of all its incoming edges, and associated local conditional distributions *P~i~* and parameters *Θ~i~*, are referred to here as "network fragments". We interpret each of these network fragments *M~i~* to characterize both the functional variation of its output variable *X~i~* with respect to its parent input variables *Y~j1~, ..., Y~jKi~* and the residual variation in *X~i~*. For integrative genomics we consider several specific functional forms for network fragments and used linear regression. First, consider the case where all of the input variables *Y~j1~, ..., Y~jKi~* are continuous, then we model the centroid of *X~i~* by: and *X~i~* by a normal distribution about that value:The parameters σ*~i~, θ~0~, θ~j1~, ..., θ~jKi~* can be thought of as adjusted to best fit the data in the Maximum Likelihood Estimation (MLE) sense. The likelihood function gives the posterior distribution of the parameter values about the MLE point. Next, consider the case where one of the *Y* variables is discrete. To model its influence its linear term in Eq. 2 is dropped and the discrete value is used to switch the value of the remaining linear fitting parameters. That is to say, for each value of the discrete variable a *different* set *θ~0~, θ~j1~,* ..., *θ~jKi~* of fitting parameters is introduced. Finally, if multiple *Y*\'s are discrete all of their linear terms are dropped from Eq. 2 and their joint discrete state is used as the switching value. In this study all of the output variables *X~i~* are continuous: discrete variables are only taken as inputs. Parallel ensemble sampling {#s4e} -------------------------- To determine which factorizations are likely given the data we use a Bayesian framework to compute the posterior probability of the model *P(M\|D)* from Bayes\' Lawwhere *P(D)* is the probability of *D, P(M)* is the prior probability of the model and is the integral of the data likelihood over the prior distribution of parameters Θ. We assume that data is complete. Assuming that parameters *Θ* are independent, all models are equally likely, and *P(D)* is constant, we factor *P(M\|D)* in Eq. 4 into the product of integrals over the parameters local to each network fragment *M~i~* ~.~ Eq. 4 now becomeswhere *P*(Θ*~i~* \|*M~i~*) is the network specific prior for its parameters. For this work we use Schwartz\'s Bayesian Information Criterion approximation to the above integral (asymptotically exact as the number of samples increases):where κ(*M~i~*) is the number of fitting parameters in model *M~i~* and *N* is the number of samples. We refer to *S* as a "score", but note the minus sign in the definition (to agree with the simulated annealing analogy described below) and so lower scores are more likely. *S~MLE~* is the negative logarithm of the MLE value of the likelihood function. The total network score is:a sum over the scores of each network fragment in the candidate graph model. In principal the repository of candidate network fragments can be constructed by exhaustive enumeration over variables and network fragment forms. We selected models that provided highest likelihood [@pcbi.1001105-Heckerman1], [@pcbi.1001105-Woolf1] and considering at most 2 edges for a particular vertex. However, even with these constraints, the space of all possible graphs is still too large to be sampled by exhaustive enumeration. Instead we use the Metropolis method (Markov Chain Monte Carlo) to generate samples from an equilibrium Boltzmann distribution of candidate structures [@pcbi.1001105-Ding1] from *P(M\|D)*. Each step in a Metropolis Markov Chain corresponds to local transformations such as adding or deleting network fragments. To accelerate convergence we used simulating annealing were we applied the Metropolis method to a sequence of distributions with decreasing *T~j~* (annealing temperature). At each stage *j* the equilibrated samples from *T~j~* initialize the Metropolis method at *T~j+1~*. Convergence of the random walk is monitored along the way and the annealing schedule is dynamically modified to take more Monte Carlo steps when the barriers present a larger obstacle to diffusion through the space of networks. The method for doing this was to estimate rate of change with respect to *T* of the mean total score and also its variance (the angle brackets denote Monte Carlo averages over networks at the current *T*.) From these values the change in temperature, , is selected so that the distribution of at will have 80% overlap with the distribution at *T*. This process of maintaining overlap helps ensure that the sampling will be correct when *T = 1* is reached. In addition, shorter runs were performed to confirm that results are consistent with the longer runs. In the normal usage of simulated annealing to find a global optimum, the control parameter *T* is allowed to go below 1; as long as better solutions are still being found the temperature is allowed to decrease. In our approach we stop at *T = 1* because the sampling there corresponds directly to the posterior distribution *P(M\|D)* in Eq. 4; going to lower values of *T* would lead to over fitting the data. Model intervention simulations {#s4f} ------------------------------ Stochastic simulation of a probabilistic model *M* allows predictions about the distribution of a variable *Xi* to be made under different conditions. The conditions can be interventions with variables in the model and/or different values of inputs to the model. We used Gibbs sampling in which each variable *Xi* is sampled from its conditional Gaussian distribution, such as Eq. 2, 3, whose parameters take on most likely values given data *D*. For simulation of subjects not seen in the training data, only roots of the graph G had values. A simulation routine iteratively sweeps the network and generates samples of variables whose parents have already acquired a value in previous iterations until all variables have values. One full sweep produces one sample (one vector of values of all variables). Interventions such as a knockdown of gene transcript expression level variables are done by removal of the network fragment from M that outputs to the variable and the network is swept as described previously. For each subject in the data analysis, the contribution of each gene simulation was assessed conditioned by the genotype and the other gene expression measures of the subject. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### Predicted phenotypic values of training data for untreated model. Plot of simulated data for phenotypes (number of tender joints, number of swollen joints, pain score, and CRP levels) based on untreated model vs. observed values. Dashed line represents the line of unity (y = x) and each circle represents a patient in the training data. Correlation coefficients (r) for each phenotype are printed as part of the title. (0.75 MB EPS) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### Predicted phenotypic values of training data for treated model. Plot of simulated data for phenotypes (number of tender joints, number of swollen joints, pain score, and CRP levels) based on treated model vs. observed values. Dashed line represents the line of unity (y = x) and each circle represents a patient in the training data. Correlation coefficients (r) for each phenotype are printed as part of the title. (0.74 MB EPS) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S3 ::: {.caption} ###### Predicted phenotypic values of test data. Plot of simulated data for phenotypes (number of tender joints, number of swollen joints, pain score, and CRP levels) based on treated model vs. observed values. Dashed line represents the line of unity (y = x) and each circle represents a patient in the test data. Correlation coefficients (r) for each phenotype are printed as part of the title. (0.74 MB EPS) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S4 ::: {.caption} ###### Comparison of REFS with statistical models & Network topology of example gene. A. Comparison of genes identified by REFS™ and correlative statistical models for swollen joint (SJ) counts of untreated data. B. A Schematic representation of genes upstream of SJ from untreated model. (3.03 MB EPS) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S1 ::: {.caption} ###### Significant transcripts identified from untreated network ensemble. Transcripts from the untreated network ensemble were predicted to significantly modulate any of the DAS28 component scores (p\<0.05). (0.03 MB XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S2 ::: {.caption} ###### Significant transcripts identified from the TNF-α blocker treated network ensemble. Transcripts from theTNF-α blocker treated network ensemble were predicted to significantly modulate any of the DAS28 component scores (p\<0.05). (0.03 MB XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: Text S1 ::: {.caption} ###### Assessment of the accuracy of the network Ensemble. (0.06 MB DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: We are grateful to Peter K. Gregesen and the Autoimmune Biomarkers Collaborative Network (ABCoN) collaborating investigators for access to the ABCoN dataset. The authors would also like to thank Normand Allaire for laboratory support and Dr. Evan Beckman for support. Jadwiga Bienkowska and John Carulli are employees of and shareholders in Biogen Idec. The data in this study were provided by The Autoimmune Biomarkers Collaborative Network (ABCoN), supported by grants from the National Institutes of Health NO1-AR-1-2256 to Dr. Peter K Gregersen. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: **¤:** Current address: Novartis Institute of Biomedical Research, Cambridge, Massachusetts, United States of America [^2]: Conceived and designed the experiments: HX PDM RR IGK JC. Performed the experiments: JB. Analyzed the data: HX PDM. Contributed reagents/materials/analysis tools: HX PDM JB KR REM DD BC. Wrote the paper: HX PDM TC KR JC.

PubMed Central

2024-06-05T04:04:19.663800

2011-3-10

PMC3053316

Introduction {#s1} ============ The fact that survival and reproduction are sometimes a matter of luck rather than fitness, has arguably left many traces in the history of evolution [@pcbi.1002005-Nei1]--[@pcbi.1002005-Wagner1]. Random accidents in the reproductive process lead to sampling errors in the chain of generations. When accumulated over time, these sampling errors can cause significant changes in the abundance of genetic variants. This phenomenon, called random genetic drift, can represent a significant hurdle for adaptation [@pcbi.1002005-Lynch1]. For instance, newly arising beneficial mutations are usually lost by chance and need to occur many times, until they succeed in reaching fixation [@pcbi.1002005-Hartl1]. More generally, random sampling errors tend to reduce diversity by eliminating rare variants from the gene pool. Spatially extended populations are thereby fragmented into patches in which different genetic variants (alleles) dominate [@pcbi.1002005-Kimura1]. Allele frequency gradients between patches are maintained by a balance of genetic drift and dispersal [@pcbi.1002005-Hartl1]. Such spatial structure has important consequences for the process of adaptation. In a spatial setting, novel beneficial mutations occur in one place and need to spread across the habitat to reach fixation [@pcbi.1002005-Novembre1]. These mutant invasions proceed in the form of waves, first described by R.A. Fisher and A. Kolmogorov et al. in 1937 [@pcbi.1002005-Fisher1], [@pcbi.1002005-Kolmogorov1]. Their deterministic analysis of the combined effects of selection and diffusion reveals a characteristic wave speed, which depends on migration and growth rates. Subsequently, it was found that Fisher-Kolmogorov waves appear in most complex systems and control the speed of numerous important dynamical processes, such as chemical reactions [@pcbi.1002005-Xin1], bacterial colony growth [@pcbi.1002005-Matsushita1] or epidemic outbreaks [@pcbi.1002005-Grenfell1], [@pcbi.1002005-Brockmann1]. As a result, Fisher-Kolmogorov waves have been investigated not only in biology, but also in chemistry and physics [@pcbi.1002005-Douglas1]--[@pcbi.1002005-Marquet1]. An entirely deterministic analysis of traveling waves is incomplete as it neglects genetic drift, which is inevitable in finite systems. Already R.A. Fisher noticed that random fluctuations play an important role in selecting a unique wave speed. The sensitivity of traveling waves to genetic drift started to become fully appreciated when stochastic computer simulations became feasible [@pcbi.1002005-Mollison1]. This spurred intensive research efforts, in particular in the statistical physics community, to augment the deterministic analysis by random sampling noise [@pcbi.1002005-vanSaarloos1]. The ensuing stochastic Fisher--Kolmogorov waves are characterized by fluctuating wave fronts and strongly reduced wave speeds. Noise acts as a drag force in these nonlinear systems, with the result that the deterministic wave speed is a threshold that is only approached slowly as population sizes tend to infinity [@pcbi.1002005-Tsimring1]--[@pcbi.1002005-Hallatschek1]. Here, we show that random sampling errors can also *drive* traveling waves. We analyze the stochastic mechanism underlying these noise-driven waves and quantify the conditions under which they emerge in complex biological or physical systems. In the context of evolution, noise-driven waves ensue from the competition for a single limited resource in a spatially extended habitat. Importantly, this phenomenon promotes the evolution of the economical use of a limited resource, which has been hypothesized as one of the earliest forms of altruism, already present at the level of microbial biofilms. Capturing the phenomenon of noise-driven waves requires a fundamental extension of the Fisher-Kolmogorov model for the invasion of mutants. This standard model (and its variants) exclusively deals with mutations that change the growth rate while having no effect on the growth "yield" -- that is the biomass produced per unit of resource. The spread of such growth rate mutations is slowed down by noise, as described above. However, a change in growth rate is expected to involve a change in growth yield as well, as several recent studies have advocated on the basis of thermodynamic principles [@pcbi.1002005-Pfeiffer1], [@pcbi.1002005-Kreft1]. To account for such a trade-off between growth and yield, we extend the Fisher-Kolmogorov model in a minimal way to be able to describe mutations that change both rate and yield. The resulting model supports noise-driven waves because in the presence of number fluctuations, nearby individuals are related, and all gain from an increase in local density. Noise driven waves, hence, arise from a form of kin selection (a version of group selection), which we quantify using simulations and a novel analytical approach. Model {#s1a} ----- Our computer model, illustrated in [figure 1](#pcbi-1002005-g001){ref-type="fig"}, provides the setting for the competition of two types, mutants and wild type, in a spatially extended population. It consists of a linear array of sub-populations, called demes. Individuals have a chance per generation to jump to one of the neighboring demes. The growth of mutants (0) and wild type (1) within a deme from generation to is simulated by the following rule where and are the numbers of wild type and mutants in generation , respectively, and is the total population size of the deme. The first term on each of the right hand sides describes the logistic growth of the deme population : The growth rate declines linearly with increasing population size and vanishes at certain maximal occupancy . This "carrying capacity" is the equilibrium population size per deme at which resource production and consumption just balance. It represents the population density that the environment can sustain, given the available necessities. The second term on the right hand side of each of the equations (1) and (2) accounts for a small difference in the growth rate of mutants and wild type. This implements natural selection against the mutant type in a standard way. Notice that we have chosen selection to act on the ratios of both types but not directly on the total deme population : The -dependent terms in equations (1) and (2) add up to zero. Finally, genetic drift arises in our model from the sampling noise in equations (1, 2), which we generate using standard Wright-Fisher sampling [@pcbi.1002005-Hartl1]. ::: {#pcbi-1002005-g001 .fig} 10.1371/journal.pcbi.1002005.g001 Figure 1 ::: {.caption} ###### Noise can drive traveling waves. A computer model is used to simulate the competition for a common resource between two species, mutants (blue) and wild type (green). Mutants are assumed to use resources more economically than the wild-type. As a consequence, higher population densities can be sustained in the mutant regions. Yet, mutants are unable to invade the wild-type population unless the randomness in the reproduction process (genetic drift) is implemented in the computer model. a) The spatially extended population is represented by a linear array of local populations, called demes. Individuals migrate between neighboring demes at a rate per generation. The population size of the demes ranges from for demes that are occupied by wild-type only to for mutant only demes. Due to the diffusive mixing of both types, the transition from MT to WT occurs in general over more than one deme. b) For very low migration rates, demes are either fixed for the wild-type (WT) or the mutant type (MT), and the transition between both regions is step-like. c) Representative results of stochastic (left) and deterministic (right) simulations with parameters , and . Horizontal and vertical axes represent space and time, respectively. The color shading intermediate between blue ( mutants) and green ( wild-type) indicates the mixture between both types at a given deme. Note that i) mutants invade the wild-type population only in the stochastic simulations, ii) the transition region between mutants and wild type remains stable in the stochastic case but gradually blurs in the deterministic simulations. :::  ::: With constant carrying capacity , the above model simply represents a discretized version of the standard Fisher-Kolmogorov model. For , the wild type sweeps through the population in the form of a traveling wave, thereby displacing the mutant type. However, as we demonstrate below, the assumption of a constant carrying capacity has to be relaxed to account for mutations that change the organism\'s growth yield (biomass produced per unit resource). Therefore, we go beyond the Fisher-Kolmogorov setting and allow for the possibility that the carrying capacity depends on the local composition of the population. Specifically, we assume that a population entirely consisting of mutants has a carrying capacity as opposed to in a purely wild-type population, see [Fig. 1a, b](#pcbi-1002005-g001){ref-type="fig"}. The (small) parameter quantifies the strength of the mutation. In a mixed population with mutant frequency , the carrying capacity is assumed to be given by . Biologically, such a frequency dependent carrying capacity arises whenever the mutant type consumes less resource per generation than the wild type (equivalently, whenever mutants produce more biomass per unit of resource). Such yield-mutants will leave more of the limited resource to its immediate neighbors, notwithstanding their identity, with the net-result of an increased carrying capacity. Natural realisations of this scenario are provided by many microbial species that can boost their growth rates by (partially) shifting catabolic substrate flow into less-energy-conserving branches, resulting in lower biomass yields [@pcbi.1002005-MacLean1]. For instance, yeasts can switch their metabolism from respiration to fermentation plus respiration [@pcbi.1002005-MacLean2], [@pcbi.1002005-Merico1]. Respiration results in higher yield but slower substrate turnover and growth rate. Using fermentation in addition to respiration results in lower yield but higher substrate turnover and growth rate. Mutations with immediate effect on carrying capacity also occur when bacteria compete for space rather than nutrients, as in a tightly packed biofilm [@pcbi.1002005-Kreft1]. A mutation that reduces slightly the space requirements of a mutant cell will effectively increase the local carrying capacity: A population containing a fraction of mutants will be able to reach higher cell densities than an all wild type population. As these microbial examples show, a frequency dependent carrying capacity is an important biological alternative when different types compete for the same limited resource (nutrients, water, sunlight, space, etc.). To highlight the novel effects associated with such a frequency dependent carrying capacity, which lies outside the scope of traditional wave models, we begin our analysis by assuming that the growth rates of mutants and wild type are identical. In the second part of the analysis, however, we will assign a growth rate cost () to the mutants because it is quite unlikely that an increase in population density comes without any cost. Indeed, in the case of microbes competing for the same nutrient source, it is predicted that an increase in metabolic efficiency is usually associated with a decreased growth rate [@pcbi.1002005-Pfeiffer1], [@pcbi.1002005-Kreft1], [@pcbi.1002005-MacLean2]. This case of a trade-off [@pcbi.1002005-Pfeiffer1] between growth rate and yield has received particular attention in the recent literature, and will be discussed in the second part of the analysis. At first, however, we will investigate the above model assuming in order to answer the question whether mutations with will prevail despite the fact that they lack a direct fitness difference. To this end, we stage a "tug of war" between both types. That is, we assume that, initially, all individuals in one half space () are mutants and the entire population in the other half-space () is wild-type. As individuals migrate and reproduce, this initially step-like transition between both types evolves into a more or less smooth interface. Shape and motion of this mixing zone determine whether the mutant invasion will succeed or fail. Results {#s2} ======= We find that, in any *finite* population, mutants can invade (only) with the help of local number fluctuations. That is, the interface between mutants and wild-type gradually shifts towards the wild-type region, as in the simulation [Fig. 1c](#pcbi-1002005-g001){ref-type="fig"} (left). The importance of sampling noise can be verified in purely deterministic simulations that neglect genetic drift, see [Fig. 1c](#pcbi-1002005-g001){ref-type="fig"} (right). Note that the transition region between mutants and wild-type remains at a fixed position and merely broadens diffusively over time. To quantify how strongly mutants dominate over wild-type in finite populations, we measured the invasion speed as a function of the model parameters. The simulation results, summarized in [Fig. 2](#pcbi-1002005-g002){ref-type="fig"}, suggest that the invasion dynamics is controlled by a single parameter , combining carrying capacity , diffusivity , relative increase of the carrying capacity of mutants, and the variance in the offspring number of individuals. The parameter compares the effect of diffusion with the strength of stochastic fluctuations. For large , the wave front extends over many demes, and moves slowly with weak front diffusion. For small , on the other hand, wave fronts are step-like and exhibit strong diffusion. The simulation results in [Fig. 2](#pcbi-1002005-g002){ref-type="fig"} suggest that the wave speed in both regimes can be summarized as ::: {#pcbi-1002005-g002 .fig} 10.1371/journal.pcbi.1002005.g002 Figure 2 ::: {.caption} ###### Speed of noise driven mutant invasions. The wave speed was measured in units of and plotted as a function of the parameter combination . Data sets with different migration rates were used to generate this scaling plot, as indicated in the legend box. The effect of the mutations was set to (black symbols) or (blue symbols); the variance in offspring number was chosen to be . Note that the different data sets collapse onto a single curve. For small , this "master" curve saturates at corresponding to a wave speed . On the double logarithmic scale (inset), the data approaches a straight dashed line for large , consistent with the predicted asymptotic power law dependence . :::  ::: How can one rationalise the stochastic mechanism underlying these noise driven waves? An intuitive argument can be given for the regime , which occurs when the migration rates or local population sizes are small. Then, the flux of migrants is so small compared to the fixation time within a deme, that the transition from wild-type to mutants occurs between two neighboring demes. Hence, the situation usually looks as in [Fig. 1b](#pcbi-1002005-g001){ref-type="fig"} with a step-like interface between wild-type and mutant regions. Under these conditions, the transition region shifts one deme into the wild-type region if a mutant migrates into the first wild-type deme and reaches fixation there. Such events occur at rate because mutant migrants appear in the wild-type region at a rate , and fix with probability . Conversely, the transition region may shift towards the mutant domain if a wild-type becomes established in the first mutant deme. The corresponding transition rate is given by the product of the rate at which wild-type migrants appear in the first mutant deme, , and the fixation probability of a wild-type in mutant demes, . The back and forth stepping of the transition region results in a net speed of in agreement with the small limit of our simulation results. This simple argument shows that the invasion of mutants is made possible by the fact that i) mutants more often attempt to invade wild-type demes than the other way around and ii) that invasion attempts have a higher success probability. Both effects are the result of the larger carrying capacity of mutant demes, and contribute the same amount to the average invasion speed. The situation becomes more complicated when the mixing zone between both types extends over many demes (), and the wave front is smeared out. Nevertheless, the general case can be treated analytically (see [Methods](#s4){ref-type="sec"}). This is made possible by a nonlinear variable transformation due to E. Hopf and J.D. Cole [@pcbi.1002005-Hopf1], [@pcbi.1002005-Cole1], which converts our model of noise driven waves onto the conventional Fisher--Kolmogorov model with parameters that depend on the noise strength. This exact mapping shows that the combination of migration and stochasticity confers an *effective* growth rate advantage of to the mutants. The results for the wave speed in Eq. (3) then follow from the known asymptotic results for noisy Fisher--Kolmogorov waves [@pcbi.1002005-Brunet1], [@pcbi.1002005-Doering1], [@pcbi.1002005-Hallatschek2]. Due to the noise-induced growth rate advantage, mutants will always out-compete the wild-type population provided both types have equal intrinsic growth rate, or fitness. However, as we discussed earlier, the mutants\' ability to increase population densities will usually be associated with growth rate determinant. For heterotrophic organisms, in fact, such a correlation follows from basic thermodynamic principles of ATP production [@pcbi.1002005-Pfeiffer1], [@pcbi.1002005-Kreft1], [@pcbi.1002005-MacLean2]. To account for this trade-off between growth rate and yield [@pcbi.1002005-Pfeiffer1], we have studied our model for a selective disadvantage of the mutants. We find both in simulations ([Fig. 3](#pcbi-1002005-g003){ref-type="fig"}) and theory ([Methods](#s4){ref-type="sec"}) that the noise induced excess growth rate () must be larger than the fitness cost () to ensure invasion of the mutants. As a consequence of this "force" balance, we can determine an optimal carrying capacity , at which mutations are unable to invade. To this end, we assume that relative change in carrying capacity is linearly related to the relative change in growth rate , where the number characterizes the evolutionary costs associated with a small change in carrying capacity. We expect such a linear relation to hold at least for small . Balancing the evolutionary cost for increasing carrying capacities () with the noise induced growth rate of mutants () yields which is the carrying capacity for which mutations with non-zero are unable to invade. In the frame work of evolutionary game theory [@pcbi.1002005-Hofbauer1], the condition in Eq. (5) is called an evolutionary stable strategy towards which populations are expected to evolve on long evolutionary time scales. ::: {#pcbi-1002005-g003 .fig} 10.1371/journal.pcbi.1002005.g003 Figure 3 ::: {.caption} ###### Balance between natural selection and random sampling noise. The wave speed in units of is depicted as a function of a selective disadvantage of mutants that increase the carrying capacity by a factor . Under these conditions, sampling noise and natural selection act in opposite directions - noise favors mutants, natural selection favors wild type. Note that the mutant population expands () provided that the selective disadvantage is less than the ratio of and the carrying capacity . The point at which the speed changes sign defines an evolutionary stable strategy as discussed in the text. The data exhibits a deviation of about from the predicted point of sign change. This deviation can be lowered by using smaller migration rates (data not shown). Migration rates were set to and different values of were used, see the legend box. :::  ::: Discussion {#s3} ========== The emergence of an optimal carrying capacity is intriguing because, even though using resources more efficiently seems to be good for the group, it is not clear how resource efficiency could evolve if it implies a fitness cost. The resulting evolutionary dilemma is analogous to the "tragedy of the commons", a metaphor widely used to describe evolution towards the inefficient use of a common resource [@pcbi.1002005-Hardin1]. This puzzle is particularly striking in microbial populations that exhibit a wide spectrum of phenotypes between fast growing strains with low efficiency in ATP production and slow growing high efficiency strains [@pcbi.1002005-Kreft1], [@pcbi.1002005-MacLean2]. It has been argued that the economical utilisation of resources may be one of the earliest form of altruism, since it is wide-spread already at the level of microbial systems [@pcbi.1002005-Kreft1]. The emergence of this basic form of cooperation in spatially extended habitats has been observed in individual-based simulations [@pcbi.1002005-Pfeiffer1], [@pcbi.1002005-Kreft1], [@pcbi.1002005-Wakano1], but (to our knowledge) no theoretical account could yet quantify the effect. On the contrary, attempts to describe the spread of mutations using the classical Fisher-Kolmogorov approach, which is based on deterministic reaction diffusion equations, came to the conclusion that density increasing mutations are unable to invade [@pcbi.1002005-Wakano2], [@pcbi.1002005-Hou1]. Our analytical results show that stochasticity is the key difference between the individual based simulations and the deterministic theory. Random genetic drift favors mutations that increase the carrying capacity. It thereby promotes the economical use of a limited resource even if this implies a small growth rate detriment. The strength of this effect crucially depends on the parameter , characterizing the trade-off between growth rate and yield. If, for instance, we consider microbes competing for the same nutrient source, we expect that a mutant type that consumes less nutrients will suffer from a comparable reduction in growth rate. In this case, the parameter will be on the order of 1 with the consequence that equation (5) predicts a rather small evolutionary stable carrying capacity . The opposite situation may arise, for instance, when bacteria are competing for space in a dense biofilm. Then, mutant cells would occupy smaller volumes, which could be neutral (or even beneficial) in terms of growth rates. This would imply and, because noise would be strong compared to selection, a rather large evolutionary stable carrying capacity . Thus, in systems where the density changing mutations have little effect on relative fitness but large effect on density, the carrying capacity might indeed result from the balance of noise and selection, as predicted by equation (5). Our study thus provides a predictive null model for the joint evolution of growth rate and yield, which shows that intricate interactions between individuals are not required for the evolution of resource efficiency in spatially extended populations. All that is required is (inevitable) genetic drift in conjunction with spatial structure, which is particularly strong in microbial biofilms. Real biofilms are often characterized by heterogeneous resource distributions, environmental fluctuations, intrinsic instabilities (e.g., finger or sector formation in biofilms), or self-organisation (Touring mechanism), which are beyond our simple null-model. Such spatio-temporal heterogeneities are expected to further increase the levels of genetic drift. Our predictions for the evolutionary optimal carrying capacity should therefore be interpreted as lower bounds for real systems. The mechanism underlying noise-driven waves can be understood in several ways. Within the theory of "kin selection" [@pcbi.1002005-Hamilton1], which is a special case of group selection [@pcbi.1002005-Barton1], one tries to rationalise the advantage of cooperative mutants in terms of an increased relatedness, which makes it more likely that the altruistic benefits are received by conspecifics rather than wild type. From this point of view, genetic drift generates increased relatedness in our model and allows mutants to invade despite a growth rate detriment. A more direct way of rationalizing the role of noise in our model is provided by our discussion of the regime of low migration rates in the [Results](#s2){ref-type="sec"} section. There, we showed that mutants enjoy a higher diffusion flux into the wild type demes and a higher probability of becoming fixed there. Crucially, these advantages require frequency gradients. If mutants were homogeneously distributed in the habitat, diffusion fluxes and fixation probabilities would be identical for all individuals, independent of their identity. This entirely mixed state, lacking any frequency gradients, is in fact the equilibrium state of our model in the deterministic limit of infinite population sizes ([Methods](#s4){ref-type="sec"}). Consequently, the wave speed of noise-driven waves declines as population sizes tends to infinity. For any finite population size, however, frequency gradients are continually generated by the action of genetic drift. In the scenario of our model, these (random) frequency gradients turn into an advantage for the mutants. The importance of frequency gradients for noise-driven waves is clarified mathematically in the [Methods](#s4){ref-type="sec"} section. There, we show that the local growth rate of the mutant frequency is proportional to the square of local frequency gradients. These gradients are generated by genetic drift, leading to an effective growth rate advantage of mutants. A similar mathematical structure occurs in certain reaction diffusion models of group selection, which also exhibit growth rates proportional the square of frequency gradients [@pcbi.1002005-Barton1]. Barton and Clark in Ref. [@pcbi.1002005-Barton1] gave an heuristic explanation of how this mathematical structure could lead to an effective mean growth rate, considering a balanced polymorphism in the limit of small genetic drift. Our exact analysis based on the Cole-Hopf transformation justifies the use of an effective local growth rate and shows that it is given by , which depends on the carrying capacity , the relative increase of the mutant carrying capacity, and the variance in offspring numbers. (The scaling (not the pre-factor) of our *local* effective growth rate is consistent with the *mean* effective growth rate obtained by Barton and Clark [@pcbi.1002005-Barton1].) It is quite remarkable that this effective growth rate and, consequently, the evolutionary stable strategy in equation (5) do not depend on either diffusion constant nor the dimensionality, even though migration and population structure are needed for the phenomenon of noise driven waves. In summary, we have seen that the established Fisher wave model is unable to account for a trade-off between growth rate and yield. To overcome this limitation, we have generalized the Fisher-Kolmogorov wave model such that mutations are allowed that change both the growth rate and the carrying capacity. We found that the extended model exhibits traveling waves that are driven by random sampling errors. The ensuing noise driven waves are described analytically and compared with classical Fisher--Kolmogorov waves. The most striking difference is that the speed of noise driven waves decreases (like a power law) as population sizes tend to infinity, quite in contrast to classical Fisher--Kolmogorov waves. Comparing the strength of the noise-induced driving force with natural selection led us to the prediction of an evolutionary optimal carrying capacity. This implies that random genetic drift promotes the economical use of a limited resource, one of the most basic forms of altruism. We suspect that this mechanism has been acting over long evolutionary times, because it merely rests on random genetic drift in conjunction with spatial structure, which must have been present already in the most ancient microbial systems. In the sense of Wright\'s shifting balance hypothesis [@pcbi.1002005-Wright1], our model describes a mechanism of peak shifts that relies on pure chance rather than selection. Although our model was formulated with an evolutionary application in mind, its mathematical structure arises in many problems that combine diffusion and interaction of discrete entities. Sampling errors turn into a driving force whenever reaction rates depend on the magnitude of gradients. This occurs, for instance, in problems where the diffusivities depend on population densities [@pcbi.1002005-Murray1], or vary among species, which can lead to Turing patterns [@pcbi.1002005-Wakano3]. Thus, pattern formation by genetic drift may be an important mechanism in many complex systems including biological evolution. Methods {#s4} ======= Noise as a driving force {#s4a} ------------------------ Here, we give an analytic derivation of our result equation (3) for the wave speed of noise driven waves in the absence of any direct selection against the mutant type. Our analysis is based on nonlinear variable transformation that maps the model of noise driven waves to classical Fisher-Kolmogorov waves. The following also discloses the general mathematical conditions, for which noise can act as a driving force in pattern forming systems. The main text contained a brief intuitive argument for the wave speed under conditions of small migration rates, where the transition between wild-type and mutant demes is step-like, as in [Fig. 1b](#pcbi-1002005-g001){ref-type="fig"}. This weak migration limit was relatively easy to analyze because the state of the system frequently returns to a well-defined initial state (renewal process). Next, we consider the other extreme, in which the dynamics becomes deterministic. As mentisoned in the main text, previous studies as well as our simulations [@pcbi.1002005-Wakano2], [@pcbi.1002005-Hou1] indicate the absence of traveling waves in this deterministic limit, and we would like to explain these observations analytically. The general (and most interesting) stochastic case with intermediate migration rates is treated subsequently by adding the appropriate fluctuations. In the deterministic limit, the migration of individuals between demes can be approximated by diffusion with diffusivity . In this framework, the spatially varying population density is described by a field that depends on time and a continuous deme index . The dynamics of this field is given by a spatial analog of the logistic equation,where the local growth rate depends on the ratio between total population density and local carrying capacity, and reads in our units of time. Whereas for small densities , the growth rate equals the linear birth rate per generation, the growth rate disappears at carrying capacity , which is a general feature of logistic growth. As discussed in the main text, the carrying capacity depends on the local frequency of mutants by virtue of As mutants and wild-type are subject to the same migration and growth rates, the evolution equation for the mutant density must have the same form as equation (6), It is convenient to eliminate in favor of the frequency of mutants because appears in the expression for the carrying capacity, equation (8). After a further substitution from equation (6), we obtain We seek a solution of equation (6) and equation (10) for a step function initial condition, . In the Supporting [Text S1](#pcbi.1002005.s001){ref-type="supplementary-material"}, we show that the density closely follows the carrying capacity, provided that the migration rate is small, The assumption becomes exact in the limit while const. In this quasi-static regime, we may substitute in equation (10) to arrive at a closed equation for , The nonlinearity proportional to is non-negative everywhere. Neglecting this term might cause serious errors as its integral over the whole space could be large or even divergent. In fact, the nonlinearity turns out to be a singular perturbation and, thus, the crucial point of equation (12). Fortunately, this nonlinearity can be removed by a variable transformation due to E. Hopf and J.D. Cole [@pcbi.1002005-Hopf1], [@pcbi.1002005-Cole1]. Thereto, we introduce the new dynamical field which represents the fraction of mutants to leading order in , . In terms of this new field, equation (12) transforms into a simple diffusion equation It is clear that the diffusion equation does not admit traveling wave solutions. Instead, equation (14) with a step function initial condition has a solution of the form , which can be easily found analytically. The form of the scaling variable suggests that the solution describes a front that is slowly broadening due to diffusion. The typical width and position of the front grows as the characteristic length scale for diffusion. Even though the mean position of the front moves towards the wild-type domain, it does so at an ever decreasing speed. Both observations, front broadening and vanishing front speed, are consistent with the deterministic simulations reported in the main text, [Fig. 1c](#pcbi-1002005-g001){ref-type="fig"} (right). There, we had to conclude that mutants are not able to invade in the deterministic limit. For large *but finite*, however, we can no longer neglect sampling errors (genetic drift). The mutant frequency then becomes a stochastic field that fluctuates due to population turnover from generation to generation. These sampling errors, for example generated by Wright-Fisher sampling [@pcbi.1002005-Hartl1], cause a noise term in the equations (10,12), which reads [@pcbi.1002005-Barton2], [@pcbi.1002005-Korolev1] where is the variance in offspring number and the stochastic forcing term has white noise correlations, The square of the amplitude in front of the noise term in equation (15) represents the expected variance in mutant frequency due to the sampling from generation to generation. Altogether, the *noisy* dynamics of the mutants\' frequency is described by In contrast to the deterministic case, frequency gradients remain finite in the long time limit as they are continuously generated by the noise term. It thus seems reasonable that these fluctuations could turn the gradient squared term into a veritable growth term. How strong will this stochastic driving force be? It turns out that this effect becomes manifest when we apply the above nonlinear variable transformation to the stochastic differential equation (17). In doing so, one has to appreciate that stochastic differential equations have peculiar transformation rules. These so-called Ito transformation rules result from the fact that, during a short time interval , fluctuations have an amplitude proportional to (like a random walk) instead of . As a consequence, a non-linear variable transformation automatically leads to an additional drift term in the transformed equation, called a "spurious" drift term [@pcbi.1002005-vanKampen1]. The Cole-Hopf transformation (13) therefore results in equation (14) plus a spurious drift term and a noise term. The new drift term has the form of a logistic growth term,favoring the growth of the mutants. The noise term takes the form on the right hand side. The suppressed terms of order turn out to become of higher order than the displayed terms after the following rescaling With these substitutions, the stochastic equation of motion takes the form The suppressed terms are now of higher order, , and may be neglected for small . The remaining leading order of equation (22) has the form of a noisy Fisher-Kolmogorov wave equation [@pcbi.1002005-Fisher1], [@pcbi.1002005-Kolmogorov1]. The parameter , introduced in the main text, represents the effective strength of the noise term. The asymptotic behavior of the wave speed as a function of reported in equation (3) and [Fig. 2](#pcbi-1002005-g002){ref-type="fig"} now follows from known results [@pcbi.1002005-Fisher1], [@pcbi.1002005-Kolmogorov1], [@pcbi.1002005-Brunet1], [@pcbi.1002005-Doering1], [@pcbi.1002005-Hallatschek2] on the stochastic Fisher-Kolmogorov equation. Finally, we note that the square gradient nonlinearity in equation (12) was the crucial mathematical structure from which our noise driven waves emerged. It is clear that similar waves arise in any reaction diffusion system of discrete objects provided that the reaction terms contain similar gradient square non-linearities. In these systems, noise turns into a driving force because it randomly creates and maintains gradients, which are absent in the deterministic limit. Trade-off between growth rate and yield {#s4b} --------------------------------------- As mentioned earlier, there are general reasons to posit a trade-off between growth rates and densities, at least in heterotrophic organisms [@pcbi.1002005-Pfeiffer1]. This means that mutants that use resources more efficiently (and therefore allow for higher population densities) may suffer from a reduced fitness. To account for this possibility, we have included a selective disadvantage for the mutants; i.e. we assume that the growth rate mutants is by a factor , smaller than that of the wild-type, which is in the chosen units of time. This leads to a negative logistic growth term in the equations (10, 12) for the frequency of the mutants. For , this would trigger a genetic Fisher wave of wild-type invading the mutants. To study the case, observe that the logistic term is carried through all the steps that lead from equation (10) to equation (22). In equation (22), it leads to the replacement A traveling wave of mutants invading the wild-type population will occur only if this growth term is positive. In other words, the stability condition for a trade-off between growth rate and yield is given by This criterion was used in the main text to derive the evolutionary stable strategy in equation (5). We would like to remark that, in contrast to the wave speed, the statement in equation (5) is *independent* of the control parameter . Furthermore, all steps of our analysis, including the nonlinear Cole-Hopf transformation, can be carried out in higher dimensions and result in the same stability criterion as in one dimension. Therefore, the evolutionary stable strategy formulated in equation (5) represents a fairly general result for weak selection. Supporting Information {#s5} ====================== Text S1 ::: {.caption} ###### Detailed analysis of the quasi-static assumption, , which led to the closed stochastic equation (17) for the mutant frequency. (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: We gratefully acknowledge helpful discussions with Martin Nowak and Nick Barton. The author has declared that no competing interests exist. Funding was generously provided by the Max Planck Society. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: OH. Performed the experiments: OH. Analyzed the data: OH. Contributed reagents/materials/analysis tools: OH. Wrote the paper: OH.

PubMed Central

2024-06-05T04:04:19.670816

2011-3-10

PMC3053317

Introduction {#s1} ============ Biochemical pathways and feedbacks in gene regulatory networks (GRNs) shape when and how much genes are expressed. Differential gene expression can lead to qualitative changes in cellular phenotypes, whether via alternative cell fate determination in unicellular organisms (e.g., competence [@pcbi.1002006-Solomon1], sporulation [@pcbi.1002006-Stragier1], persistence [@pcbi.1002006-Balaban1], and infected cell fate [@pcbi.1002006-Ptashne1]) or via cell differentiation in multi-cellular organisms (e.g., lineage determination [@pcbi.1002006-Morrison1]). The steps leading to qualitative changes in phenotype are not strictly deterministic. Gene regulation is an inherently noisy process involving transcription control, translation, diffusion and chemical modifications of transcription factors, all of which may be characterized by stochastic fluctuations due to low copy numbers of regulatory molecules [@pcbi.1002006-McAdams1]--[@pcbi.1002006-Kaufmann1]. As a result, genetically identical cells can have marked differences in the state of regulatory molecules even when faced with identical environmental conditions [@pcbi.1002006-Spudich1]--[@pcbi.1002006-Maamar1]. Explanations for alternative cell fate determination generally presume the existence of multiple stationary states within the GRN [@pcbi.1002006-Maamar2], [@pcbi.1002006-Losick1]. Determination of cell fate is therefore usually described as the result of the interplay between noise and deterministic dynamics of GRNs which determines the relative frequency of each decision [@pcbi.1002006-Maamar2], [@pcbi.1002006-Xiong1]. A potential problem with this explanation is that cellular decision making occurs within finite time. From a theoretical point of view, differences in asymptotic dynamics are not necessary for regulatory dynamics to reach markedly different transient states. The hypothesis that transient dynamics can drive cell fate determination has been suggested in the context of HIV-1 latency where a bistable response is observed despite the purported monostability of the GRN [@pcbi.1002006-Weinberger1]. Here, we take a generalized approach to a similar problem by considering cell fate determination as the result of stochastic transient dynamics of a GRN. Our starting point is the fact that extrinsic variation can drive substantial differences in the transient state of regulatory molecules [@pcbi.1002006-Alon1]. That is to say, ensembles of cells with the same initial state of regulatory molecules which are exposed to two different conditions can follow distinct transient trajectories on average. In such a case, gene expression will be characterized by an early period in which transient trajectories are unresolvable with respect to the stochastic noise and a middle period in which they are markedly different. However, we claim that such transient differentiation in regulatory state need not be accompanied by marked differences in asymptotic, i.e., steady-state, behavior. Instead, we hypothesize that alternative cell fate decisions can be mediated by first passage processes of regulatory molecules [@pcbi.1002006-Darling1], [@pcbi.1002006-Redner1]. We examine the effect of first passage processes in stochastic GRNs within the initial decision switch between lysis and lysogeny by phage . Bacteriophage is perhaps the simplest example of an organism with alternative developmental modes, which are quiescent (lysogenic) and productive (lytic) growth upon infecting *E. coli* cells [@pcbi.1002006-Ptashne1], [@pcbi.1002006-Hendrix1]--[@pcbi.1002006-Court1]. Here we focus on how phage\--infected cells are lysed or become lysogens as a function of the number of coinfecting phages (also known as the cellular multiplicity of infection denoted as ). Experimental infection assays have revealed that *E. coli* cells that are multiply infected tend to become lysogens whereas singly infected cells tend to be lysed [@pcbi.1002006-Kourilsky1], [@pcbi.1002006-Kobiler1]. The decision to lyse a cell or enter lysogeny is stochastic [@pcbi.1002006-Arkin1], [@pcbi.1002006-Singh1], and the fraction of lysogeny is a probabilistic function of the number of coinfecting phages and cell volume [@pcbi.1002006-StPierre1]--[@pcbi.1002006-Zeng1]. Cells that become lysogenic may later spontaneously induce leading to virion production and cell lysis. The stability of the lysogenic state has also been evaluated in light of first exit problems [@pcbi.1002006-Aurell1], many of whose concepts we adapt in the current model of the initial decision switch. A significant advantage in modeling phage is that the core pathways of lysis-lysogeny have been studied extensively. Subsequent to infection, the repressors (CIs) bind cooperatively to adjacent operator sites [@pcbi.1002006-Hochschild1], and cooperative binding can induce DNA loops which enhance the stability of the lysogenic state [@pcbi.1002006-Griffith1]--[@pcbi.1002006-Morelli1]. Early quantitative studies of the initial lysis-lysogeny decision utilized statistical thermodynamic models which described the dynamics of gene regulation by cooperative binding of CI [@pcbi.1002006-Ackers1], [@pcbi.1002006-Shea1]. Arkin *et. al.* developed a fully stochastic model based on transcription, translation and protein interactions [@pcbi.1002006-Arkin1]. Whether cells were fated to lysis or lysogeny was ascribed to intrinsic stochasticity, whose complexity rendered it intractable for mathematical analysis. More recently, theoretical work has suggested that alternative decisions of lysis and lysogeny may be due to inherent bistability of the phage GRN with respect to changes in copy number concentration ( divided by host cell volume) [@pcbi.1002006-Weitz1]. However, this model presumes that differences in asymptotic dynamics lead to changes in cell fate, without considering stochastic effects in transient dynamics. In this study, we demonstrate that biased alternative cell fate decisions are possible due to transient divergences within gene regulatory dynamics. As evidence, we develop and analyze a quantitative model of a GRN of phage based on empirical analyses of viral infection. Although the structure of the phage GRN is relatively well established, the quantitative values of most kinetic parameters involved in viral gene regulation remain either unknown or poorly constrained. We examine two sets of kinetic parameters close to consensus empirical estimates which we refer to as transiently divergent and asymptotically divergent, respectively. We show that the dynamics of the GRN with these parameter sets are similar shortly after phage infection but the asymptotic dynamics are qualitatively distinct as a function of viral genome concentration. Next, we compare the fraction of lysogeny as a function of viral genome concentration in the two parameter sets. Cell fate is determined via first passage processes of two regulatory proteins, Q and CI, corresponding to lysis and lysogeny, respectively (see [Models](#s4){ref-type="sec"}). We find that equivalent responses of cell fate to changes in viral genome concentration can be obtained with either parameter set, suggesting caution must be applied in interpreting alternative cell fate determination as a hallmark of bistability. In the process, we also discuss how thresholds of first passage processes can change the fraction of lysogeny and the time scale of decisions. Finally, we compare model results with experimental data on cell fate outcomes from single cell assays [@pcbi.1002006-Zeng1]. We propose an alternative data collapse of the observed cell fate outcomes, consistent with a previously unidentified gene dosage compensation mechanism. We show that including gene dosage compensation at the mRNA level in our stochastic model of transient fate determination also leads to the form of data collapse observed in the single cell study. We conclude by discussing means to reconcile multiple competing hypotheses for observed heterogeneity in the phage GRN. Results {#s2} ======= Deterministic dynamics of qualitatively identical phage decision switches can be asymptotically or transiently divergent {#s2a} ------------------------------------------------------------------------------------------------------------------------ We first analyze a deterministic model of a GRN of phage (see [Fig. 1](#pcbi-1002006-g001){ref-type="fig"}, and [Models](#s4){ref-type="sec"} Eq. (3)). Prior to phage infection, there are no viral proteins and mRNAs in the host cell. A cell can be infected by phages, which we vary one to five for a fixed cell volume. Cell fate, either lysis or lysogeny, is determined based on the first passages of a pair of fate-determining regulatory molecules, CI and Q (see [Fig. 2](#pcbi-1002006-g002){ref-type="fig"} (B,E)). We model lysogeny as occurring when CI exceeds a concentration threshold and lysis as occurring when Q exceeds a concentration threshold. We set the value of these thresholds at 100 nM each, and explore the impact of varying these thresholds levels. Values of kinetic parameters necessary for modeling the lysis-lysogeny decision switch are known to within a few percent error in some cases, unknown in other cases, or have estimates with significant uncertainty (see [Table 1](#pcbi-1002006-t001){ref-type="table"}). We chose two sets of parameters which are close to the consensus estimates, but that show markedly distinct asymptotic behaviors especially when . GRNs with these two sets are asymptotically and transiently divergent, respectively ([Fig. 2](#pcbi-1002006-g002){ref-type="fig"}). We define a phage GRN with a set of kinetic parameters to be asymptotically divergent if each deterministic trajectory for crosses the CI and Q thresholds only once. Otherwise, a GRN is referred to as transiently divergent. ::: {#pcbi-1002006-g001 .fig} 10.1371/journal.pcbi.1002006.g001 Figure 1 ::: {.caption} ###### Core genetic components of lysis-lysogeny decision switch in phage . \(A) Schematic diagram of genes and promoters. CI and CRO dimers are the transcription factors for and while and is controlled by CII tetramers. Black arrows represent open reading frames of promoters when activated (, and ) and antisense transcript *aQ*. (B) Interactions among gene products. Regular and blunt arrows represent positive and negative feedbacks, respectively. CI dimers are self-activators while repressing the other genes, and CRO dimers repress all the genes in the system. CII tetramers activate *cI* transcription, and suppress Q expression by transcribing antisense mRNAs. :::  ::: ::: {#pcbi-1002006-g002 .fig} 10.1371/journal.pcbi.1002006.g002 Figure 2 ::: {.caption} ###### Dynamics of regulatory proteins, CI and Q, when the GRN is asymptotically divergent and transiently divergent. \(A) Phase diagram of CI-Q dynamics when asymptotically divergent for . Note that the system is bistable. (B) Phase diagram of CI-Q dynamics starting from no viral proteins when asymptotically divergent. Thresholds of CI and Q (both at 100 nM) represent the concentrations above which decisions are lysogenic and lytic, respectively. Trajectories cross the threshold only once. (C) Asymptotically divergent dynamics of Q concentration as a function of time. (D) Phase diagram of transiently divergent system with . Note that the system is not bistable. (E) Phase diagram of CI-Q dynamics of the transiently divergent phage GRN. At the deterministic trajectory crosses the threshold three times, and decisions change from lysis to lysogeny as a function of time. (F) Transiently divergent Q dynamics. :::  ::: ::: {#pcbi-1002006-t001 .table-wrap} 10.1371/journal.pcbi.1002006.t001 Table 1 ::: {.caption} ###### Parameters for transiently divergent and asymptotically divergent GRNs. ::: {#pcbi-1002006-t001-1} Parameter Reference value Reference Asymptotically divergent Transiently divergent ----------- ----------------- --------------------------------------------------------- -------------------------- ----------------------- [@pcbi.1002006-Hawley1], 0.042 [@pcbi.1002006-Arkin1] 0.014 0.013 0.016 [@pcbi.1002006-Pakula1] 0.033 0.056 0.16 w/o CIII [@pcbi.1002006-Kobiler2] 0.13 0.22 0.0095 0.016 0.12 [@pcbi.1002006-Court2] 0.1 0.1 0.06 [@pcbi.1002006-Shea1] 0.055 0.055 0.84 [@pcbi.1002006-Shea1], [@pcbi.1002006-Hawley1] 0.82 0.70 0.50 0.83 0.78 0.63 0.66 [@pcbi.1002006-Shea1], 3.42 [@pcbi.1002006-Hwang1] 0.79 0.88 0.9 [@pcbi.1002006-Arkin1] 1.24 0.93 3.1 3.8 0.05 [@pcbi.1002006-Sauer1], 0.18 [@pcbi.1002006-Burz1] 0.060 0.079 5.8 [@pcbi.1002006-Jana1], 307 [@pcbi.1002006-Darling2] 4.6 6.7 0.065 0.020 0.093 0.068 0.14 0.10 ) 0.38 0.47 0.02 [@pcbi.1002006-Levine1] 0.15 0.068 0.52.0 ::: The transient dynamics for the phage GRN given either parameter set (either asymptotically or transiently divergent) are similar during the time scale of lysis-lysogeny decision (min). The asymptotically divergent phage GRN exhibits lysis for and lysogeny when . Note that the ratio between CI and Q changes dramatically as a function of from having far more Q to having far more CI at steady state. Only when are there two possible steady states, but the initial condition leads to lysis ([Fig. 2](#pcbi-1002006-g002){ref-type="fig"} (A)). In contrast, the transiently divergent GRN is monostable for all values of that we considered. Further, the steady-state CI and Q concentrations have far greater levels of CI than Q, suggesting that an asymptotic analysis would suggest that the transiently divergent GRN would always lead to lysogeny. However, note that when , Q increases rapidly, exceeds the threshold for lysis, and only later does it drop down and approaches a case where Q is low and CI is high ([Fig. 2](#pcbi-1002006-g002){ref-type="fig"} (D--F)). Thus, there is an inconsistency between expectations for cell fate determination as viewed in finite time vs. that viewed asymptotically. Alternative cell fates as determined by transient viral gene regulation {#s2b} ----------------------------------------------------------------------- The initial lysis-lysogeny decision of phage is sensitive to the external conditions of and cell size. Empirical analyses have shown this decision to be highly stochastic with the fraction of lysogeny between and for physiologically relevant and cell size [@pcbi.1002006-Zeng1]. To model the stochastic nature of this decision, we assume that first passage processes of CI and Q determine whether lysis or lysogeny occurs in an infected cell. Lysogeny occurs if CI reaches its critical concentration before Q does. Lysis occurs if the opposite holds true. We follow the approach of Arkin *et. al.* [@pcbi.1002006-Arkin1] and run fully stochastic simulations of the phage GRN while setting both lytic and lysogenic thresholds at (see [Models](#s4){ref-type="sec"}). We assume that reaching a decision of lysis or lysogeny brings a topological change to the GRN. Thus, we stop the dynamics at the time of a decision since our phage model cannot describe the post-decision regulatory dynamics. [Fig. 3](#pcbi-1002006-g003){ref-type="fig"} depicts a subsample of trajectories in the phase space of CI-Q labeled according to which decision is reached via a first passage process. Note that there is a delay for CI to be expressed since sufficiently abundant CII is required for initial CI expression. In contrast, Q can be produced immediately after phage infection. When the host is singly infected, lysis is the dominant decision, and CI does not build up until a significant amount of Q is produced ([Fig. 3](#pcbi-1002006-g003){ref-type="fig"} (A)). At higher ( for [Fig. 3](#pcbi-1002006-g003){ref-type="fig"} (B)), CII and Q are produced at a higher rate. Depending on the CII expression level Q can be repressed while CI becomes active which leads to lysogeny. In comparison to the deterministic dynamics described in the previous section, there is significant variability in the lysis-lysogeny bias of the GRN, though the bias itself is affected by changes in and cell volume (as described in the next section). ::: {#pcbi-1002006-g003 .fig} 10.1371/journal.pcbi.1002006.g003 Figure 3 ::: {.caption} ###### Stochastic realization of C and Q dynamics for (A) and (B) . Trajectories are sampled for every 1/4 minute. The system is transiently divergent, and thresholds are set at 100 nM for both CI and Q. Each curve represents a single realization, and 50 realizations are shown here. Red trajectories indicate that decisions are lytic whereas blue ones represent lysogeny. :::  ::: Probability of lysogeny is an increasing function of phage genome concentration {#s2c} ------------------------------------------------------------------------------- We vary the volume of host cells (denoted as ) as well as in order to investigate how cell fate responds to changes in the concentration of viral genomes (). For consistency with experimental studies and to model physiologically reasonable values, we vary from one to five, and vary from 0.5 to 2 . [Fig. 4](#pcbi-1002006-g004){ref-type="fig"} shows the fraction of lysogeny as a function of phage genome concentration. Regardless of bistability in the phage GRN, we find that first passage mediated decision making can lead to systematic biases in alternative cell fate determination. Phages preferentially enter lysogeny when multiple phages infect the same hosts while singly infected hosts tend to be fated for lysis. The relative frequencies of lysis or lysogeny can be collapsed as a function of an extrinsic parameter . Our results match the general trend of recent experimental observations which demonstrated that the fraction of lysogeny goes up as phage genome number increases or cell volume decreases [@pcbi.1002006-StPierre1], [@pcbi.1002006-Zeng1]. Importantly, the functional responses to phage genome concentration are nearly indistinguishable even for two parameter sets which have qualitatively different asymptotic dynamics ([Fig. 4](#pcbi-1002006-g004){ref-type="fig"} and [Fig. 2](#pcbi-1002006-g002){ref-type="fig"} (B,E)). The biased decision response as a function of phage genome concentration is due to the similarity of transient dynamics, irrespective of asymptotic dynamics that could have been followed. Hence, the finding that infected cell fate can change from predominantly lytic (at ) to predominantly lysogenic (at ) is not necessarily a hallmark of an underlying bistable viral GRN nor of a bifurcation in the underlying dynamics as a function of or . Despite the agreement with prior empirical studies, note that our model does not predict systematic decreases in the lysogen fraction given a fixed value of and increasing values of , as observed in a recent single-cell experimental study [@pcbi.1002006-Zeng1]. In the next section, we revisit the experimental data from Zeng *et. al.* [@pcbi.1002006-Zeng1] and in so doing, provide an alternative data collapse and a corresponding mechanism that is consistent with a modified version of the current stochastic model. ::: {#pcbi-1002006-g004 .fig} 10.1371/journal.pcbi.1002006.g004 Figure 4 ::: {.caption} ###### Response of phage to various phage genome concentrations when (A) asymptotically divergent and (B) transiently divergent. and represent the number of coinfecting phages and the host cell volume, respectively, so is the phage genome concentration. Each point is the result from 5,000 simulations. :::  ::: Mechanism of partial gene dosage compensation accounts for observed heterogeneity in lysis-lysogeny decisions {#s2d} ------------------------------------------------------------------------------------------------------------- Zeng *et. al.* [@pcbi.1002006-Zeng1] measured the fate of multiply infected cells in which the number of phages and cell volume could be measured on a per-cell basis. The experimental protocol induces viral injection with an abrupt change in temperature and hence, infections are treated as simultaneous. The experimental data demonstrate that the fraction of lysogeny increases with viral concentration, ([Fig. 5](#pcbi-1002006-g005){ref-type="fig"} (A)). This trend agrees with prior experimental works showing that increases in co-infection number increases the likelihood of lysogeny [@pcbi.1002006-Kourilsky1], [@pcbi.1002006-Kobiler1] and that increases in cell volume increases the likelihood of lysogeny [@pcbi.1002006-StPierre1]. However, there is significant amount of heterogeneity in the observed cell fate data other than strict dependence on as suggested by theory [@pcbi.1002006-Weitz1]. ::: {#pcbi-1002006-g005 .fig} 10.1371/journal.pcbi.1002006.g005 Figure 5 ::: {.caption} ###### Alternative mechanisms underlying heterogeneity of lysis-lysogeny decisions. \(A) Fraction of lysogeny plotted from single cell assays[@pcbi.1002006-Zeng1]. (B) Rescaled probability of . Each phage within a host is completely independent from other phages, and decision of lysogeny becomes a function of host volume. Note that rescaled curves do not collapse into a single curve. (C) Rescaled probability of proposed by Zeng *et. al.* [@pcbi.1002006-Zeng1] representing the probability of lysogeny for each individual infecting phage. Each phage independently "chooses" lysis or lysogeny. However, since the fraction of lysogeny for a single phage is a function of , phages sense the presence of other phages. Note that data from different -s collapse into a single curve. (D) Probability of lysogeny plotted against rescaled when , corresponding to a mechanism in which gene expression from multiple copies is partially compensated. Due to partial dosage compensation, the transcription rate is not linearly proportional to , and the effective copy number is given as where . Note that the data from different -s collapse into a single curve. Black lines represent nonlinear curve fits into Hill functions. :::  ::: In particular, Zeng *et. al.* [@pcbi.1002006-Zeng1] observed that the fraction of lysogens decreases with increasing for a given ratio of . Zeng *et. al.* [@pcbi.1002006-Zeng1] suggested that the remaining heterogeneity in cell fate not explained by a strict dependence on is due to a voting mechanism that takes place at the single-cell level. In this view, a unanimous decision of phages is required by phages for lysogeny [@pcbi.1002006-Zeng1] (presumably because a single phage that is fated to lysis would over-ride a decision by other phages for lysogeny). If each coinfecting phage is totally independent from each other, then one would expect the probability of lysogeny to be: where is the probability that a cell of volume infected by a single phage would become a lysogen. [Fig. 5](#pcbi-1002006-g005){ref-type="fig"} (B) shows the fraction of lysogeny scaled with power based on the empirical observations for the singly infected case. The re-scaled data for the five values of should agree with in an independent phage voting model. However, this rescaling does not form a single line. This suggests that there might be some inter-dependence between phages. Indeed, the voting model proposed by Zeng *et. al.* [@pcbi.1002006-Zeng1] is actually a "quasi-independent" voting model. In this view, a unanimous decision of phages is required by phages for lysogeny [@pcbi.1002006-Zeng1]. However, the probability that any given phage decides for lysogeny becomes a function of the viral genome concentration, . Thus the fraction of lysogeny becomes where is the probability that a single phage reaches a lysogenic decision state given that it is in a cell of volume with a total of phages. The re-scaled probability of entering lysogeny at the whole cell level, , is shown in [Fig. 5](#pcbi-1002006-g005){ref-type="fig"} (C). Notably, the re-scaled experimental data collapses on a single line, presumably . Thus, this mechanism captures the characteristics of experimental data phenomenologically. However, the mechanism involves both independence and inter-dependence among phage genomes that remains un-identified at the subcellular level. Here, we revisit the cell fate data of Zeng *et. al.* [@pcbi.1002006-Zeng1] and propose a mechanism of partial gene dosage compensation as an alternative explanation for the scaling collapse they observe. In this context, partial gene dose compensation means that a cell with multiple copies of a viral genome has smaller per-copy viral gene expression than a cell with a single viral genome. Indirect support already exists for this hypothesis. For example, Zeng *et. al.* [@pcbi.1002006-Zeng1] showed that the fraction of cells with halted growth increases with the number of co-infections, suggesting that viral genomes have adverse effects on cellular metabolism in addition to or instead of lysis. Earlier studies showed that phage infections repress host synthesis activity at the level of transcription [@pcbi.1002006-Terzi1] and translation [@pcbi.1002006-Howes1]. The degree of repression depends on the number of coinfections, and more coinfections lead to greater repression. Broadly speaking, the mechanism (or mechanisms) underlying gene dosage compensation remains an open question. However, it has been widely noted that copy numbers of genes and chromosomes can differ among cells and individuals, but the resulting gene expression need not be a linear function of gene copy number [@pcbi.1002006-Birchler1]--[@pcbi.1002006-Springer1]. Here, we assume that partial gene dosage compensation occurs at the level of transcription. Specifically, we assume that the total transcription rate of a gene is proportional to where (see [Models](#s4){ref-type="sec"} and Eq. (3)). is the quantitative measure of partial gene dosage compensation and RNA synthesis repression by phage genomes. When , increases in viral genome have no effect on transcriptional rates, whereas when , transcriptional rates increase linearly with (as in the original model described previously in this paper). Hence, if a partial gene dosage mechanism is at work, then the lysogeny data should collapse when plotted against . [Fig. 5](#pcbi-1002006-g005){ref-type="fig"} (D) shows the fraction of lysogeny against which incorporates the effect of partial gene dosage compensation. Note that the data collapses into a single line, similar to the quasi-independent decision mechanism. The estimate of from experimental data is about 0.5, suggesting that the overall viral transcriptional activity in the host cells on a per-viral genome basis scales with . Hence, two distinct mechanisms: (i) quasi-independent decision making; and (ii) partial gene dosage compensation, can explain heterogeneous decision making from single cell assay based experiments using data collapse. Note that we cannot evaluate the quasi-independent mechanism using our model because doing so would require incorporating genome-specific changes (such as anti-termination events) or compartmentalizing the cell with respect to transcription and translation events (requiring even more unknown parameters than the current model). However, it is possible to explicitly incorporate partial gene dosage compensation in stochastic simulations (see [Models](#s4){ref-type="sec"}). In brief, we modified transcriptional rates so that transcription increased with instead of and ran stochastic simulations with all other parameters as before. [Fig. 6](#pcbi-1002006-g006){ref-type="fig"} shows the fraction of lysogeny resulting from the stochastic fate determination model incorporating partial gene dosage compensation against and rescaled . Stochastic simulations with partial dosage compensation exhibit the heterogeneous, yet strong dependence of lysogeny on . Moreover, the cell fate results of stochastic simulations collapse into a single line when is rescaled as . Given the new scaling collapse, cells with the same have a lower chance of lysogeny given increasing values of , consistent with the pattern observed in the experimental study ([Fig. 5](#pcbi-1002006-g005){ref-type="fig"} (D)). Hence, we propose that partial gene dosage compensation should be considered as an alternative mechanism to explain the heterogeneous cell fate of bacteria infected by bacteriophage . ::: {#pcbi-1002006-g006 .fig} 10.1371/journal.pcbi.1002006.g006 Figure 6 ::: {.caption} ###### Effect of gene dosage compensation from stochastic simulations. \(A) Fraction of lysogeny from stochastic simulations. Simulations with partial dosage compensation exhibit the nested pattern of dependence as seen in the experimental data (see [Fig. 5](#pcbi-1002006-g005){ref-type="fig"}). (B) Simulation results on the fraction of lysogeny from Fig. 6 (A) plotted with rescaled when . The outcome of stochastic simulations with partial dosage compensation is consistent with experimental data (see [Fig. 5](#pcbi-1002006-g005){ref-type="fig"} (A,D)). In this case, the GRN is asymptotically driven with CI and Q threshold at 100 nM and 120 nM, respectively, all other parameters are set according to [Table 1](#pcbi-1002006-t001){ref-type="table"} - transiently divergent. Each point is the result from 3,000 simulations. :::  ::: Discussion {#s3} ========== In this paper, we have proposed and analyzed a transient mechanism of cell fate determination in terms of first passage processes of regulatory proteins. We applied this mechanism to the study of the initial lysis-lysogeny decision in bacterial cells infected by phage . We found that stochastic simulation of parametrized viral GRNs lead to changes in the the frequency of alternative fates for infected cells, either lysis or lysogeny, as a function of the genome concentration of infecting viruses. The biased response in cell fate outcome occurs despite intrinsic noise in the system and does not require the bistability of the underlying GRN. Hence, alternative and seemingly adaptive cell fate decisions may be due to transient divergence in stochastic trajectories of regulatory molecules and not necessarily due to underlying bistability. Finally, we showed that a partial gene dosage compensation is a candidate mechanism underlying noise in lysis-lysogeny decisions, as supported by both our quantitative model and experimental data. Our central result is in contrast to the conventional perspective that multistability is required for alternative decisions [@pcbi.1002006-Losick1], [@pcbi.1002006-Xiong1]. Multistability often requires cooperative binding as a necessary condition for the emergence of the two or more stable steady states in the GRN [@pcbi.1002006-Gardner1], [@pcbi.1002006-Cherry1]. A recent study showed that a switch system can arise in the absence of cooperative bindings [@pcbi.1002006-Lipshtat1]. Our study suggests that cooperative binding may occur and affect transient dynamics but not necessarily lead to bistability in asymptotic dynamics. Together these results suggest that GRNs which do not have bistability or cooperative bindings might be able to lead to alternative cell fate determinations. Thus, it might be possible for a GRN to evolve (by natural selection) or to be designed (via synthetic means) to perform a complex task of alternative decision making in response to external stimuli without multistability. Note that such a transiently excitable GRN which differentiates transient and asymptotic phenotypes was experimentally demonstrated in *Bacillus subtilis* [@pcbi.1002006-Sel1]. Generally, there exist examples of GRNs which are responsive to environmental signals and robust to changes of kinetic parameters [@pcbi.1002006-Ziv1] while other are sensitive to kinetic parameters. Sensitivity of transient dynamics to a GRN\'s kinetic parameters and thresholds might be a target of selection over evolutionary time scales. In this context, we examined how modifying thresholds for decisions can lead to systematic changes in lysis vs. lysogeny as well as decision times (see Fig. S2). The general result from the present analysis is that alternative determination requires separation of thresholds, which comes at the expense of slower decisions. Hence, transiently driven cellular decisions have the potential to be highly evolvable. As we have detailed, stochastic simulations of the phage GRN proposed here can reproduce a number of characteristics for how the fraction of lysogeny changes with and cell volume. Importantly, we find that lysogeny increases with increasing [@pcbi.1002006-Kourilsky1], [@pcbi.1002006-Kobiler1] and decreasing cell volume [@pcbi.1002006-StPierre1], and remains between approximately 20%--90% for physiologically reasonable values [@pcbi.1002006-Zeng1]. The bias in cell fate outcome in favor of lysogeny with increasing may be adaptively significant. On average, high implies that phages infect hosts frequently on the time-scale of decision-making and further, that phages are more abundant than their bacterial hosts. Lysis will further increase the phage-host ratio, and a previous study has speculated that phages seem to avoid depletion of hosts by entering lysogeny predominantly at high [@pcbi.1002006-Stewart1]. However, if lysogeny is adaptively favorable at high , why is it that a small fraction of phages still enter the lytic pathway? The answer could be due to constraints in the resolvability of the GRN due to the strength of intrinsic stochasticity in the GRN [@pcbi.1002006-Zeng1]. Or the stochasticity itself may be adaptive. Phages may have evolved to respond to changes in intracellular phage genome concentration in order to minimize the chance of extinction [@pcbi.1002006-Avlund1] by maintaining phage and lysogen population as a bet-hedging strategy [@pcbi.1002006-Veening1]. Any such speculations require careful consideration of selective pressures imparted by ecological dynamics, game theoretic issues arising from co-infections by non-identical strains, and biophysical constraints and trade-offs arising at the intracellular scale [@pcbi.1002006-Gudelj1]. However, the first set of stochastic simulations of the phage GRN presented in this manuscript fail to predict the systematic decrease in the fraction of lysogeny given a fixed value of and increasing values of [@pcbi.1002006-Zeng1] (see [Fig. 4](#pcbi-1002006-g004){ref-type="fig"}). We revisited the original single-cell data and demonstrated the existence of an alternative scaling collapse owing to a proposed partial gene dosage compensation mechanism. When we incorporate partial gene dosage compensation within our stochastic model, we are able to recover the alternative scaling collapse consistent with the empirical measurements of Zeng *et. al.* [@pcbi.1002006-Zeng1] (see [Fig. 5(D)](#pcbi-1002006-g005){ref-type="fig"} and 6(B)). What might cause partial dosage compensation to occur in multiple infected cells? In stochastic simulations here, dosage compensation is modeled explicitly at the transcriptional level, whereas in reality multiple factors can contribute to it, and may occur at both transcriptional and post-transcriptional levels. The degree of compensation might change depending on copy numbers of genes and chromosomes as well as other intracellular factors. Copy number variation (CNV) is common in biological organisms [@pcbi.1002006-Perry1], [@pcbi.1002006-DeLuna1], and previous studies suggested that gene expression can depend sensitively on CNV when uncompensated [@pcbi.1002006-Mileyko1]. Indeed, one hypothesis is that gene regulatory networks have been selected for their lack of dosage sensitivity to avoid problems in gene expression that may arise when CNV occurs naturally [@pcbi.1002006-SchusterBckler1]. Previous studies showed that phage represses overall activity of RNA and protein synthesis within infected hosts depending on the number of coinfections [@pcbi.1002006-Terzi1], [@pcbi.1002006-Howes1]. Viruses are known to control host cell cycle in eukaryotic cells [@pcbi.1002006-BenIsrael1], but how viruses affect the overall host transcriptional and translational activity in bacterial hosts is an open question. We believe that elucidating intracellular mechanisms of gene dosage compensation would be an important step toward understanding CNV and its resulting change in gene expression, at both the transient and steady state. In doing so, we also hope to provide a cautionary note: deducing explicit mechanisms from data collapses can be difficult, particularly when multiple data collapse schemes are consistent with observations. In summary, this study proposed a novel intracellular decision-making mechanism to explain the variability in cell fate determination in multiply infected hosts. However, there can be other sources of variability underlying the lysis-lysogeny decision switch. First, the viral concentration, , in naturally infected hosts may be dynamic. Multiple phages infect a host sequentially, and a host can keep growing while being infected. Subsequent infections increase over time, and infected cells may spend a substantial fraction of the time prior to cell fate determination with a value of which is smaller than the final . Next, host cell growth decreases whereas viral genome replication increases during the infection cycle. Clearly the dynamic nature of viral genome concentration needs to be addressed even if experimental protocols have been designed to synchronize infections. Second, despite our incorporation of stochasticity in the model, we assume the bacterial cytoplasm is well-mixed. Previous studies demonstrated that bacterial DNA, RNA and proteins have spatial patterns [@pcbi.1002006-Sherratt1]--[@pcbi.1002006-Thanbichler1]. Bacteriophages are known to target cellular poles of hosts preferentially [@pcbi.1002006-Edgar1] which suggests phage genomes might be localized within bacterial cytoplasm. Hence, cell fate decision may be determined by local concentrations of regulatory proteins and quasi-independent cell fate determination by each virus. Finally, we assumed decision making as strict first passage processes arising from the consideration of thresholds as absorbing states within a GRN dynamics. It is possible that decision making involves soft thresholds over which cells make decisions with some probability. There are studies which show duration of signals is critical to cellular decisions [@pcbi.1002006-Marshall1], [@pcbi.1002006-Ebisuya1], and there might be some minimum time interval during which the system is above a threshold to make a decision [@pcbi.1002006-Mangan1]. Even if experimental protocols can minimize the impact of one of these mechanisms, the evolution of the phage GRN would surely be impacted by all of them. Progress in identifying the importance of each of these issues at the molecular and evolutionary scales is relevant not only to the study of transient fate determination in phage , but to the study of cellular decision making in general. Models {#s4} ====== Gene regulation in phage {#s4a} ------------------------ The fate of *E. coli* cells infected by phage are decided soon after infection by a set of so-called early viral genes [@pcbi.1002006-Ptashne1]. Among them we consider four genes, *cI, cro, cII* and *Q*, and one antisense mRNA (*aQ*) (see [Fig. 1](#pcbi-1002006-g001){ref-type="fig"} (A)). The expression of these genes are controlled by four promoters, , , and . and share three operator sites which are targeted by CI and CRO. The natural form of CI is a dimer, and CI dimers act as self activators and repressors for other genes by binding to . CII tetramers can bind to to transcribe mRNA and to produce CI [@pcbi.1002006-Parua1]. Dimers of CRO bind to to inhibit all the genes in the system ([Fig. 1](#pcbi-1002006-g001){ref-type="fig"} (B)). Immediately after phage infections there are no viral gene products. At this initial stage is active which leads to an increase of CRO, CII and Q levels. If Q becomes sufficiently abundant, it will turn on genes which make progeny phages, and the infected host will be lysed. However, as CII concentration increases CII tetramers can activate CI transcription from , and CI expression level become further enhanced by the positive feedback loop of CI at . CII also represses Q by transcribing *aQ* which facilitates *Q* mRNA degradation, and sufficiently high CI level leads to lysogeny [@pcbi.1002006-Ptashne1]. Hence, lysis or lysogeny is determined based on which of either CI and Q reaches the threshold concentration first. When CI reaches its threshold, CI dimers begin to form tetramers and octamers which lead to DNA looping [@pcbi.1002006-Rvet1]. DNA looping is very stable while maintaining lysogeny and repressing genes which trigger lysis [@pcbi.1002006-Morelli1]. When Q reaches its threshold, a group of late genes responsible for making progeny phages will be turned on, and the host will eventually be lysed. Since translation occurs with a single protein at a time, simultaneous crossings of lytic and lysogenic thresholds are forbidden, and the decisions are mutually exclusive. In reality, decisions would not be triggered by infinitesimally short bursts over decision thresholds, but for simplicity we assume a decision is made when either CI or Q concentration reaches its threshold for the first time. The use of step functions instead of Hill function type responses has been used extensively in the study of quantitative gene regulatory networks [@pcbi.1002006-Alon1]. Note that when phages multiply infect cells in natural settings, they do not do so simultaneously, and so increases sequentially in time. However, for simplicity we only consider simultaneous coinfections, for which becomes a parameter in determining cell fate rather than a dynamic variable. This choice of modeling simultaneous infections is also motivated by the the experimental protocol of Zeng *et. al.* [@pcbi.1002006-Zeng1] in which rapid temperature changes were used to synchronize phage infection of DNA into host genomes. Quantitative model of phage decision switch {#s4b} ------------------------------------------- Here we express the interactions among *cI, cro, cII* and *Q* as well as *aQ* mRNA described in the previous section as a set of ordinary differential equations. If we apply quasi-steady-state approximation for dimers and tetramers, the system can be described as where , , and represent the total concentration of CI, CRO, CII and Q, respectively. represents the number of coinfecting phages while is the cell volume. represents the mRNA concentration, and denotes the degradation rate where each subscript represent the species of associated gene/protein. *Q* and *aQ* mRNA become degraded by binding to each other and the adsorption rate is denoted as . , and represent the basal, CI-mediated and CII-mediated transcription rates with subscripts indicating the species of mRNA. Note that , and is inversely proportional to since the concentration change by a transcription event is proportional to . We assume that the concentrations of dimers and tetramers are at quasi-steady states such as where the subscripts 1, 2 and 4 represent the concentration of monomers, dimers, and tetramers of each respective protein. and are the dimerization and tetramerization constants, respectively. , , and in Eq. (3) denote the probability of transcribable configurations for each promoters based on free energy change of possible states, and we follow the calculation of Shea and Ackers for and [@pcbi.1002006-Shea1] and Arkin *et. al.* for [@pcbi.1002006-Arkin1] (see Supplementary [Text S1](#pcbi.1002006.s001){ref-type="supplementary-material"}). has two modes of transcription denoted as basal and activated depending on . Response of is a first order Hill function which is For stochastic simulations, we chose two parameter sets which lead to a transiently and asymptotically divergent lysis-lysogeny decision switch. Parameter values for the transiently divergent and asymptotically divergent cases are listed in [Table 1](#pcbi-1002006-t001){ref-type="table"}. To calculate the fraction of lysogeny, we used at least 3,000 realizations of a stochastic model. Our simulations are based on Eq. (3) and are fully stochastic as implemented using the Gillespie algorithm [@pcbi.1002006-Gillespie1] (see Supplementary [Text S1](#pcbi.1002006.s001){ref-type="supplementary-material"} for details). Modeling gene dosage compensation {#s4c} --------------------------------- When gene dosage is compensated, the effective copy number, which is the fold change of transcription rate, is smaller than the actual copy number. Here we assume the effective copy number scales as where . When , the system is completely compensated without any copy number dependence. On the contrary, when , transcription rate is linearly proportional to the copy number. The experimental data ([Fig. 5](#pcbi-1002006-g005){ref-type="fig"} (A)) supports that is between 0.4 and 0.6. For stochastic simulations, we replace all the terms of in Eq. (3) with , and set . Supporting Information {#s5} ====================== Text S1 ::: {.caption} ###### Additional details on methods. (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: We thank Ido Golding and Lanying Zeng for sharing experimental data and for their comments on the manuscript. The authors have declared that no competing interests exist. This work was supported by the James S. McDonnell Foundation and the Defense Advanced Projects Research Agency under grants HR0011-05-1-0057 and HR0011-09-1-0055. Joshua S. Weitz, Ph.D. holds a Career Award at the Scientific Interface from the Burroughs Wellcome Fund. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: RIJ JSW. Performed the experiments: RIJ. Analyzed the data: RIJ JSW. Contributed reagents/materials/analysis tools: RIJ JSW. Wrote the paper: RIJ JSW.

PubMed Central

2024-06-05T04:04:19.674501

2011-3-10

PMC3053318

Introduction {#s1} ============ Transcriptional and translational regulatory networks control the phenotype of modern cells, regulating gene expression in response to changing environmental conditions and/or biological stimuli. It has been well established that intrinsic noise in gene regulation results from the discrete biochemical nature of the process [@pcbi.1002010-Thattai1]. There is also an extrinsic component to the total noise arising from cell-to-cell variation in the number of copies of the transcription and translation machinery (transcription factors, RNA polymerases, ribosomes, etc) [@pcbi.1002010-Elowitz1]--[@pcbi.1002010-Paulsson1]. Stochastic noise can lead to different phenotypic outcomes for a cellular population and, in certain fluctuating environments, the resulting heterogeneous population can be more optimal for growth than would be a population containing a single phenotype [@pcbi.1002010-Thattai2], [@pcbi.1002010-Acar1]. Theoretical modeling of stochasticity in gene expression has been a topic of intense study in the last decade and has greatly increased our understanding of the effect that statistical noise has on gene regulation (for reviews see [@pcbi.1002010-Paulsson2]--[@pcbi.1002010-Cheong1]). Without detailed information regarding spatial heterogeneity within a cell, models of stochastic gene expression are typically expressed in terms of the chemical master equation (CME), which describes the time evolution of the probability for a chemical system to be in a given state [@pcbi.1002010-McQuarrie1]. Various analytical methods including moment generating functions [@pcbi.1002010-Thattai1], [@pcbi.1002010-Swain1], [@pcbi.1002010-Paulsson3], the Langevin and Fokker-Planck equations [@pcbi.1002010-Hasty1], linear noise approximation [@pcbi.1002010-Paulsson1], and many-body theory [@pcbi.1002010-Sasai1] are used to study such models of gene expression. Computer simulations, usually based on a variant of Gillespie\'s stochastic simulation algorithm (SSA) [@pcbi.1002010-Gillespie1] are also widely employed to analyze gene network models that are too complex to be amenable to analytical modeling [@pcbi.1002010-McAdams1], [@pcbi.1002010-Arkin1]. Such theoretical studies have predicted and experimental measurements have shown [@pcbi.1002010-Elowitz1], [@pcbi.1002010-Ozbudak1]--[@pcbi.1002010-Raj2] that populations of cells can be quite heterogeneous, even when starting from an initially identical state. The large variance in the population distribution is usually ascribed to bursting in the process of gene transcription. Two models have been developed which can be used as a framework for quantitatively analyzing population distributions to infer the underlying gene expression kinetics. The burst model ([Figure 1A](#pcbi-1002010-g001){ref-type="fig"}) of Friedman *et al.* [@pcbi.1002010-Friedman1] is based on the assumption that an mRNA\'s lifetime is short compared with that of its protein product. In that case, proteins will be produced in independent bursts with exponentially distributed sizes. The solution to the stationary probability distribution of protein in the continuous CME formulation of the model was shown to be the Gamma distribution where the and parameters were interpreted to be the frequency of transcriptional bursts relative to the protein lifetime and the mean number of proteins produced per burst, respectively. ::: {#pcbi-1002010-g001 .fig} 10.1371/journal.pcbi.1002010.g001 Figure 1 ::: {.caption} ###### Three models for stochastic gene expression. \(A) Burst model in which transcription of the DNA is always active. (B) Two-state model in which the DNA switches with constant rates between active and repressed states. (C) Inducible genetic switch in which an inducer both controls the rate of switching between active and inactive transcription states and is also positively regulated by the protein product -- a positive feedback loop (PFB). The gray dotted connection indicates a weak effect of the inducer in promoting the unbinding of repressor at high inducer concentrations. :::  ::: Shahrezaei and Swain [@pcbi.1002010-Shahrezaei2] further developed the analytical theory of gene expression, by deriving not only the time-dependent probability distribution for the burst model, but also the steady-state distribution for a two-state model of gene expression ([Figure 1B](#pcbi-1002010-g001){ref-type="fig"}; three-stage model in their nomenclature). In the two-state model a gene alternates between transcriptionally active and inactive states with constant rates. Their analytical distributions show that in addition to large variance within a population, bimodality can appear when transitions between the active and inactive states are slow. A similar model has also been used to analyze the switching behavior of a population due to rare large events versus the cumulative effect of many small events [@pcbi.1002010-Choi2]. Computational modeling can greatly assist in understanding genetic systems where complexity exceeds the capacity of analytical solutions. In a model for an inducible genetic switch incorporating more of the complexity present in real biological systems ([Figure 1C](#pcbi-1002010-g001){ref-type="fig"}), the transitions between the active and inactive transcriptional states are no longer constant but depend upon an external inducer likely in a nonlinear manner. The positive feedback (PFB) loop changes the network topology by introducing an additional regulatory link. Both of these differences provide additional sources of noise in the circuit that may affect the probability distributions. Combining computer modeling of a complete genetic circuit with analysis using simplified analytical models can help to provide an overall picture of the dynamics of such a system. Further complexity in modeling real biological systems comes from the spatial heterogeneity within a cell and molecular crowding in the *in vivo* environment. It is becoming apparent that the cell is not a well-stirred system [@pcbi.1002010-vanZon1]--[@pcbi.1002010-Takahashi1]. Studies using cryoelectron tomography techniques [@pcbi.1002010-Ortiz1]--[@pcbi.1002010-Khner1] have revealed that individual macromolecules are not necessarily uniformly distributed inside the cell, but may be clustered in a spatially dependent manner. Spatial organization can affect reaction kinetics by increasing local concentrations of reactants and enzymes. Additionally, crowding and non-specific molecular interactions in the *in vivo* environment can lead to anomalous subdiffusive behavior for macromolecules, as measured experimentally [@pcbi.1002010-Banks1], [@pcbi.1002010-Golding1] and by computational modeling of bacterial cytoplasmic environments [@pcbi.1002010-Roberts1]--[@pcbi.1002010-McGuffee1]. Accounting for spatial heterogeneity is a challenge to computational biology that must eventually be met and several such modeling studies have been undertaken [@pcbi.1002010-Roberts1]--[@pcbi.1002010-Arjunan1]. Stochastic modeling of gene expression circuits in a three-dimensional bacterial cell poses several difficulties, both computational and informational in nature. Recently a "lattice microbe" method [@pcbi.1002010-Roberts1] was developed using GPU (graphics processing unit) computational accelerators to simulate diffusion of macromolecules within a modeled *Escherichia coli* cell packed with a distribution of obstacles according to reported proteomics data. It implemented a multiparticle reaction-diffusion algorithm on a three-dimensional lattice to perform simulations of cell-scale systems. With the lattice microbe method one can observe anomalous diffusion of macromolecules and track diffusive-reactive processes over the timescale of the cell cycle, with spatial resolution from 2--16 nm. On the informational side, painstaking efforts must be undertaken to obtain parameters for the models. Kinetic parameters, which are often obtained under *in vitro* conditions, must be validated by comparing modeling results to published experiments. Recent time-lapse fluorescence microscopy experiments have been able to track dynamic behavior for individual macromolecules *in vivo* [@pcbi.1002010-Yu1], [@pcbi.1002010-Golding2], providing an additional source for model parameters. Parameters obtained from *in vivo* single-molecule experiments are uniquely suited for stochastic modeling, as they provide population distributions not simply mean values from ensemble measurements. Equally importantly such parameters are measured under *in vivo* conditions and incorporate the effects of the cellular environment. Also, super-resolution imaging studies [@pcbi.1002010-Yildiz1]--[@pcbi.1002010-Biteen1] provide further spatial information to complement the cryoelectron tomography data. We present here a computational study of gene expression noise in the inducible genetic switch shown in [Figure 1C](#pcbi-1002010-g001){ref-type="fig"} using both well-stirred and spatially resolved models. Spatial models of *E. coli* cells were constructed to approximate cytoplasmic crowding under both rapid and slow growth phenotypes, with the latter being based on data from cryoelectron tomography [@pcbi.1002010-Ortiz2]. Both spatial models were simulated using the lattice microbe method [@pcbi.1002010-Roberts1]. The genetic switch was based on the well-characterized *E. coli* lactose utilization system, parameterized using measurements from a recent series of *in vivo* single-molecule fluorescence studies [@pcbi.1002010-Yu1], [@pcbi.1002010-Choi1], [@pcbi.1002010-Elf1] as well as published *in vitro* rate constants. We report the contributions to intrinsic noise from the regulatory elements of the inducible genetic circuit as well as the extrinsic noise due to *in vivo* crowding. Using the slow-growth model we investigate the effect of using experimentally determined cellular architecture in reaction-diffusion models, with implications for effects due to cell growth. Comparing the noise from the inducible genetic switch to the bursting and two-state models described above ([Figure 1A,B](#pcbi-1002010-g001){ref-type="fig"}), we consider what improvements in both modeling and experimental efforts are needed to develop stochastic models of gene expression with predictive power regarding phenotype switching and heterogeneity in cellular populations. Methods {#s2} ======= Lac circuit kinetic model {#s2a} ------------------------- ### Lactose uptake in *Escherichia coli* {#s2a1} The lactose utilization system in *E. coli* is a model system for studying inducible genetic circuits [@pcbi.1002010-Wong1]--[@pcbi.1002010-Noel1]. The overall genetic system is described in [Figure 2](#pcbi-1002010-g002){ref-type="fig"}. Briefly, the *lac* repressor (LacI; in model annotation) [@pcbi.1002010-Bell1]--[@pcbi.1002010-Xu1] binds to the *lac* operator ( in model annotation) upstream of the DNA encoding for the genes responsible for lactose uptake and metabolism, repressing their expression in the absence of lactose. In the presence of lactose or another inducer ( in model annotation), LacI binds the inducer preferentially and is prevented from binding to the operator region allowing expression of the proteins in the *lac* operon. One protein in the operon, lactose permease (LacY; in model annotation), establishes positive feedback in the circuit by inserting into the membrane and actively transporting lactose into the cell, ensuring that LacI remains sequestered; the cell switches to the induced state. Theoretical and experimental studies have investigated the behavior of *lac* system and shown it to be stochastic, depending on random fluctuations to switch between the off and on states [@pcbi.1002010-Mettetal1], [@pcbi.1002010-Stamatakis1]. We assumed the same overall kinetic structure for our model of the *lac* system as Stamatakis and Mantzaris [@pcbi.1002010-Stamatakis1], but where possible derived the rate parameters from single molecule *in vivo* experiments. The reactions and stochastic rate constants, derived using a well-stirred approximation, are given in [Table 1](#pcbi-1002010-t001){ref-type="table"}. ::: {#pcbi-1002010-g002 .fig} 10.1371/journal.pcbi.1002010.g002 Figure 2 ::: {.caption} ###### Overview of the *lac* genetic circuit in *E. coli*. \(A) In the absence of inducer, the lac repressor (LacI) binds to the *lac* operator preventing transcription of genes in the *lac* operon. (B) Following an increase in the extracellular inducer concentration, inducer enters the cell via both diffusion across the membrane and active transport by lactose permease (LacY). Once inside, inducer binds free LacI molecules preventing them from binding to the operator. (C) After the intracellular inducer concentration reaches a threshold, any bound repressor is "knocked-off" the operator leading to expression of the *lac* genes. (D) At high intracellular inducer concentrations the genes for lactose metabolism are fully induced. (E) After inducer is removed, repressor rebinds to the operator preventing further expression of the *lac* operon and the enzymes for lactose metabolism are either degraded or diluted through cellular division. :::  ::: ::: {#pcbi-1002010-t001 .table-wrap} 10.1371/journal.pcbi.1002010.t001 Table 1 ::: {.caption} ###### Reactions and rate constants used in the stochastic model of the *lac* circuit. ::: {#pcbi-1002010-t001-1} Reaction Param Stochastic Rate Units Source[a](#nt101){ref-type="table-fn"} Pub. *in vitro* Rate ------------------------------------------------- ------- ----------------- ---------- ---------------------------------------- ------------------------------------------------- ------ --------------------------------------------- **Lac operon regulation** 2.43e+06 *M* 4.0-20.0e+08[b](#nt102){ref-type="table-fn"} 1.21e+06 *M* -- 2.43e+04 *M* -- 6.30e-04 *S* 1.4-2.3e-02[b](#nt102){ref-type="table-fn"} 6.30e-04 *S* -- 3.15e-01 *M* -- **Transcription, translation, and degredation** 1.26e-01 *M* -- 4.44e-02 *S* -- 1.11e-02 *S* -- 2.10e-04 *M* -- **Inducer--repressor interact.** TMG IPTG TMG IPTG IPTG 2.27e+04 9.71e+04 *M* *K* 9.2-9.8e+04[c](#nt103){ref-type="table-fn"} 1.14e+04 4.85e+04 *M* *K* 4.6-4.9e+04[c](#nt103){ref-type="table-fn"} 6.67e+02 2.24e+04 *M* *K* 2.0-2.3e+04[c](#nt103){ref-type="table-fn"} 3.33e+02 1.12e+04 *M* *K* 1.0-1.2e+04[c](#nt103){ref-type="table-fn"} 2.00e-01 *K* 2.0e-01[c](#nt103){ref-type="table-fn"} 4.00e-01 *K* 4.0e-01[c](#nt103){ref-type="table-fn"} 1.00e+00 *K* 0.5-1.0e+00[c](#nt103){ref-type="table-fn"} 2.00e+00 *K* 1.0-2.0e+00[c](#nt103){ref-type="table-fn"} **Inducer transport** 2.33e-03 *K* 2.3e-03-1.4e-01[d](#nt104){ref-type="table-fn"} 2.33e-03 *K* 2.3e-03-1.4e-01[d](#nt104){ref-type="table-fn"} 3.03e+04 *K* -- 1.20e-01 *K* -- 1.20e+01 *K* 1.2e+01[e](#nt105){ref-type="table-fn"} a *S* = *in vivo* single molecule experiment, *K* = *in vitro* (kinetic) experiment, *M* = model parameter fit to single-molecule distributions. b [@pcbi.1002010-Goeddel1], c [@pcbi.1002010-OGorman1], [@pcbi.1002010-Dunaway1], d [@pcbi.1002010-Maloney1], [@pcbi.1002010-Chung1], e [@pcbi.1002010-Dornmair1]. ::: ### Lac operon regulation: Activation and inactivation of transcription {#s2a2} The regulatory behavior of the *lac* circuit results from the binding of the repressor to the *lac* operator, thereby inhibiting transcription initiation. There are three possible inducer--repressor species and we modeled the binding and unbinding reactions of each to the operator: The stoichiometry of inducer--repressor binding is currently subject to debate [@pcbi.1002010-Oehler1]; it is unclear whether the affinity of for the operator is of the same order as that of or much lower. We therefore compared the effect on our model of both a high (comparable to ) and a low (). In either case, the affinity of for the operator is thought to be low and we assumed and . Values for the rate constants were obtained by fitting the model with experimental LacY distributions from single cells, as presented in [Results](#s3){ref-type="sec"}. ### Transcription, translation, and degradation {#s2a3} In the cell both transcription and translation are multistep processes involving numerous intermediates. Knowledge of the rate constants for each step of these processes *in vivo* is limited. We therefore assumed that each process was controlled by a single rate-limiting event and used pseudo first order rate equations with effective transcription and translation rates in the model. Such an approximation is reasonable as long as i) the concentration of the transcription and translation machinery is high and constant and ii) the non-Poissonian time delays missed by modeling such multi-step processes as a single step are not significant. For our model these conditions appear to be satisfied as the components of transcription and translation are among the most abundant in the cell and even though some, such as ribosomes, diffuse quite slowly there should nevertheless be a ready supply available at all times. Also, the times delays in these processes are on the order of seconds [@pcbi.1002010-vanZon1], while the dynamics of the cell response (here determined by the protein lifetime) is on the order of an hour. Transcription of a LacY messenger RNA (mRNA) () from the *lac* operon was modeled as a first order process dependent on a free operator: The effective transcription rate constant () was a free parameter determined during model fitting. Decay of and translation from were modeled as a competition between RNase E enzymes [@pcbi.1002010-Condon1] and ribosomes for an \'s ribosomal binding site (RBS). The rate of degradation of by RNase E was chosen to result in a mean lifetime () of 90 s, as reported by Yu *et al.* [@pcbi.1002010-Yu1]. The effective translation rate was chosen to produce a mean of four LacY proteins over the lifetime of an average messenger, also as reported in [@pcbi.1002010-Yu1]: with the effective rates given by and . The loss of membrane proteins in *E. coli* is primarily from dilution as a result of cellular growth over the cell cycle [@pcbi.1002010-Akiyama1]. Therefore, degradation of LacY was modeled as a first order reaction with a half-life corresponding to the cell doubling time (),where . ### Lac repressor/inducer kinetics {#s2a4} LacI rapidly dimerizes with very high affinity and the dimers further associate to form tetramers with a in the nanomolar range [@pcbi.1002010-Royer1]. The tetrameric form enhances repression by binding multiple *lac* operators simultaneously [@pcbi.1002010-Oehler2]. However, modeling the DNA loops formed by this process would require additional inactive kinetic states, and as the focus of the current study is a two-state switch, we assumed a mutant form of LacI that did not tetramerize. This assumption allowed us to connect our model to single cell data from *lac* operator mutants incapable of DNA looping [@pcbi.1002010-Choi1]. Furthermore, we assumed that the dimerization was sufficiently low that LacI only existed in the dimer state, the species . Ten molecules of were placed in the cell, and we assumed the cell regulated this number to be constant, so that the noise from the transcription/translation of the repressor gene was ignored. Although noise from expression of LacI has been shown to have an effect on the induction rate [@pcbi.1002010-Stamatakis1], we chose to ignore this effect here to focus on the noise originating in the circuit itself. The inducer molecules, isopropyl -D-1-thiogalactopyranoside (IPTG) and thiomethyl\--D-galactoside (TMG), are small sugar-like solutes that can either passively diffuse or be actively transported by LacY across the cellular membrane using an electrochemical proton gradient for energy. Inducer molecules in the extracellular space () and those in the intracellular space (,) can diffuse across the membrane freely in both directions. The diffusive influx and efflux are modeled as first order reactions with the same kinetic constants,where  =  [@pcbi.1002010-Noel1], [@pcbi.1002010-Maloney1]. The active transport of inducer molecules into the cell from the extracellular space is modeled as an irreversible Michaelis-Menten reaction, The value of has been reported to be 12 [@pcbi.1002010-Dornmair1] and the Michaelis constant for the reaction with TMG to be ∼500 [@pcbi.1002010-Kepes1]. Similarly, the intracellular TMG concentration has been reported to be ∼70-fold higher than the extracellular concentration in fully induced cells [@pcbi.1002010-Stamatakis1], [@pcbi.1002010-Kepes2]. Under pseudo steady state conditions, the ratio of to is related to by the expression . In our model a maximal enrichment ratio of 70 corresponded to a of 400 , which is the value we used for TMG. The active influx of inducer molecules through the membrane is given by the standard Michaelis-Menten expression , with and . However, for simulations of stochastic kinetics, we need unique values for and . As long as the relationship between and corresponds to the required value, the influx of inducer will be correct irrespective of the values used. Only fluctuations in the external environment will be affected by the choice. In particular, a large value for relative to would require a correspondingly large value for , which would produce many unproductive inducer binding events outside the cell. These small fluctuations in the external inducer environment are expected to have little to no effect on gene expression dynamics inside the cell, but can have a significant impact on simulation cost. In the absence of any experimental data regarding the actual kinetic rates, we chose to make the rate be 1% of the rate to improve simulation performance by minimizing the calculation of such non-productive binding events. The value of was then fixed by the relationship . Upon entering the cell, inducer molecules can bind to free LacI and the repressor-operator complex, albeit with a much lower affinity. As each LacI monomer binds a single inducer molecule, there are three possible repressor dimer species, , and , which interconvert according to the following reactions: *In vitro* kinetic data suggests non-cooperative binding (Hill coefficient of 1) of inducer to in the absence of *lac* operator DNA [@pcbi.1002010-Oehler1], [@pcbi.1002010-Ohshima1], [@pcbi.1002010-OGorman1], corresponding to and (see Supporting [Text S1](#pcbi.1002010.s001){ref-type="supplementary-material"}). Although there is some equilibrium data suggesting the binding to the complex is cooperative with a Hill coefficient of 1.45 [@pcbi.1002010-OGorman1], such cooperativity was not observed in kinetic measurements of binding and unbinding [@pcbi.1002010-Dunaway1]. For simplicity, a non-cooperative model was assumed: with and . Experimentally, different aspects of the *lac* circuit have been investigated using the inducers IPTG and TMG. To make use of the data one has to take into account the differences in their value, defined as the inducer concentration at which half of the LacI monomers are bound with an inducer. Kinetic and equilibrium binding measurements [@pcbi.1002010-OGorman1], [@pcbi.1002010-Dunaway1] were available for IPTG binding to both free repressor and the repressor-operator complex. From the kinetic measurements, the rate constants for inducer binding and unbinding were and 0.2 for the free repressor and and 1.0 for the repressor-operator complex. This yielded a for binding to the repressor-operator complex (89 ) that is ∼20 times higher than for free repressor (4.1 ). [Figure 3](#pcbi-1002010-g003){ref-type="fig"} shows the results of using these rate constants in stochastic simulations of inducer binding; good agreement between simulations and experiments were seen for both kinetic and equilibrium measurements. TMG has been reported to have a for binding to free repressor greater than that for IPTG by a factor of ∼10 [@pcbi.1002010-Barkley1]. However, since neither kinetic data nor detailed equilibrium studies were available, we assumed the same unbinding rate constants for TMG as IPTG and left binding to both the free repressor and repressor-operator complex as free parameters in the model to be fitted from single molecule experimental data. ::: {#pcbi-1002010-g003 .fig} 10.1371/journal.pcbi.1002010.g003 Figure 3 ::: {.caption} ###### Fits of rate constants for IPTG binding to the *lac* repressor. \(A) Pseudo first order rate constants observed during stochastic simulations of IPTG binding to (blue) repressor and (red) repressor-operator complex. At each inducer concentration 1000 simulations starting with 2 free (or operator-complexed) repressor dimers in a volume of L were performed. The mean fraction of free repressor monomers as a function of time was fit to a single exponential to obtain the observed rate constant for binding at the inducer concentration. x and o are data from Dunaway *et al.* [@pcbi.1002010-Dunaway1]. (B) Equilibrium binding of IPTG to (blue) repressor and (red) repressor-operator complexes. In a stochastic simulation at each inducer concentration, 20 free (or operator-complexed) repressor dimers in L were first equilibrated with inducer to reach the steady state. Following, 5 minutes of data were collected from which the equilibrium fraction of inducer bound repressor monomers was calculated. x and o are data from O\'Gorman *et al.* [@pcbi.1002010-OGorman1]. :::  ::: Well-stirred and spatially resolved simulations {#s2b} ----------------------------------------------- Since the stochastic switch model is more complex than can be solved using analytic methods, we used computational Monte Carlo methods to sample the master equation and estimate the probability distributions. Two stochastic approaches were used to simulate the *lac* kinetic model: a well-stirred method using the CME and a spatially resolved method based on the reaction-diffusion master equation (RDME). The RDME model of the *lac* circuit can be thought of as a superset of the CME model in that all of the kinetic rates used for modeling reactions in the CME based model are also used in the RDME model, but with additional parameters regarding the spatial localization of particles and their diffusion in three-dimensional space. ### Well-stirred model {#s2b1} The well-stirred model was sampled using a version of Gillespie\'s SSA algorithm [@pcbi.1002010-Gillespie1] implemented in CUDA and running on the GPU. The *lac* model includes independent volumes for the extra- and intracellular space. These were tracked separately during the simulations and, to balance the flux of inducer across the membrane at equilibrium, the internal and external volumes were taken to be equal. Inducer in the extracellular space was maintained at a constant concentration for the propensity calculations by holding the number of molecules fixed. Unless otherwise noted, we ran 10,000 replicates of each simulation to obtain sufficient sampling of the probability distributions. ### Spatial model {#s2b2} The RDME model was sampled using our lattice microbe method. The multi-particle, *in vivo* diffusion operator for this method, based on that of Karapiperis and Blankleider [@pcbi.1002010-Karapiperis1], has been presented previously [@pcbi.1002010-Roberts1] and the reaction operator is provided in Supporting [Text S1](#pcbi.1002010.s001){ref-type="supplementary-material"}. Complete implementation details will be presented in a forthcoming publication. The method discretizes space onto a three-dimensional lattice of uniform spacing and time into uniform time steps. Chemical species randomly diffuse and react on the lattice according to rules defined using spatially dependent stochastic rate constants. A virtual microbe is constructed on a lattice by placing particles and obstacles on the lattice and specifying the reaction and diffusion properties of the lattice sites to mimic the spatial organization of the cell. The lattice microbe method uses the GPU as a computational coprocessor and the whole-cell simulations performed for this study were run for one hour of simulation time (slightly longer than a cell cycle of 55 min) using a time step of 50 and a lattice spacing of 16 nm. For each external condition, 100 independent cell-scale simulations were run. Simulations were carried out on the NCSA Lincoln Intel 64 Tesla Cluster containing two NVIDIA Tesla GPU accelerators per node. Approximately 200 GPU-hours were required per hour of simulation time. Simulations of a smaller volume using 5 ns time steps and 2 nm lattice spacing were also performed to examine the variation in repressor rebinding as a function of packing density at the operator site. Spatial models were developed for two *E. coli* phenotypes: fast- and slow-growth. For the fast-growth cells, the cellular volume was constrained to a typical *E. coli* cell shape: a cylinder with spherical end caps 2 long by 0.8 in diameter. The volume was surrounded by an impermeable cytoplasmic membrane separating the extracellular environment from the cytoplasm. We excluded the outer membrane from the model as its permeability is not thought to be a limiting factor for inducer transport. The intracellular space was randomly filled with stationary, *in vivo* obstacles to 50% volume fraction approximating the *E. coli* intracellular environment (see [Table 2](#pcbi-1002010-t002){ref-type="table"}). The model was then coarse-grained onto a 16 nm resolution lattice, as described in Supporting [Text S1](#pcbi.1002010.s001){ref-type="supplementary-material"}. ::: {#pcbi-1002010-t002 .table-wrap} 10.1371/journal.pcbi.1002010.t002 Table 2 ::: {.caption} ###### Obstacle abundance in *in vivo* spatial models. ::: {#pcbi-1002010-t002-2} Fast Growth*^a^* ^,*b*^ Slow Growth*^b^* ----------------- ------ ------ ------------------------- ------------------ -------- ------ Ribosome 10.4 2700 35005 17.8 3021 5.7 Generic Protein 5.2 346 290908 18.6 96992 25.5 \" 4.3 186 18610 0.7 6205 1.2 \" 4.1 162 9907 0.3 3303 0.4 \" 4.0 156 59862 1.7 19959 2.4 \" 3.8 133 50261 1.2 16758 1.8 \" 3.5 107 47365 0.9 15792 1.3 \" 3.4 91 140212 2.5 46748 3.6 \" 3.0 67 162894 2.0 54311 2.8 \" 2.7 46 226358 2.0 75470 2.7 \" 2.3 29 321118 1.8 107064 2.3 \" 1.7 11 163939 0.4 54659 0.6 DNA*^c^* -- 89 0 0 31000 8.8 a Based on data from Ridgway *et al.* [@pcbi.1002010-Ridgway1]. b Total occupied volume (excl. DNA) of 50%. c Per cylindrical persistence length 2 nm in diameter and 50 nm long. ::: For the slow-growth phenotype, we based the spatial model on cryoelectron tomography (CET) of an *E. coli* cell undergoing slow growth. The full cell was approximately 3 long by 0.4 wide and tomograms encompassed approximately one-third of the cell length. Ribosomes were matched and located in the tomograms as described previously [@pcbi.1002010-Ortiz2] as was a portion of the cytoplasmic membrane. The position of the missing membrane was extrapolated to form a contiguous surface and ribosomes were placed at their measured positions in lieu of random placement. The completed one-third cell model was then mirrored to produce the opposite cell pole. The middle third was randomly generated using ribosome densities from the adjacent CET data. Ribosomes were not observed in the central volume of the cell, which we inferred to be the condensed nucleoid. A random walk algorithm was used to place a full-length *E. coli* chromosome in the nucleoid region. Starting from the center of the nucleoid region, cylinders representing DNA persistence lengths 50 nm in length and 2 nm in diameter were randomly added end-to-end such that the angle between successive cylinders was constrained to . If the random walk left the nucleoid region, the path was unwound a number of steps and a new random path started. This process was repeated until all 31,000 persistence lengths had been added. Since later coarse graining of the model onto a 16 nm resolution lattice spread out the nucleoid density, we did not constrain the chromosome to be circular in this model. Following nucleoid addition, the non-ribosomal *in vivo* obstacles were proportionally placed in the model at random locations (including within free space in the nucleoid region) to reach an occupied volume fraction of 50% ([Table 2](#pcbi-1002010-t002){ref-type="table"}). The final slow growth model was then coarse-grained to a 16 nm lattice for simulation. In *E. coli*, translation of an mRNA containing the sequence for an integral membrane protein is thought to be coupled with translocation of the resultant protein across the cytoplasmic membrane by the Sec translocase [@pcbi.1002010-Driessen1], *i.e.*, cotranslational translocation. Specifically, LacY has been observed to require the bacterial signal recognition particle (SRP) pathway for functional membrane integration [@pcbi.1002010-Macfarlane1]--[@pcbi.1002010-Facey1]. However, it is not presently clear whether or not transcription and translation of membrane proteins are also coupled such that the gene being transcribed is also physically located near the site of translocation. For this reason, we modeled two variants of operator placement for comparison: in the fast-growth phenotype the operator site was located in the center of the cell and in the slow-growth phenotype the operator site was placed on the nucleoid near the membrane (∼32 nm away) close to a cell pole. These two different configuration allowed us to compare the mRNA dispersions for close versus far gene--translocation distances. Messenger molecules were created at the location of the operator following transcription and then allowed to diffuse in the cytoplasm with a diffusion constant of 0.1 [@pcbi.1002010-Fusco1], [@pcbi.1002010-Ishihama1]; in the slow growth model molecules were precluded from entering the nucleoid. In the spatial model, then, was required to diffuse to the membrane before translation could occur; Equation 6 was limited to membrane sites. Since ribosomes likely attach to an mRNA\'s RBS while transcription is still ongoing [@pcbi.1002010-Gowrishankar1], the model assumed that molecules were protected from degradation by RNase E until after reached the membrane; Equation 5 was also limited to membrane sites. Translation of produced LacY proteins at the same location as the mRNA in the membrane. LacY molecules were constrained to diffuse in the membrane with a diffusion coefficient of 0.1 . molecules were randomly placed in the cell and diffused at 1 within the intracellular volume. Small inducer molecules diffuse at ∼1000 in extracellular space and ∼100 in intracellular space. However, in order to reach simulation times on the order of the cell cycle the maximum diffusion coefficient in the model, which depends on the lattice spacing and time step, was 1.28 . Therefore, the diffusion coefficient of the inducer molecules was set to 1.28 . Since inducer molecules are present in large numbers and they diffuse faster than the repressor this approximation was not expected to have a noticeable effect. The lattice was connected to a infinite reservoir of molecules through the use of constant concentration boundary conditions to maintain the extracellular space at a constant inducer concentration. Maximum likelihood fitting of gene expression models {#s2c} ---------------------------------------------------- We analyzed the capability of the burst and two-state analytic models of gene expression to recover parameters from our stochastic simulations of an inducible switch by fitting molecular distributions. We used a maximum likelihood method to estimate the model parameters. Briefly, the likelihood of the model parameters having produced a set of observations is given bywhere is the conditional probability of observation occurring given the parameters . The parameters that maximize this likelihood function are those that describe the best fit of the model to the data, assuming a uniform prior distribution for the parameter probabilities. To find the best parameters for a model of gene expression, was calculated using the model\'s steady-state probability density function with the values being the protein counts from the 10,000 simulations. The parameter values that minimized the negative log of the likelihood function were then found using downhill simplex minimization as implement in the Matlab fminsearch function. We estimated the confidence intervals for different sample sizes by taking 1000 random sets of either 50 or 200 cells from the full set of 10,000 and performed maximum likelihood estimation on each of these data sets. The confidence range for each parameter was then defined by the middle 95% of the values obtained during these random resamplings. The burst model was first expressed in terms of parameters and by Friedman *et al.* [@pcbi.1002010-Friedman1] as the Gamma distribution. However, since our stochastic simulations produced discrete protein counts, we used the discrete formulation for the steady-state probability density derived by Shahrezaei and Swain [@pcbi.1002010-Shahrezaei2] in terms of a negative binomial distribution with parameter being the burst frequency (bursts per mean protein lifetime) and being the burst size (proteins produced per burst). The two-state model was fit using the steady-state probability density function derived by Shahrezaei and Swain [@pcbi.1002010-Shahrezaei2]: In this expression the parameters are , , (the activation rate), and (the inactivation rate), the latter two being expressed in units of mean protein lifetime. Additionally, , , , and is Gauss\'s hypergeometric function. Fitting with all four parameters free often resulted in convergence in a local minima, so we adopted a fitting procedure whereby we first constrained the and parameters and fit only and to obtain initial estimates of these two parameters. In the fully induced state the above probability density function reduces to a negative binomial distribution with no dependence on or , only and . Since neither nor depend on inducer concentration, it is a reasonable approximation to use the values for and in the fully induced state as initial estimates for all inducer concentrations. After obtaining an initial fit for and , we then performed another fit with and unconstrained and with and allowed to vary ±5%. This procedure resulted in convergence at a higher likelihood score than when all four parameter were fit simultaneously for all distributions except one. Results {#s3} ======= Here we present the result of our study into the noise effects in the inducible *lac* genetic switch. The first two sections describe the fitting of model free parameters to data from single-molecule fluorescence studies on *E. coli* populations. The next two sections analyze noise in the well-stirred circuit due to its regulatory control elements. The final two sections report on changes to the behavior of the circuit from *in vivo* effects, using a model of a spatially heterogeneous, crowded cell and then an experimentally determined cell structure under an alternate growth phenotype. Linear relationship between transcriptional burst size and inducer concentration {#s3a} -------------------------------------------------------------------------------- In a recent *in vivo* single-molecule fluorescence study, Choi *et al.* measured the distributions of a fluorescent reporter protein under control of the *lac* operator in individual *E. coli* cells at various inducer (TMG) concentrations [@pcbi.1002010-Choi1]. They performed the measurements in the absence of LacY\'s positive feedback by replacing its gene with that of the membrane protein Tsr in the *lac* operon. This enabled an accurate determination of the protein distribution produced by the circuit at a given inducer concentration without any confounding non-linear effects due to enhancement of the internal inducer concentration by LacY. In the absence of DNA looping, they were able to fit their observed distributions to a gamma distribution , where was interpreted as the frequency of transcriptional bursts relative to the protein lifetime and as the mean number of proteins produced per burst. They observed a relatively constant value for the burst frequency of 3--4 and a linearly increasing relationship between burst size and inducer concentration at low to intermediates concentrations. To understand the origin of the linear relationship between burst size and inducer concentration and to reproduce this behavior in our model, we derived an expression for the burst size as a function of kinetic parameters in our model. As long as bursts are infrequent relative to protein degradation, *i.e.* once a free operator is bound with a repressor it remains bound for a significant fraction of the cell cycle, transcriptional bursting from the *lac* operon can be modeled as a Markov process with competition between RNA polymerase (RNAP) and the various LacI species for binding to the free operator (see [Figure 4](#pcbi-1002010-g004){ref-type="fig"}). ::: {#pcbi-1002010-g004 .fig} 10.1371/journal.pcbi.1002010.g004 Figure 4 ::: {.caption} ###### Markov diagram for transcriptional bursting in the *lac* circuit. Under low-to-moderate inducer concentrations, a burst begins when the operator enters the state and ends when it transitions to a repressor bound state. . :::  ::: Transcription initiation by RNAP was modeled as a pseudo first order process (Equation 4), with a rate constant of . The two repressor states with potentially significant binding affinity were and , shown in Equations 1 & 2. Free repressor binds with free operator with a rate constant of resulting in a pseudo first order rate of . Given the current debate surrounding the binding affinity of the state to the operator, we set the rate constant to be proportional to the free repressor binding constant and analyzed the effect of varying the proportionality constant on the pseudo first order rate . This model of transcriptional bursting assumes that the binding of to the free operator is negligible at low inducer concentrations by assuming and ignoring Equation 3. In practice, this condition was satisfied when . We used the upper limit in our model, which is within the range experimentally reported [@pcbi.1002010-Barkley1]. Following the unbinding of a repressor from the repressor--operator complex, the probability of transcription initiation (and subsequent mRNA creation) occurring at the free operator as opposed to a repressor re-binding is The probability of a given number of consecutive transcription initiation events (the size of the mRNA burst) then follows a geometric distribution with of which the mean is . However, repressor unbinding events that produce no mRNA are not observable as a burst, therefore the mean number of mRNA produced in a transcription bursts (B) is Combining Equations 13 and 14 gives the expression for the mean transcription burst size in terms of the rate constants for transcription initiation and repressor binding Given the inducer mass balances (see Supporting [Text S1](#pcbi.1002010.s001){ref-type="supplementary-material"}) and the expression for the total number of repressor dimers , one can derive the equilibrium concentrations of the two repressor specieswhere is the inducer concentration at which half of the repressor monomers are bound to an inducer molecule. Substituting and into Equation 17 gives the expression for the transcription burst size as a function of inducer concentration From this last equation it is clear that the transcription burst size will be linear over the entire range of inducer concentrations only when . [Figure 5](#pcbi-1002010-g005){ref-type="fig"} shows the effect of varying , of particular interest are the very low values. When , the transcription burst size does not linearly increase over the range of inducer concentrations for which this behavior has been reported (0--200 ). In the model here formulated, a linear relationship between size and inducer concentration exists only when the binding affinity of for the free operator is comparable to that of . For our simulations, we therefore chose , such that , as this value assumed no effect on the unbound repressor monomer due to a single bound inducer and gave a strictly linear relationship for all inducer concentrations. ::: {#pcbi-1002010-g005 .fig} 10.1371/journal.pcbi.1002010.g005 Figure 5 ::: {.caption} ###### Parameter space of repressor binding parameter . \(A) Mean burst size as a function of inducer concentration for various values of , where . Parameters used were  =  M,  =  M,  =  , and  =  . (B) The rate of change in the burst size with respective to the inducer concentration. :::  ::: Fitting transcription and inducer/repressor rate constants to single-cell distributions {#s3b} --------------------------------------------------------------------------------------- To obtain values for the model parameters , , and , we used the distributions for LacY reported by Choi *et al.* [@pcbi.1002010-Choi1], specifically the inferred burst frequency (bursts per cell cycle) and size parameters ( and ) from their gamma distribution fits. From Equation 18, the mean transcription burst size as a function of inducer concentration is . This equation is linear in inducer concentration and by fitting it (multiplied by the mean number of proteins produced per mRNA) to the experimental protein burst sizes, as shown in [Figure 6](#pcbi-1002010-g006){ref-type="fig"}, one can constrain the kinetic parameters. The y-intercept of the line fixes the ratio of transcription to repression in the uninduced state () and the slope can then be used to obtain  = 17.6 for TMG. ::: {#pcbi-1002010-g006 .fig} 10.1371/journal.pcbi.1002010.g006 Figure 6 ::: {.caption} ###### Linear fit of burst size to inducer concentration. x are data from Choi *et al.* [@pcbi.1002010-Choi1]. :::  ::: The linear fit, however, only fixes the ratio between and . To recover unique values for these two rate constants, we next considered the mean duration of each transcription burst. The interpretation of the shape parameter of the gamma distribution as the burst frequency is only meaningful if the burst duration is short compared to the protein lifetime. In that case, individual exponentially sized bursts can be considered exponentially distributed in time and therefore act independently to give rise to a gamma distribution of protein abundance. In setting rate constants for the model, then, we wanted to ensure that the burst duration was appropriately short. The burst duration is simply the mean time for a repressor to bind to a free operator. Given a constant , a linear relationship between burst size and inducer concentration also implies a linear relationship between and inducer concentration as can be seen fromwhere in the last step. For TMG, the linear relationship between burst size and inducer concentration extended to at least ∼200 , which is ∼11 times the value for TMG of 17.6 . From [Figure 7](#pcbi-1002010-g007){ref-type="fig"} it can be seen that the interpretation of as the burst frequency begins to break down once is \>5% of the protein lifetime. Using 5% of the protein lifetime as for 200 , we can compute the value for that gives the appropriate : , using a cell doubling time of 55 minutes. With this value for the repressor binding rate, a single repressor molecule in an *E. coli* cell would take ∼200 s to find a free operator. This is somewhat faster than the 354 s reported by Elf *et al.* [@pcbi.1002010-Elf1]. Using the above value for and the ratio of to from the linear fit of the experimental data we obtained the value for the transcription rate  =  . This rate for transcription initiation resulted in a steady state concentration of ∼2500 LacY molecules per cell in the fully induced state, within a factor of two of the ∼1000--1200 reported in the literature [@pcbi.1002010-Choi1], [@pcbi.1002010-Choi2]. The value also falls within the range of 1000--3000 seen for other highly expressed proteins in *E. coli* [@pcbi.1002010-Taniguchi1]. Accurate measurements of the burst duration in the *lac* system, particularly in the fully induced state, would increase the accuracy of our model. ::: {#pcbi-1002010-g007 .fig} 10.1371/journal.pcbi.1002010.g007 Figure 7 ::: {.caption} ###### Burst analysis of stochastic simulations of a simple two-state process. The two-state process was described by: . Rate constants were chosen such that on average bursts of Z with a constant burst size were produced during Z\'s mean lifetime with the mean duration of each burst lasting for the indicated fraction of the lifetime. At each point, 250 stochastic simulations were run until the probability density was stationary and then the distributions of Z were fit to gamma distributions to obtain the and parameters. The ratios of (A) / and (B) / as a function of the burst duration show the range of burst durations for which a gamma distribution fit can reliably recover the original parameters. In this example and . :::  ::: In order to reproduce a burst frequency of over the mean LacY lifetime in the model, the repressor should dissociate from the operator with a frequency , assuming that each dissociation event produces a burst and that the cell cycle. The burst frequencies inferred by Choi *et al.* for TMG levels ≤100 are relatively constant with a mean of ∼3 bursts. This corresponds to  =  . Since the dissociation of a repressor dimer is not thought to be significantly affected by the binding of a single inducer molecule,  = . The affinity of a repressor dimer with two bound inducer molecules, however, is thought to be much lower, *i.e.*, the binding of a second inducer molecule essentially knocks the repressor off of the operator. In the absence of this effect, the response to an increase in inducer concentrations would take a significant fraction of the cell cycle. Elf *et al.* reported a response time of \<60 seconds for addition of IPTG to concentrations from 50 -- 1 mM [@pcbi.1002010-Elf1]. Therefore, we fit such that the response of the model to increase in IPTG agreed with the published data. The best fit value was obtained for (shown in [Figure 8A](#pcbi-1002010-g008){ref-type="fig"}). ::: {#pcbi-1002010-g008 .fig} 10.1371/journal.pcbi.1002010.g008 Figure 8 ::: {.caption} ###### Parameter fitting for inducer--repressor--operator interactions. \(A) Fraction of operator regions bound by a repressor as a function of time following an increase of IPTG to the indicated concentration. In these simulations, . (B) Number of bursts over the mean protein lifetime as a function of inducer concentration for a variety of values of the parameter. x are data from Choi *et al.* [@pcbi.1002010-Choi1]. :::  ::: The final kinetic rates to be defined were those regarding the binding of TMG to the repressor--operator complex (Equations 10 & 11). As discussed in [Methods](#s2){ref-type="sec"}, we used the same dissociation rates as for IPTG, leaving only the association rates and , both of which can be derived from the value, which is the inducer concentration at which half of the repressor--operator complexes have a bound inducer. [Figure 8B](#pcbi-1002010-g008){ref-type="fig"} shows the effect of varying on the burst frequency. As approaches , the burst frequency begins to diverge from its expected value. This is due to the increasing occupancy of the O state, which can decay much more quickly into a free operator than the other repressed states; with operator free more often, there are more bursts over the lifetime of a protein. A value of 3 mM for gave the best agreement with the experimental burst frequencies for TMG. Population distributions without positive feedback {#s3c} -------------------------------------------------- Using the derived rates, we performed well-stirred stochastic simulations of the *lac* model in the absence of LacY positive feedback (NPF model), obtaining the stationary LacY distributions as a function of internal inducer concentration shown in [Figure 9](#pcbi-1002010-g009){ref-type="fig"}. Compared to the intrinsic noise of the two-state model, the NPF model contains additional noise contributions from the non-constant rates for transitioning between active and inactive transcriptional states. The distributions showed the widest cell-to-cell variability due to the intrinsic noise of the system at intermediate inducer concentrations of 50--400 . At high inducer concentrations the population migrated toward a less variable distribution, as expected. Up to 100 , the population distributions agreed well with those reported by Choi *et al.* but at 200 the agreement began to break down. This discrepancy at concentrations \>100 was caused by two primary factors: the burst duration and the action of inducer knocking repressor off of the operator. Increasing the repressor binding rate would improve the fit by decreasing the duration of each burst, but would cause a large increase in the total number of LacY molecules in the fully induced state, which is not supported experimentally. Alternatively, one could increase the value, causing less inducer instigated dissociation of the repressor--operator complex, but this would decrease the responsiveness of the circuit to addition of inducer, which is also not supported experimentally. Clearly, in order for the model to have greater predictive power, additional features would be necessary. For example, adding a delay between production of mRNA to account for the steps of RNAP open complex formation or more detailed modeling of translation. But lacking the *in vivo* experimental results to validate any additional complexity, we chose to ignore these effects and analyzed the model as described. ::: {#pcbi-1002010-g009 .fig} 10.1371/journal.pcbi.1002010.g009 Figure 9 ::: {.caption} ###### Steady state LacY distributions from the well-stirred NPF model. Distributions at inducer concentrations of (A) 0, (B) 100, and (C) 200 TMG. Shown are (gray bars) histograms from 10,000 Gillespie trajectories and (red dash) gamma distributions from Choi *et al.* [@pcbi.1002010-Choi1]. (D) Mean LacY as a function of inducer concentration along with 95% ranges. (E) The noise in the LacY distributions as quantified by the Fano factor (variance over the mean). (F) The fraction of time spent in the transcriptionally active state. :::  ::: The gene regulation function (GRF) of an genetic system describes the relation between the activity of a gene and its regulatory control elements [@pcbi.1002010-Setty1]--[@pcbi.1002010-Mayo1]. In the steady state, protein production is balanced by protein degradation/dilution. The mean protein count as a function of the control elements provides a method to analyze a GRF. The mean number of LacY per cell as function of the TMG concentration ([Figure 9D](#pcbi-1002010-g009){ref-type="fig"}) and the fraction of time spent in the transcriptionally active state ([Figure 9F](#pcbi-1002010-g009){ref-type="fig"}) show the regulatory behavior of the NPF model. We saw a typical sigmoidal regulatory response that was well fit by a Hill equation with an inflection at 312 and a Hill coefficient of 2.11. In a stochastic system, though, the mean rate of gene expression is just one piece of information. As important for a stochastic GRF is how the distribution changes with inducer concentration. The Fano factor (variance/mean) provides a measure of the variation of the distribution. For reference, the Fano factor of a Poisson process is 1. For the NPF model ([Figure 9E](#pcbi-1002010-g009){ref-type="fig"}) the Fano factor monotonically increases until 100--200 where it peaks at a value of ∼60 and then begins to decrease ending at a lower value of relative noise than at zero inducer. Noise due to positive feedback {#s3d} ------------------------------ Next we investigated noise in the inducible genetic switch when the positive feedback regulatory link was active (PFB model). The *lacY* gene located in the *lac* operon codes for the integral membrane protein LacY, which actively imports inducer molecules (lactose/ co-transport) establishing a positive feedback loop as shown in [Figure 1C](#pcbi-1002010-g001){ref-type="fig"}. The presence of active LacY in the membrane creates a concentration gradient enriching the intracellular environment with inducer molecules relative to extracellular space. For a fixed concentration, the underlying GRF for the *lac* operon therefore operates not only at an increased inducer concentration but, since the number of LacY is different for each cell, across a distribution of internal inducer concentrations. We calculated the population distributions for the PFB model using well-stirred stochastic simulations at various concentrations. Starting from a stationary population distribution in the absence of inducer, each population of 10,000 cells was subject to an instantaneous increase in and simulated for twenty-four hours. Above an concentration of ∼10 , cells in the population began to switch to an induced state in which LacY expression was near its maximum value (see [Figure 10A and B](#pcbi-1002010-g010){ref-type="fig"}). Above ∼25 the transition to full expression was relatively concerted throughout the population. In the range of 10--25 , though, there were two transiently stable subpopulations, one uninduced and the other induced -- the overall population was bimodal for a time. ::: {#pcbi-1002010-g010 .fig} 10.1371/journal.pcbi.1002010.g010 Figure 10 ::: {.caption} ###### Response of an uninduced PFB population to the addition of external inducer. \(A) Probability density (arbitrary units, darker = higher) of the number of LacY in a cell over the course of 24 hours. Shown are representative responses for populations in the uninduced range (0--10 ; left), the bimodal range (10--25 ; center), and the concerted induction range (\>25 ; right). Lines show the mean value of the (green) uninduced and (red) induced subpopulations. (B) Fraction of the cells in each of the subpopulations. (C) The (solid) mean and (dotted) variance of LacY in the uninduced subpopulation. :::  ::: To quantify the switching behavior of the population, we classified cells at regular time intervals as uninduced with \<400--600 LacY (best fit for each ) or induced with \>1750 LacY. Each subpopulation was then analyzed separately. The mean and variance of the distributions ([Figure 10C](#pcbi-1002010-g010){ref-type="fig"}) show that, after an initial response phase, the distribution of the uninduced subpopulation was stable over time. This was true even as the total number of cells in the uninduced population was decreasing as cells within it were switching to the induced state. At intermediate inducer concentrations, the uninduced cell population appeared to reach a stationary distribution from which cells independently and stochastically transitioned to the induced state. In contrast, at higher inducer concentrations the population migrated as a whole in a more downhill-like manner. Noise in a GRF can be expressed in terms of its effect on the phenotypic variance in a population under identical environmental conditions. To compare noise between the NPF and PFB models, we first mapped concentrations to mean concentrations in the uninduced and induced subpopulations (in the NPF model  = ). We then compared both the mean of the LacY distributions and the Fano factor for the two models. The mean values for the LacY distributions ([Figure 11](#pcbi-1002010-g011){ref-type="fig"}) were similar but the noise in the uninduced subpopulation was significantly higher in the model with positive feedback. Since the underlying GRF is equivalent between the two models, it is the action of the GRF on the distribution of concentrations that gives rise to the increase in intrinsic noise in the PFB model. ::: {#pcbi-1002010-g011 .fig} 10.1371/journal.pcbi.1002010.g011 Figure 11 ::: {.caption} ###### Effect of positive feedback on GRF noise. \(A) Mapping of the mean internal inducer concentration for a given external concentration for the (green) uninduced and (red) induced subpopulations. (black dotted) The values for the lac circuit without positive feedback are shown for reference. (B) The mean number of LacY in the subpopulations as a function of internal inducer concentration. (C) The noise in the LacY distribution. :::  ::: Differences in circuit behavior due to *in vivo* crowding {#s3e} --------------------------------------------------------- Having established the well-stirred PFB stationary distribution, we next evaluated the effect of *in vivo* molecular crowding on the distributions, the PFB+IV model. One obvious reaction subject to spatial effects is the rebinding of the repressor to the operator following an unbinding event. Immediately after unbinding, a repressor is necessarily localized near the operator, *i.e.* it has a memory of its location. As was shown by van Zon *et al.* [@pcbi.1002010-vanZon1], this memory effect increases the probability of repressor rebinding at very short times compared to a well-stirred approximation. Previous studies only considered the effect of normal diffusion following unbinding but there is an additional effect caused by anomalous diffusion due to *in vivo* crowding. To investigate repressor rebinding in an *in vivo* environment, we performed reaction-diffusion simulations of a volume centered on an operator immediately following unbinding of a repressor. We varied the packing density of the approximated *in vivo* environment to study its effect on rebinding. [Figure 12A and B](#pcbi-1002010-g012){ref-type="fig"} shows that there is an anomalous effect at short time scales (\<1 ms). Repressor diffusion at very short time scales is normal at the *in vitro* rate, but between 1--100 there is a period of anomalous behavior, and at very long time scales repressor diffusion returns to normal diffusion behavior with a lower diffusion coefficient D. Brownian dynamics simulations of proteins in a virtual *in vivo* environment [@pcbi.1002010-McGuffee1] show a similar anomalous behavior when including only steric constraints with a minimum in the time exponent of ∼0.8 for proteins slightly larger than the 75 kDa repressor dimer. When electrostatic effects are included in the Brownian dynamics simulations, however, the apparent diffusion coefficient as well as the anomalous exponent change greatly, so our results should only be considered an upper bound on the *in vivo* effects. Including further electrostatically driven interactions such as non-specific binding, will increase the anomalous behavior of the repressor. ::: {#pcbi-1002010-g012 .fig} 10.1371/journal.pcbi.1002010.g012 Figure 12 ::: {.caption} ###### The effect of *in vivo* crowding on repressor rebinding. Each line represents the mean of 5000 trajectories. (A) The observed diffusion coefficient, , as a function of time scale for a repressor diffusing in a volume with the indicated fraction occupied by *in vivo* obstacles. (B) --exponent arising from fitting to a model of anomalous diffusion, . (C) The probability for a repressor to rebind with the operator before diffusing into the bulk (64 nm from operator) following unbinding, as a function of the *in vivo* packing. (D) The distribution of escape times for repressors that diffuse to bulk rather than rebind, at three packing values. :::  ::: The anomalous behavior of the repressor causes it to spend more time near the operator following unbinding than would be expected for purely Brownian diffusion, leading to more encounters with the operator and a potentially greater probability of rebinding. To measure the change in rebinding probability, we counted the number of repressors that rebound to the operator following unbinding versus the number that escaped into bulk solution, defined here as leaving the simulation volume. As can be seen in [Figure 12C](#pcbi-1002010-g012){ref-type="fig"}, as the density of *in vivo* crowding increases, the probability of rebinding goes up. Compared to an *in vitro* unpacked environment at 15% probability of rebinding, at 50% packing the probability of rebinding is ∼24%. The distribution of escape times also broadens ([Figure 12D](#pcbi-1002010-g012){ref-type="fig"}) with particles in general taking longer to diffuse away. The anomalous memory effect resulted in the duration of some bursts being significantly shorter than expected. To study the effect of burst duration differences on the stationary LacY distributions in a population, we used our lattice microbe method to generate PFB+IV trajectories of spatially resolved rapid-growth *E. coli* cells (see [Methods](#s2){ref-type="sec"}). Beginning with the stationary distribution from the well-stirred PFB population, 100 cells were simulated at five internal inducer concentrations for one hour, slightly longer than the duration of a cell cycle (55 minutes), see [Video S1](#pcbi.1002010.s002){ref-type="supplementary-material"}. Over the course of the simulations, distributions in the *in vivo* models gradually migrated to lower mean values and lower noise, as can be seen in [Figure 13](#pcbi-1002010-g013){ref-type="fig"}. Two factors caused this migration: First, the shorter burst durations due to the anomalous diffusion effect described above resulted in fewer proteins being produced per burst and more time spent in the inactive state led to more frequent bursts and less noise. Second, the effective increase in repressor due to the decreased reaction volume. In contrast to spatial effects in an *in vitro* environment [@pcbi.1002010-vanZon1], it appears that *in vivo* crowding lowers both the mean value and the noise in distributions of observables. Since bacterial cells such as *E. coli* are known to have packing density changes during different portions of the cell cycle and/or growth conditions, this presents the possibility of measuring these *in vivo* effects on living cells if the observable distributions can be accurately quantified as a function of the cell cycle or growth conditions. ::: {#pcbi-1002010-g013 .fig} 10.1371/journal.pcbi.1002010.g013 Figure 13 ::: {.caption} ###### LacY PFB+IV *in vivo* distributions. \(A) The distribution of LacY in (orange bars) 100 modeled *E. coli* cells at 13 TMG concentration compared with (green dotted) the PFB well-stirred distribution. (B) Mean number of LacY proteins in the (circles) PFB+IV and (green dotted) PFB models. (C) The noise in the distributions. :::  ::: Whole-cell modeling using experimentally determined cell architecture {#s3f} --------------------------------------------------------------------- As a first attempt at addressing how changes in the cellular environment due to growth conditions affect gene expression noise, we used CET of *E. coli* cells under slow growth to build a whole-cell model of an individual bacteria ([Figure 14A](#pcbi-1002010-g014){ref-type="fig"}). Under conditions of slow growth in minimal media *E. coli* B/r K grows as elongated cylinders with diameter ∼400 nm [@pcbi.1002010-Woldringh1], which are amenable for CET [@pcbi.1002010-Ortiz2]. The tomograms were used to identify the membrane-enclosed volume of an individual cell along with the three-dimensional position of ribosomes within it. The *E. coli* B/r K cell under slow growth had only of the volume of typical fast growing cells. A central region of the cell was devoid of ribosomes and inferred to be the location of the condensed nucleoid. ::: {#pcbi-1002010-g014 .fig} 10.1371/journal.pcbi.1002010.g014 Figure 14 ::: {.caption} ###### Analysis of cryoelectron tomography based cell model. \(A) Slow growth *E. coli* cell model based in part on data from a tomographic reconstruction. Shown are (orange) ribosomes, (light gray) membrane, (dark grey) condensed nucleoid, and (red) *lac* operator. (B+C) Distribution of repressor--operator complex lifetimes for the fast and slow growth models, respectively. Curves show fits to an exponential distribution with the given mean. (D) Position of mRNA--membrane contact after diffusion of mRNA produced at the *lac* operon in (blue x) fast growth and (red o) slow growth models. Dotted lines show the length of the respective cells. :::  ::: We studied the operation of the *lac* circuit in the slow-growth phenotype (PFB+IV+CET) using 100 random replica cells. Each replica used the same experimentally measured cellular geometry and ribosome positions, but a random distribution of other molecules including a condensed chromosome (see [Methods](#s2){ref-type="sec"} for details). Cells were simulated using the lattice microbe method in 15 external TMG, starting with LacY and mRNA counts sampled from the uninduced stationary distribution of the well-stirred PFB model, see [Video S2](#pcbi.1002010.s003){ref-type="supplementary-material"}. Simulations were run for either one hour or until the cell had induced, whichever came first. There were clear differences between the slow- and fast-growth *in vivo* models. Of the 100 slow-growth cells, 11 induced within one hour whereas only a single fast-growth cell induced in the same time period. Also, the mean number of LacY molecules in the uninduced slow-growth population increased ∼15% over the course of one hour, compared to the fast-growth population which decreased ∼15%. Analysis of the simulation trajectories revealed that the primary cause of the differences in LacY distributions between the slow- and fast-growth models was an increased mean inducer concentration in the smaller cells, 100 versus 42 . For a given number of LacY proteins, the cells with the smaller volume had an increased internal inducer concentration. The increased levels of inducer caused a slight lengthening of the mean duration of free operator events, 68 seconds versus 64 seconds, and a corresponding larger burst size. A bigger change was observed in the mean lifetime of the repressor--operator complex, which decreased to 430 seconds from 730 seconds ([Figure 14B,C](#pcbi-1002010-g014){ref-type="fig"}). The decrease effected an increase in the mean number of transcription bursts per hour, to 4.3 from 2.6. The slow-growth model provides a first approximation as to the effect of differences in cellular architecture on stochastic gene expression. The model assumed the same number of repressor molecules for smaller cells, which may not be accurate as repressor is known to regulate its own expression. However, since the largest effect was due to an increased rate of repressor unbinding due to elevated inducer levels, which is independent of repressor concentration, we consider the general result of increased burst frequency and rate of induction in smaller cells to be intriguing. It implies that there might be a difference in the switching properties during the first part of the cell cycle following division when a large burst of LacY would have an increased influence on switching due to the reduced cellular volume. Such an effect could potentially be measured using cell synchronization techniques. Although specific ribosome placement likely also influenced repressor rebinding in the slow-growth model, any differences were overshadowed by the effect of the cell volume change. Nevertheless, in a situation where the placed macromolecules are involved in the reaction kinetics, we anticipate accurate (non-uniform) placement will take on much greater importance. Another large difference between the slow- and fast-growth models arose due to the presence of a condensed nucleoid coupled with the smaller cell diameter. In the fast-growth cells the chromosome was assumed to be diffuse and not an obstacle for mRNA diffusion. In the slow-growth cells, the chromosome was randomly placed in the ribosome-excluded region observed in the tomograms and it represented an obstruction for mRNA diffusion. Additionally, the operator was positioned in the center of the fast-growth cells and at the edge of the nucleoid in the slow-growth cells. As can be seen in [Figure 14D](#pcbi-1002010-g014){ref-type="fig"} there was a dramatic increase in localization of mRNA in the slow-growth cells as a result of this arrangement. A recent report of mRNA localization in bacteria [@pcbi.1002010-Llopis1] suggests that the relative locations of transcription and translation in bacteria may indeed be correlated. If that is generally true, then in systems where the location of protein synthesis affects the reaction kinetics it will be important to know the actual position of the gene in the cell and measurement of the dispersion of the transcripts might be one way to quantify whether the gene is physically located near the site of translation and translocation. Discussion {#s4} ========== Fitting population distributions to models of stochastic gene expression {#s4a} ------------------------------------------------------------------------ Fitting protein population distributions to gene expression models will be a key step in developing simulations of other stochastic cellular systems with predictive power. Parameters obtained from fitting the distributions will drive the computations. Our stochastic simulations of the inducible *lac* switch provide an opportunity to test the process of extracting parameters from a population distribution arising from a complex gene expression system using simplified but analytically tractable models. To do so, we fit the stationary population distributions from our simulations to both the burst and two-state models ([Figure 1A & B](#pcbi-1002010-g001){ref-type="fig"}) and evaluated their capability to recover the stochastic rate constants used in the simulations (*e.g.* , , etc). The analysis was performed for each of the different noise variations described above, corresponding to the NPF, PFB, and PFB+IV simulations. We excluded the PFB+IV+CET simulations from this study as they were not performed over a range of inducer conditions. The best fit parameter values were obtained by maximum likelihood estimation using the stationary probability density function (PDF) for the burst and two-state models, Equations 12 & 14 in [Methods](#s2){ref-type="sec"}. Fits were performed using 10,000 cells for NPF and PFB simulations and 100 cells for PFB+IV simulations. [Figure 15A & B](#pcbi-1002010-g015){ref-type="fig"} show parameter estimates obtained from fitting using the burst model\'s gamma distribution PDF (Equation 12). The and parameters (the B subscript indicates parameters for the burst model) reliably recover the burst frequency and burst size, respectively, in the NPF simulations at low inducer concentrations, but diverge from the simulation values above ∼100 . This is as expected as the model is only valid when the duration of each burst is short enough that sequential bursts can be considered as occurring independently, \<5% of the protein lifetime as shown in [Results](#s3){ref-type="sec"}. In particular the divergence occurs near the switching threshold, making this model most suitable for analyzing the system in the uninduced state with low expression levels. However, the clearness of the biological interpretation for the model parameters as the burst frequency and size make the model extremely valuable over the regime it is valid. ::: {#pcbi-1002010-g015 .fig} 10.1371/journal.pcbi.1002010.g015 Figure 15 ::: {.caption} ###### Maximum-likelihood fitting of two models for gene expression to stochastic simulations of an inducible genetic circuit. (A and B) Parameter fits from the burst model. (C--F) Parameter fits from the two-state model. Shown are fits for (black dotted) NPF simulations, (green dotted) PFB simulations, and (orange circles) PFB+IV simulations. Also shown are (blue solid) actual parameter values calculated from the simulation data. Shaded areas indicate the 95% confidence intervals for ML fits using distributions from 50 and 200 NPF cells. :::  ::: Fitting the NPF simulation data to the stationary PDF of the two-state model ([Figure 1B](#pcbi-1002010-g001){ref-type="fig"}; Equation 14) provides good parameter estimates over a wider range of inducer concentrations. The fits are shown in [Figure 15C--F](#pcbi-1002010-g015){ref-type="fig"} for the parameters (; the TS subscript indicates two-state), (), (the rate constant for operator activation), and (the rate constant for operator inactivation), respectively. As the inducer concentration increases, though, many more cells are required to obtain reliable estimates. Using even 10,000 cells, we were unable to obtain good fits for the highest expression levels. At these inducer levels so little time is spent in the inactive state that the difference in likelihood values for different switching rates is insufficient to find a unique maximum using 10,000 samples. However, as the time spent in the inactive state approaches zero () the probability distribution approaches a negative binomial distribution without dependence on or , so it is possible to estimate the and parameters in the fully induced state by fitting to a negative binomial. The two-state model therefore appears to be a reasonable method for fitting the NPF simulations. Using the fitting parameters (along with known or estimated mRNA and protein degradation rates), one can readily recover the transcription and translation rates as well as the rates of the operator switching between active and inactive states at a given inducer concentration. Even though switching between active and inactive states in the *lac* switch is not a first order process -- it is controlled by 14 reactions -- at a given inducer concentration the steady state switching times are reasonably well-approximated by a single exponential. A further improvement in the two-state model would allow and to depend on the inducer concentration using, *e.g.*, a Hill function. An analytic solution to such a model would allow extraction of parameters from a multivariate fit using data across all inducer concentrations. However, to the best of our knowledge, the analytic form of such a model has not been derived. Using the steady state distributions from the PFB simulations, neither model achieves good fits. For the two-state model, the and parameters are recovered correctly, but the fits for the and parameters are lower than expected. The poor fit for these parameters is due to noise in the switching rates of the cell population caused by differences in internal inducer concentrations. With positive feedback, it will be very difficult to reliably estimate model parameters from population distributions due to non-linear noise. Fitting to experimental data should be done in the absence of positive feedback, such as by using gene knock-outs to eliminate circuit components responsible for positive feedback. However, if an analytic model were developed including positive feedback effects, comparison of systems with and without these effects could provide estimates of positive feedback parameters, *e.g.* inducer transport rates. Fits to the PFB+IV simulations as well show deviations from the expected values; *in vivo* crowding noise changes the parameter fits. For these simulations, an additional source of discrepancies with the models is the non-Poissonian behavior of repressor rebinding -- there is a positional memory in the system for a short time following unbinding. In our simulations the effect from *in vivo* conditions due to excluded volume is modest, but there are other *in vivo* factors still not accounted for in them, especially non-specific binding as recently reported by McGuffee and Elcock [@pcbi.1002010-McGuffee1], which would have an even larger effect on repressor rebinding. Also, repressor rebinding most likely occurs via a series of 1D sliding and 3D hopping steps, the effect of which on rebinding in a crowded environment is not known. Accounting for *in vivo* effects when deriving parameter from experimental population distributions, which would include *in vivo* noise contributions, will be difficult. Possibly an iterative process of refinement may be required, starting with model estimates and proceeding through multiple rounds of spatial simulation. Overall, it appears that fitting population distributions to the two-state model could prove to be an effective way of obtaining rate constants for stochastic simulations of gene regulation. Single-molecule *in vivo* fluorescence imaging provides a way to experimentally measure these distributions. Measurements over a range of regulatory conditions could then be used to build a stochastic gene regulation function, provided the actual probability distributions from single-molecule experiments were available at each condition. However, it is important to acknowledge that our simulations did not include a contribution from global extrinsic noise. Noise in our simulations under conditions of high expression approaches Poissonian, as expected from the intrinsic noise of an uncorrelated random process. A recent study has clearly shown, though, that there is a constant level of global extrinsic noise in gene expression in *E. coli*, maintaining population heterogeneity even at high levels of gene expression [@pcbi.1002010-Taniguchi1]. This implies that a way to correct for the global extrinsic noise will be needed in order to fit experimental population distributions at high expression levels. Probability landscape of an inducible *lac* switch {#s4b} -------------------------------------------------- The probability distribution for a stochastic biochemical system to be in a particular state represents the totality of information about the system. From it various measures of the behavior of the system such as the mean first passage time between two states or their relative population at the steady state can be obtained. For models of stochastic gene expression, two relevant reaction coordinates are the number of protein and mRNA molecules in the system. We used our stochastic simulations to reconstruct the two-dimensional probability landscapes (negative log of the PDF) of the NPF and PFB models at two external TMG concentrations ([Figure 16](#pcbi-1002010-g016){ref-type="fig"}). ::: {#pcbi-1002010-g016 .fig} 10.1371/journal.pcbi.1002010.g016 Figure 16 ::: {.caption} ###### Probability landscape of protein--mRNA abundances in the inducible *lac* switch model. \(A) Steady-state probability landscape (arbitrary units, darker = higher) for the NPF model at 500 TMG. The dotted line shows the trajectory of a representative cell during a ∼3 hour interval starting at the open circle and ending at the closed circle. (B) Probability landscape of the PFB circuit over a period of 24 hours following the addition of external TMG to 16 . The line follows a single cell switching from the uninduced to the induced state over the course of ∼13 hours. :::  ::: The steady-state landscape of the NPF simulations at 500 inducer shows a bistable mRNA distribution that has been reported by others [@pcbi.1002010-Shahrezaei2], [@pcbi.1002010-Raj3]. One minima is located near 0 /1700 LacY and the other near 10 /1800 LacY. Note that the stable 10/1800 point does not imply that 180 LacY were produced per , as the degradation rates of the two molecules differ. At 500 inducer, the net time some cells stochastically spend in the inactive state is greater than the typical lifetime of the mRNA bursts. These cells then drift to a zero mRNA abundance. The higher density is caused by an accumulation of these cells near the zero mRNA level until their next mRNA burst pushes them back into a random cycle around the mean mRNA burst size. Interestingly, though, the protein distributions at the two mRNA minimum are different, with the protein abundance being slightly lower in the lower mRNA minimum. This means that the protein and mRNA probability distributions are not completely independent of each other; the joint probability distribution has cross terms. While not a large difference, it is nevertheless possible that the joint protein--mRNA distribution could be used to obtain better parameter fits for the two-state model with fewer cells, if the mRNA counts were known. The bimodal distributions seen in the LacY distributions from the PFB simulations ([Figure 10](#pcbi-1002010-g010){ref-type="fig"}) are recapitulated in the probability landscape for switching. The two-dimensional landscape allows classification of both the uninduced and induced states in terms of their relative protein and mRNA abundances. Additionally, the landscape reveals the transition path for switching from the uninduced to the induced state. One can imagine two possible scenarios for the transition, either the gradual build-up of protein by a series of small bursts, or alternatively, by the random occurrence of a small number of larger bursts. For the *lac* system with DNA looping, Choi *et al.* [@pcbi.1002010-Choi2] have persuasively argued for the random large bursts as the switching initiator. The switching mechanism of the stochastic system in the absence of looping, however, is not so clear. The probability landscape of our *lac* model suggests that it is actually the occurrence of several large mRNA bursts, on the order of \>10 molecules, in quick succession that is responsible for putting the cell on the path to induction. Cells can spend a significant amount of time in a high LacY but low state without inducing. This behavior is apparent in the cell trajectory plotted in [Figure 16B](#pcbi-1002010-g016){ref-type="fig"}. Switching therefore is a process in which not only a protein threshold must be crossed, but also an mRNA threshold. Conclusions {#s4c} ----------- Our goal with this study was to go beyond previous stochastic simulations of the *lac* circuit by using information from single molecule protein distributions and experimentally determined cellular architecture to constrain the kinetic parameters and estimate the effect of spatial heterogeneity on the response of the switch. The kinetic model of the inducible *lac* genetic switch presented in this study illustrates the utility of incorporating single-molecule, single-cell data when modeling cellular biochemical systems. The model was derived using a kinetic framework reproducing a linear relationship between protein burst size and inducer concentration at low concentrations, as has been reported experimentally. Analysis of the linear relationship in terms of inducer--repressor--operator interactions suggests that the stoichiometry of repressor binding is such that repressor dimers with one bound inducer still have significant affinity for the *lac* operator. Furthermore, single-cell population distributions were used to obtain estimates of the effective rate constants for transcription and repression in the cell. With future increases in performance of the lattice microbe simulation method it should be possible to iteratively refine the kinetic rate constants to account for the effects of cellular architecture, such as we obtained here from CET experiments, and cytoplasmic crowding. Using such *in vivo* adjusted rate constants the *in vivo* models should then more accurately reproduce experimental population distributions, which are after all measured under *in vivo* conditions, than the well-stirred models. The *lac* model without positive feedback provided a baseline for the noise in the regulation of the *lac* operon. Intrinsic noise at low gene expression was significantly higher than Poissonian and peaked when the promoter was active 10--30% of the time. The model with positive feedback produced similar mean values for a given intracellular inducer concentration, but the noise was substantially greater. We attribute this effect to the non-linear gene regulatory function operating on a distribution of intracellular inducer levels. Global extrinsic noise in the transcription/translation machinery is a large contributor to population heterogeneity at high levels of expression, but we excluded such noise from the current study. Fitting of data from stochastic simulations of the *lac* switch with the burst and two-state models of gene expression showed both the potential and limitations of these models to interpret stochastic gene regulation. The burst model described the data well under conditions of low expression, when the gene was active for ≤5% of the mean protein lifetime, but diverged for increasing expression levels. The two-state model better described the data at higher levels of expression, but near full induction the error in the activation and inactivation rates became significant. Additionally, the fits provided estimates of the number of cell measurements necessary to produce reliable parameter estimates. With 50 cells the worst-case relative error was ±90%, but with 200 cells it dropped to ±32%. Fitting to joint mRNA--protein distributions might improve parameter estimation. Fits to data with positive feedback indicated that both models were unable to reliably extract parameters from populations with such feedback. Switching of cells from the uninduced to the induced state was observed in the positive feedback model without DNA looping over a range of low inducer concentrations. During switching, the uninduced population maintained a stable stationary distribution while cells stochastically transitioned to the induced population. The probability landscape showed that both an mRNA and a protein threshold must be crossed for a cell to switch to the induced state. The probability landscape for the DNA looping case is likely different, but additional model states would be required to accurately represent DNA looping. Finally, we have presented what we believe to be the first whole-cell simulations of stochastic gene expression using experimentally obtained cellular architecture. These simulations showed that *in vivo* conditions can impact the stochastic noise in biological systems. Positional memory of transcription factors following unbinding, amplified by anomalous diffusion due to molecular crowding, introduces non-Poissonian statistics. In the case of our *lac* model in fast-growth cells, this effect caused a decrease in the mean value of the LacY distribution by up to 10% and its noise by up to 20%, for a given environmental condition. In a slow-growth cell phenotype we saw a large increase in burst frequency due to the smaller cell size, as determined from cryoelectron tomography. From this difference we infer that changes in cellular size and/or shape during the cell cycle can have an impact on stochastic processes. Since spatial noise can vary from cell-to-cell or even during the cell cycle so we consider it a type of extrinsic noise. The necessary computational resources and experimental data are becoming available such that computational biologists should consider adding spatial degrees of freedom into physical models of cellular biochemical networks. Supporting Information {#s5} ====================== Text S1 ::: {.caption} ###### Supporting methods. Further methods describing the derivation of the rate constant relationship for non-cooperative ligand binding, the lattice microbe reaction operator, and the lattice coarse-graining technique. (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Video S1 ::: {.caption} ###### Simulated colony of *E. coli* cells responding to inducer. Video composite of trajectories from six spatial PFB+IV simulations at 15 inducer. Yellow circles are LacY proteins and red circle are mY mRNA molecules. Two cells begin the process of switching to the induced state. (MOV) ::: ::: {.caption} ###### Click here for additional data file. ::: Video S2 ::: {.caption} ###### Trajectory of a PFB+IV+CET cell responding to inducer. Visualization of a single slow-growth CET modeled cell responding to 15 inducer. Gray spheres are ribosomes and the blue region the nucleoid. Yellow circles are LacY proteins and red circles are mY mRNA molecules. The repressor--operator complex is green and the free operator is white. (MOV) ::: ::: {.caption} ###### Click here for additional data file. ::: The authors would like to acknowledge John Stone for helpful discussions on implementation of the code on the GPU. The authors have declared that no competing interests exist. This research was supported by the Department of Energy Office of Science (BER), the National Science Foundation (MCB-0844670, PHY-0822613), and the Foundation Fourmentin-Guilbert. Computational resources were provided by the NSF through the TeraGrid and NCSA (TG-MCA03S027) and also by the CUDA Center of Excellence at UIUC. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: ER AM JOO WB ZLS. Performed the experiments: ER AM JOO. Analyzed the data: ER AM JOO ZLS. Wrote the paper: ER ZLS.

PubMed Central

2024-06-05T04:04:19.678446

2011-3-10

PMC3053319

Introduction {#s1} ============ The Warburg effect, a phenomenon discovered by Otto Warburg in 1924, reflects a shift to an inefficient metabolism in cancer cells, in which an increase in the inefficient production of adenosine 5′-triphosphate (ATP) via glycolysis leads to the secretion of non-oxidized carbons in the form of lactate, even in the presence of oxygen (termed *aerobic glycolysis*) [@pcbi.1002018-Warburg1], [@pcbi.1002018-Warburg2]. Specifically, aerobic glycolysis allows the production of only 2 ATP molecules per one glucose molecule, whereas oxidative phosphorylation permits the generation of 32 ATP molecules per one molecule of glucose [@pcbi.1002018-Lehninger1]. Nevertheless, the importance of aerobic glycolysis to cancer cells has been experimentally demonstrated [@pcbi.1002018-Schulz1], [@pcbi.1002018-Ristow1]. Over the years, several hypotheses were raised regarding the potential cause of the Warburg effect: (i) Defective mitochondrion hypothesis -- suggesting that cancer cells have defective mitochondria and hence rely on glycolysis [@pcbi.1002018-Warburg3], however subsequent research revealed that mitochondrial function is not impaired in most cancer cells [@pcbi.1002018-Funes1], [@pcbi.1002018-Mori1]. (ii) Hypoxia-- suggesting that tumor hypoxia selects for cells dependent on anaerobic metabolism [@pcbi.1002018-Gatenby1], but previous studies have shown that cancer cells already resort to aerobic glycolysis before exposure to hypoxic conditions [@pcbi.1002018-Gottschalk1], [@pcbi.1002018-Elstrom1]. (iii) Avoiding ROS-mediated DNA damage -- it was suggested that reducing oxidative phosphorylation in proliferating cells due to the Warburg shift reduces ROS and hence protects cells from DNA damage and subsequent apoptosis [@pcbi.1002018-Chiaradonna1]. (iv) A game theoretical approach suggesting that the Warburg effect occurs as glycolysis provides higher ATP production rate than oxidative phosphorylation [@pcbi.1002018-Pfeiffer1], [@pcbi.1002018-Pfeiffer2], [@pcbi.1002018-Schuster1]. (v) An approach suggesting that a trade-off between the enzyme-synthesis costs and the ATP production yields of the different pathways that catabolize carbon sources may cause the Warburg effect: the high-yield oxidative phosphorylation pathway also has high enzyme costs, leading to a sub-optimal ATP production strategy, as it has lower production rates than glycolysis [@pcbi.1002018-Molenaar1]. (vi) Metabolic adaptation to fast proliferation - it was argued that as opposed to metabolism in differentiated cells that is geared towards efficient ATP production, the aerobic glycolysis observed in cancer cells is adapted to facilitate biomass accumulation and high proliferation. Accordingly, in order to satisfy the requirements of anabolic metabolism in addition to the production of ATP, nutrients must be used to generate both the carbon building blocks of macromolecules and the reducing power needed for biosynthesis [@pcbi.1002018-VanderHeiden1]. Previous computational investigations of the Warburg effect studied the role of either energy or biomass production in causing the Warburg effect, focusing solely on central carbon metabolism. For example, the study of Vander Heiden *et al.* manually computed the metabolic requirements for producing one essential biomass precursor, palmitate (a major constituent of cellular membranes) considering the stoichiometry of a few central metabolic pathways. They found that aerobic glycolysis enables maximal palmitate production yield due to specific reducing power requirements. In another recent study, Vazquez *et al.* employed a schematic model of ATP production in human cells (considering two lumped reactions representing aerobic glycolysis and oxidative phosphorylation), elegantly showing that a switch to aerobic glycolysis should result from cellular maximization of ATP production [@pcbi.1002018-Vazquez1]. Their schematic model accounts not only for the stoichiometry of glycolysis and oxidative phosphyrylation but also for the enzyme-volumetric costs of activating these pathways (the latter bounded by the total cellular solvent capacity, also known as a *macromolecular crowding* constraint [@pcbi.1002018-Schuster2]). A similar approach was previously employed in the study of over-flow metabolism in *E. coli* [@pcbi.1002018-Beg1], [@pcbi.1002018-Vazquez2]. Another interesting theory explaining overflow metabolism was suggested by Molenaar *et al.*, where the production costs of the metabolic enzymes involved were accounted for in a self-replicating model [@pcbi.1002018-Molenaar1]. In this paper, we study the causes of the Warburg effect by accounting for both energy production and anabolism of essential biomass constituents, in a genome-scale stoichiometric network model [@pcbi.1002018-Duarte1] employing enzyme solvent capacity constraints. The usage of a large-scale metabolic network is essential if one aims to correctly account for the inter-connectivity of pathways that produce the various energy and biomass precursors required for proliferation, rather than examining just single factors in isolation, as has been previously performed in [@pcbi.1002018-VanderHeiden1], [@pcbi.1002018-Vazquez1]. Towards this goal, we rely on a constraint-based modeling (CBM) framework that serves to analyze the function of metabolic networks by solely relying on simple physical-chemical constraints [@pcbi.1002018-Price1]. CBM has already been successfully used in the past to predict the metabolic state of various microorganisms [@pcbi.1002018-Feist1], [@pcbi.1002018-Mo1], and recently for studying human cellular metabolism [@pcbi.1002018-Duarte1]. The potential clinical utility of the human CBM model was previously demonstrated by its ability to identify functionally related sets of reactions that are causally related to hemolytic anemia, and potential drug targets for treating hypercholesterolemia [@pcbi.1002018-Duarte1], to predict metabolic biomarkers in inborn errors of metabolism [@pcbi.1002018-Shlomi1] and to predict a variety of metabolic behaviors of different human tissues, including the brain, liver, kidney and more [@pcbi.1002018-Shlomi2], [@pcbi.1002018-Jerby1]. Our analysis shows that while strictly stoichiometric considerations are insufficient for explaining the Warburg effect, the incorporation of enzyme solvent capacity constraints successfully predicts the emergence of the Warburg effect. The analysis is shown to accurately predict an experimentally observed metabolic trajectory occurring during oncogenic progression, as well as the preference of cancer cells for a high rate of glutamine uptake. Results {#s2} ======= Metabolic requirements of cellular proliferation subject to enzyme solvent capacity constraints lead to the Warburg effect {#s2a} -------------------------------------------------------------------------------------------------------------------------- We utilized a genome-scale human metabolic network that includes 3,742 reactions [@pcbi.1002018-Duarte1], adding a pseudo *biomass reaction* that represents the production of a pre-defined set of essential biomass precursors required for cellular proliferation, as conventionally done in Flux Balance Analysis (FBA, [@pcbi.1002018-Varma1], see [Methods](#s4){ref-type="sec"}). The biomass precursors include amino-acids, nucleotides, deoxy-nucleotides, ATP, lipids, etc (based on prior knowledge of their relative concentrations; [Methods](#s4){ref-type="sec"}). In our simulations, we assume a minimal growth medium with glucose as a carbon source, as glucose is known to serve as a major fuel in cancer cells (below and in [Text S1](#pcbi.1002018.s004){ref-type="supplementary-material"} we show that qualitatively similar results were obtained when considering also the presence of an additional major nutrient taken by cancer cells, glutamine). To predict plausible metabolic fluxes in cancer, we first employed a standard FBA method to identify a feasible flux distribution that satisfies stoichiometric mass-balance, while maximizing biomass production yield (see [Methods](#s4){ref-type="sec"}). We found that the predicted flux distribution does not display the prime characteristic of the Warburg effect, *i.e.* lactate secretion (see also [Text S1](#pcbi.1002018.s004){ref-type="supplementary-material"}). Interestingly, this finding is in accordance with a previous study showing a conceptually similar failure of FBA to predict the Crabtree effect in yeast, in which glucose is fermented into ethanol under aerobic conditions [@pcbi.1002018-Famili1]. Thus, stoichiometric considerations alone are insufficient for explaining the Warburg effect and its relation to the metabolic requirements of highly proliferating cells. Notably, these results stand in difference from those presented by Vander Heiden *et al.* [@pcbi.1002018-VanderHeiden1], claiming that strictly stoichiometric considerations directly lead to the Warburg effect due to metabolic demands for cellular proliferation. A strictly stoichiometric analysis, such as the one presented above, implicitly assumes that metabolic flux rates can be tuned to achieve high biomass production yields, without considering constraints imposed by enzyme concentrations and catalytic rates, which are prime determinants of metabolic flux. Specifically, while cells might be free to regulate enzyme concentrations according to metabolic demands [@pcbi.1002018-VanderHeiden1], the total enzymes\' concentration in the proliferating cells is bounded by the cell\'s solvent capacity, quantifying the maximum amount of macromolecules that can occupy the intracellular space [@pcbi.1002018-Vazquez1]. To account for the functional effects of this additional fundamental constraint, we follow [@pcbi.1002018-Vazquez1], [@pcbi.1002018-Vazquez2] and extend our stoichiometric genome-scale CBM analysis to compute for each enzyme the concentration required to facilitate the predicted flux, utilizing data on known human enzyme catalytic rates (taken from the literature; see [Methods](#s4){ref-type="sec"}). This modeling approach enables the prediction of metabolic flux distributions that maximize the biomass production rate and concomitantly obey the solvent capacity constraints -- rather than predicting flux distributions that only maximize the biomass production yield as done in standard FBA. We applied the approach described above (FBA with solvent capacity constraint) to predict human cellular flux distributions that maximize the biomass production rate. To simulate varying growth rates we performed the optimization across a wide range of different glucose uptake rates. Indeed, under these combined sets of constraints we find that biomass yield does decline at high growth rates -- in accordance with the Warburg effect [@pcbi.1002018-VanderHeiden1]; [Figure 1a](#pcbi-1002018-g001){ref-type="fig"}). Specifically, the predicted metabolic behavior manifests three distinct growth phases ([Figure 1b](#pcbi-1002018-g001){ref-type="fig"}): *(i) optimal yield metabolism* at a growth rate that is below 43% of the maximal possible rate, characterized by low glycolytic vs. high oxidative phosphorylation (OXPHOS) flux ([Figure 2a](#pcbi-1002018-g002){ref-type="fig"}, phase I), with low oxygen uptake rates ([Figure 1b](#pcbi-1002018-g001){ref-type="fig"}, phase I*). (ii) Intermediate yield metabolism* at growth rate between 43-92%, characterized by increased glycolytic and oxidative phosphorylation flux ([Figure 1a](#pcbi-1002018-g001){ref-type="fig"}, phase II), the latter involving a significantly increased oxygen consumption ([Figure 1b](#pcbi-1002018-g001){ref-type="fig"}, phase II). Notably, our prediction for an intermediate phase, involving increased oxygen consumption, presents a remarkable resemblance to two recent experimental studies examining the metabolic activity at different oncogenic progression stages ([@pcbi.1002018-deGroof1], [Figure 1c](#pcbi-1002018-g001){ref-type="fig"} and [@pcbi.1002018-Ramanathan1], [Figure 2b](#pcbi-1002018-g002){ref-type="fig"}). Neither the stoichiometric model [@pcbi.1002018-VanderHeiden1] nor an analysis using the schematic model of [@pcbi.1002018-Vazquez1] give rise to similar predictions. *(iii) Low yield metabolism* at a growth rate above 92% of the maximal possible growth rate, characterized by a sharp increase in glycolytic flux and a decrease in oxidative phosphorylation (and hence of O~2~ uptake). The increase in aerobic glycolysis flux ([Figure 2a](#pcbi-1002018-g002){ref-type="fig"}, phase III) leads to a rise in lactate secretion rates - a prime characteristic of the Warburg effect ([Figure 1b](#pcbi-1002018-g001){ref-type="fig"}, phase III). ::: {#pcbi-1002018-g001 .fig} 10.1371/journal.pcbi.1002018.g001 Figure 1 ::: {.caption} ###### Metabolic behavior across increasing growth rates. \(A) Predicted maximalgrowth yield of human cells (per unit of glucose uptake; y-axis) for a range of growth rates (x-axis), based strictly on reactions\' stoichiometry (dotted) and by considering also enzyme mass and enzyme solvent capacity (solid). Vertical dashed lines indicate the borders between: phase I (high yield, no lactate secretion), phase II (medium yield, increased oxidative phosphorylation) and phase III (low yield, lactate secretion). (B) Predicted lactate secretion flux (red lines) and oxygen consumption flux (blue lines) for a range of growth rates. Growth rates were manipulated by varying the glucose uptake rate limit from 0 until the uptake value needed to reach the maximal growth rate. Fluxes were normalized by the glucose uptake rate. (C) Experimentally determined lactate secretion rates (red; squares) and oxygen uptake rates (blue; circles) during tumor development of H-RasV12/E1A transformed fibroblasts. NRFU: Normalized relative fluorescence units; see [@pcbi.1002018-deGroof1] for more details. :::  ::: ::: {#pcbi-1002018-g002 .fig} 10.1371/journal.pcbi.1002018.g002 Figure 2 ::: {.caption} ###### Pathway activity differences. \(A) as predicted across phases I-III in the model and (B) based on experimental measurements taken from BJ fibroblast cell lines representing the path towards tumorigenic conversion (CL1-CL4; [@pcbi.1002018-Ramanathan1]). The model\'s predictions are compatible with the experimental evidences for increased glycolytic activity (expressed by increased lactate production) during full cancerous development (phase III, CL4) preceded by an increase in the oxidative phosphorylation (OXPHOS) activity (expressed by the mitochondrial gene expression). Experimental results for CL2-CL4 are given as the fold change relative to the same measurement in the CL1 cell line. In (B), the bars represent the mean fold change for each set of metabolites/genes and the error bars represent the standard deviation. :::  ::: To further validate the plausibility of the model, we examined the correlation between its enzyme concentration predictions (based on predicted flux distributions; see [Methods](#s4){ref-type="sec"}) and mRNA expression values measured for 1,269 metabolic genes across 60 cancer cell lines of the NCI-collection [@pcbi.1002018-Lee1]. The enzyme concentrations predicted with FBA accounting for the solvent capacity constraint show significant rank correlations with the gene expression data across the different cancer cell-lines (mean Spearman correlation of 0.28, mean p-value  =  6.5e−21). Notably, the strictly stoichiometric analysis provides significantly lower correlations with the expression measurements (with a mean correlation of 0.1; Wilcoxon p-value  =  3.5e−21), further demonstrating the advantage of the genome-scale approach that accounts for enzyme solvent capacity. Explaining the shift to aerobic glycolysis under high proliferation rates {#s2b} ------------------------------------------------------------------------- The shift towards *low yield metabolism* at high growth rates can be intuitively explained considering, on one hand, a flux distribution A with high growth yield (*Y~A~*) and high 'cost' in terms of the required enzyme concentrations per unit of glucose uptake (*C~A~*), and, on the other hand, a flux distribution B with a lower growth yield (*Y~B~*) and lower cost (*C~B~*). Considering a bound on the total enzyme concentration cost, one can observe that when the glucose uptake is unlimited flux distribution *B* will provide a higher growth rate if its growth yield normalized by its cost is higher than that of flux distribution A (*i.e.* ; [Figure 3](#pcbi-1002018-g003){ref-type="fig"}). When the glucose uptake rate is limited, maximal growth rate is achieved solely via flux distribution *A* or by a combination of *A* and *B*. ::: {#pcbi-1002018-g003 .fig} 10.1371/journal.pcbi.1002018.g003 Figure 3 ::: {.caption} ###### A plane describing the feasible region in our model. The axes (A, B) describe the growth rate obtained from flux distributions A and B, respectively. The blue lines represent two different constraints on the glucose uptake rate, and the red line represents the maximal concentration constraint. Green dashed lines are the contours of the growth rate maximization objective function -- the further the line is from the origin, the higher the growth rate. When the glucose uptake U is limiting (dark grey feasible region), the maximal growth rate is obtained via A only (Solution 1; left green diamond). When both the uptake and the enzyme concentration constraints are limiting (light grey feasible region), maximal growth rate (G) is obtained via a combination of A and B (Solution 2; right green diamond), resulting in a shift to a less efficient metabolism and lactate secretion. This can be explained by the fact that the slope of the growth-rate (middle green) line (-1) is larger than the slope of the enzyme concentration limit (red) line (), that is the yield-to-cost ratio of flux distribution B is greater than that of flux distribution A (). :::  ::: Concretely, analyzing the results of our model, flux distribution *A* stands for a typical metabolic state in phase I, which is characterized by high mitochondrial oxidative phosphorylation, with a high growth yield of 0.094 and a high cost of 0.302 (with a yield to cost ratio of 0.31). Flux distribution B stands for a typical metabolic state at phase III, which involves a high rate of aerobic glycolysis, with a low growth yield of 0.035 and a low cost of 0.050 (yield to cost ratio  =  0.7). These values indeed transcribe to a higher growth yield per unit of concentration cost of the enzymes participating in B, as alluded above. [Figure 3](#pcbi-1002018-g003){ref-type="fig"} shows that, indeed, at low growth rates the glucose uptake rate is the sole limiting factor and hence the high yield oxidative phosphorylation route is taken; in contrast, at higher growth rates, the enzyme concentration constraint takes effect, and mixed solutions involving lactate secretion are necessarily formed. Notably, the predicted flux distributions across the range of growth rates described in this paper cannot be obtained from linear combinations of just two states (as in the above simplified example), but are rather composed of multiple flux distributions with different growth yields per concentration cost (as evident for example by the non-linear curve showing the predicted oxygen uptake rates across growth rates; [Figure 1b](#pcbi-1002018-g001){ref-type="fig"}). Thus, the flux distributions actually obtained in genome-scale models markedly differ from those that can be captured by a simplified analysis that describes the transition between just two metabolic states with different growth yields as above, or as in a previous study of Vazquez *et al.* [@pcbi.1002018-Vazquez1]. Metabolic adaptation to fast proliferation leads to a preference to high glutamine uptake rates {#s2c} ----------------------------------------------------------------------------------------------- The role of glutamine in cancer has been a topic of major interest as cancer cells are known to have a significant high glutamine uptake rate [@pcbi.1002018-DeBerardinis1]. Repeating the previous analyses in the presence of both glucose and glutamine in the growth media shows qualitatively similar results to those described above. However, as expected, the addition of glutamine yields a higher maximal biomass production rate than the one obtained when only glucose was available in the medium ([Text S1](#pcbi.1002018.s004){ref-type="supplementary-material"}). To investigate the preference of cancer cells specifically to glutamine over other amino-acids, we applied FBA analysis to predict the contribution of each amino-acid separately to biomass production in the human model that accounts for enzyme solvent capacity constraints ([Methods](#s4){ref-type="sec"}). We find that, indeed, the contribution of glutamine to the proliferation rate is markedly higher than that of all other amino-acids ([Figure 4B](#pcbi-1002018-g004){ref-type="fig"}). We further show that this result is robust to changing the bound on maximal amino-acid uptake rate, and that it remains valid across a large number of random samplings of enzyme turnover rates ([Text S1](#pcbi.1002018.s004){ref-type="supplementary-material"}). Repeating this analysis without accounting for enzyme solvent capacity constraints (*i.e.* by considering only the network stoichiometry in the vanilla FBA model) fails to predict the preference for glutamine ([Figure 4A](#pcbi-1002018-g004){ref-type="fig"}). ::: {#pcbi-1002018-g004 .fig} 10.1371/journal.pcbi.1002018.g004 Figure 4 ::: {.caption} ###### Amino-acid growth rate contribution. The increase in proliferation rate achievable by the increased uptake of each of the 20 amino-acids in addition to glucose (relative to the baseline growth rate achieved when only glucose is available), as predicted by the stoichiometric model (A) and by the model accounting for the solvent capacity constraint (B). Glutamine uptake (highlighted in yellow) enables to achieve the highest increase in growth rate according to the solvent capacity model, in agreement with experimental data showing preference for high glutamine uptake rates in cancer. :::  ::: We carefully examined the flux distribution obtained with glutamine in the growth medium (achieving a growth rate of 0.062 1/h) vs. the one obtained with glutamate (growth rate  =  0.056 1/h). Interestingly, when glutamate is present in the medium, a large quantity of it is transformed into glutamine in an ATP consuming reaction catalyzed by the enzyme glutamine synthetase (EC 6.3.1.2). This satisfies the glutamine biomass requirement as well as the production of nucleotide precursors, among others. When removing the ATP requirement from this reaction, the growth rate achieved with glutamate in the medium increases to 0.059 1/h, which explains 50% of the growth rate difference. Notably, while this provides some intuitive explanation for the predicted preference for glutamine, we cannot identify a simple explanation for the entire effect due to the high complexity of the network model employed. Discussion {#s3} ========== Metabolic adaptation to elevated growth requirements during cancer development has been recently suggested as the possible cause of the Warburg effect, a long-standing enigma of cancer metabolism. In this work we rigorously study this hypothesis using a genome-scale human metabolic model and demonstrate that stoichiometric considerations solely are insufficient to explain the shift to inefficient metabolism, in difference from recent claims [@pcbi.1002018-VanderHeiden1]. However, integrating these constraints in a genome-scale model of human metabolism together with a constraint on enzyme solvent capacity does lead to the emergence of the Warburg effect at high proliferation rates. Furthermore, it accurately predicts a three phase metabolic behavior experimentally observed during oncogenic progression, as well as a marked preference to a high uptake rate of glutamine. The importance of enzyme solvent capacity in metabolic modeling has already been recognized in the earlier work of Beg *et al.* [@pcbi.1002018-Beg1], where applying such a constraint to the *E. coli* model improved phenotypic predictions. In their work, however, Beg *et al.* assumed an upper bound on the total cell-volume occupied by metabolic enzymes, as opposed to the method introduced here where we assume a bound on the enzyme mass per cell mass (*i.e.* a bound on enzyme fractional concentration). In order to account for enzyme volumes, Beg *et al.* estimated enzyme volumes by assuming a uniform *specific volume* parameter (representing the ratio between enzyme mass and volume) for all enzymes. Here, we employed a simpler approach that does not depend on specific volume estimations, and explicitly constrains the total sum of enzyme mass. Notably, we further tested the effect of accounting for volumes instead of masses, and obtained results which are very similar to those obtained with masses only ([Text S1](#pcbi.1002018.s004){ref-type="supplementary-material"}). In a recent study by Molenaar *et al.* [@pcbi.1002018-Molenaar1], a metabolic shift at high growth rates was predicted based on a general self-replicating model. Another recent work (by Vazquez *et al.*) already pointed to the significance of the solvent-capacity constraint in explaining the Warburg effect [@pcbi.1002018-Vazquez1]. Notably though, the work presented here provides a marked contribution over both studies: First, both employ abstract small-scale models. Specifically, the Vazquez *et al.* work introduces a schematic model of ATP production in central metabolism including just a handful of variables. Furthermore, similarly to the work of Pfeiffer *et al.*, their work does not explicitly account for the entire biomass composition and the associated energy requirements. In contrast, here we study a genome-scale biomass producing human model that, despite the scores of alternative biomass and energy production pathways existing in the human network, successfully shows that highly proliferating cells such as cancer cells are forced to display Warburg related phenotypes at high growth rates (phase III). Additionally, and in contrast to the small-scale models, our genome-scale model correctly predicts an experimentally observed transitional phase (II). Furthermore, on a mechanistic level, the genome-scale metabolic description provided by our analysis is significantly correlated with the gene expression patterns across the wide array of NCI-60 cancer cell-lines (much stronger than the association displayed by the stoichiometric model alone), a result which could not have been predicted by the Vazquez *et al.* model. Lastly, the model was able to predict the marked contribution of glutamine to rapid cellular growth. As a further demonstration of the robustness of our results, we repeated the analyses using a model accounting for maintenance ATP production, obtaining qualitatively similar results ([Text S1](#pcbi.1002018.s004){ref-type="supplementary-material"}). While the data on reactions\' stoichiometry is considered accurate and comprehensive, enzyme kinetic constant data are noisy and are currently available for only about 15% of the reactions in the model. In the analysis presented here, we addressed this problem by assigning enzymes with missing turnover rates with the median rate computed over the set of known turnover rates. Notably, the model\'s main findings are robust to random sampling of turnover rates from a distribution of known rates, as shown in [Text S1](#pcbi.1002018.s004){ref-type="supplementary-material"}. However, repeating the analysis when assigning all reactions in the model with the median turn-over rate shows no Warburg characteristics - testifying to the importance of utilizing known turnover rates even if this data is sparse. Future measurements of additional enzyme turnover rates and improved methods for accurately predicting these parameters (*e.g.* [@pcbi.1002018-Borger1]) are expected to further refine the predictions of cancer metabolic phenotypes using stoichiometric metabolic models with an enzyme solvent capacity constraint. In our work we accounted for a solvent capacity constraint assuming a limited protein mass per cell, without considering the effect of enzymes\' sub-cellular compartmentalization. To investigate how the latter would affect our predictions, we repeated the analysis while considering separate solvent capacity constraints for cytoplasm and mitochondria ([Text S1](#pcbi.1002018.s004){ref-type="supplementary-material"}), yielding quantitatively similar results to those described above. The incorporation of solvent capacity constraints for different cellular compartments may lead to further improved prediction accuracy in the future, when additional data on enzyme turnover rates becomes available. Specifically, the addition of membrane-specific constraints may be a promising direction, as many metabolically important proteins are confined to membranes (e.g. those of respiratory chain and membrane biosynthesis). The presented modeling approach is likely to contribute to more accurate metabolic modeling of highly proliferating human cells in general (as was already shown regarding genome-scale models of microorganisms [@pcbi.1002018-Vazquez2]) and of cancer cells. The latter may be in turn utilized for anti-cancer drug target prediction and specifically, for predicting drugs that work to reverse the Warburg effect. While the current analysis has relied on the available human generic model, future studies may utilize a similar methodology in modeling the metabolism of specific cancers. These may be generated by integrating cancer-signature expression data with the generic human model to carve out different cancer types models (using methods such as those outlined in [@pcbi.1002018-Shlomi2] or [@pcbi.1002018-Jerby1]), and thus further advance the development of anti-cancer drugs specific to different cancers. Materials and Methods {#s4} ===================== Modeling biomass production using a stoichiometric model {#s4a} -------------------------------------------------------- The Duarte *et al.* [@pcbi.1002018-Duarte1] human genome-scale metabolic model, accounting for 1,496 ORFs, 3,742 reactions and 2,766 metabolites, was used. The metabolic network is represented in a stoichiometric matrix *S*, where *m* is the number of metabolites, *n* is the number of reactions, and represents the stoichiometric coefficient of metabolite *i* in reaction *j*. Biomass production was modeled by adding a new growth reaction to the human model: this reaction was compiled using the steady state concentrations of 30 biomass compounds including amino acids (0.78 g/gDW; [@pcbi.1002018-Barle1], [@pcbi.1002018-Triguero1]), nucleotides (0.06 g/gDW; [@pcbi.1002018-Sheikh1]), lipids (0.16 g/gDW; [@pcbi.1002018-Rabinowitz1]) as well as the growth-associated energy requirement (24 mmol/gDW of ATP; [@pcbi.1002018-Kilburn1]). Essential amino acids were not accounted for since they were assumed not to take active part in the metabolic model besides flowing directly into the biomass reaction. The full list of biomass metabolites and their relative concentrations is available in [Dataset S1](#pcbi.1002018.s001){ref-type="supplementary-material"}. The biomass reaction was defined as the objective function of the CBM method Flux Balance Analysis (FBA; [@pcbi.1002018-Varma1]). FBA looks for a flux distribution *v* that maximizes the objective function (Equation 1) subject to steady-state, thermodynamic and growth medium constraints: Equation 2 imposes the steady state constraints on the system, assuming that the metabolite concentrations remain constant in time. Thermodynamic constraints determining the reaction directionalities are accounted for via the flux limits and in Equation 3. The uptake and secretion of a pre-defined set of metabolites from and to the environment is facilitated via the definition of exchange reactions in the stoichiometric matrix. The growth medium is defined via an upper bound on the glucose uptake exchange reaction (as the carbon source) and by allowing an unlimited uptake flux of oxygen, sodium, potassium, calcium, iron, chlorine, phosphate, sulfate and ammonia (based on the RPMI- 1640 medium definition; as none of these substances can be used as a carbon source). Growth yield (growth rate divided by the glucose uptake rate), oxygen uptake and lactate secretion rates were computed under a wide range of glucose uptake rates (varying from 0 to 1.55 umol/mgDW/h, the uptake achieving maximal growth rate) using Flux Variability Analysis (FVA) [@pcbi.1002018-Mahadevan1], allowing us to determine minimal and maximal flux bounds on the reactions of interest. Accounting for enzyme solvent capacity {#s4b} -------------------------------------- A constraint on the total enzyme concentration was added to the biomass production FBA model: The enzyme mass (per mg dry weight (DW) of cells) required to maintain the flux in the i-th reaction (*v~i~* \[mmol/(mgDW\*h)\]) is given by the product of *v~i~* and the enzyme\'s molecular weight (*MW~i~* \[mg/mmol\]) divided by its turnover number ( \[1/h\]) [@pcbi.1002018-Vazquez2]. The limit on the total metabolic enzyme mass (*C* = 0.078 \[mg/mgDW\]) was estimated based on dry cell weight protein biomass measurements (0.779 \[mg/mgDW\]; [@pcbi.1002018-Davidson1]) multiplied by the fraction of metabolic genes out of the total cellular protein mass, which was evaluated as the sum of metabolic gene expression readouts divided by the total sum of gene expression readouts ([@pcbi.1002018-Lee1]; equal to 0.1). Notably, the reliability of this value was validated based on a recently published protein abundance dataset ([@pcbi.1002018-Vogel1], [Text S1](#pcbi.1002018.s004){ref-type="supplementary-material"}). In order to account for positive fluxes only, each bidirectional reaction was split into two unidirectional reactions, resulting in a total of 4,894 reactions. Enzyme molecular weights were obtained from the BRENDA database ([@pcbi.1002018-Schomburg1]; [Dataset S2](#pcbi.1002018.s002){ref-type="supplementary-material"}) while turnover number data was taken from BRENDA and from the SABIO-RK databases ([@pcbi.1002018-Rojas1]; [Dataset S3](#pcbi.1002018.s003){ref-type="supplementary-material"}), and assigned as following: each reaction with a known Enzyme Commission (EC) number was queried against BRENDA for the maximal human wild-type *k~cat~* value. In case a human *k~cat~* value was not available, the maximal non-human wild-type turnover number was assigned. In case BRENDA data was not available, the SABIO-RK database was used in a similar manner. As a result, 729 reactions were assigned with *k~cat~* values while the other 4,165 reactions were assigned with the median *k~cat~* value across the set of known *k~cat~* values (25 1/s). Pathway activity analysis {#s4c} ------------------------- Flux distributions were computed under maximal growth rates in the three growth phases (phase I -- 0.0243 1/h; phase II -- 0.0515 1/h; phase III -- 0.0557 1/h). For each phase, the median flux distribution across 1000 different uniform samples was calculated using ACHR sampling [@pcbi.1002018-Kaufman1]. Mean pathway flux was calculated as the mean flux across the reactions belonging to the pathway of interest. Data on relative metabolomic measurements for lactate, and on relative transcriptomic measurements for genes which are important for mitochondrial biogenesis (PGC-1-α, NRF-1, TFAM and ATP5E) was taken from [@pcbi.1002018-Ramanathan1]. Correlation with gene expression data {#s4d} ------------------------------------- Gene expression readouts for 1,269 metabolic genes across 60 cell lines from the NCI-60 collection [@pcbi.1002018-Lee1] were correlated with enzyme concentrations predicted by (i) a stoichiometric only model and by (ii) a model accounting also for enzyme solvent capacity. Given a flux distribution vector *v*, for each reaction *i*, the enzyme concentration needed to maintain its flux (the i-th entry in *v*) was calculated as the product ofand the molecular weight of the enzyme catalyzing this reaction (denoted ), divided by its turnover number (denoted ), that is, . Total enzyme concentrations (per gene) were given by summing the enzyme concentrations across all of the reactions associated with the gene of interest (*i.e.* reactions catalyzed by enzymes encoded by this gene), based on a gene-to-reaction mapping given in the human metabolic model. The Spearman correlation between the gene expression vector and inferred enzyme concentration vector was calculated for the two models in each of the 60 cell lines. The robustness of the results was validated against 1,000 uniformly sampled flux distributions from the solution spaces of the two models using ACHR sampling [@pcbi.1002018-Kaufman1]. Modeling amino-acid uptakes {#s4e} --------------------------- Each of the 20 amino acids was added, in turn, to the growth media, resulting in 20 different maximal biomass production rates calculated based on (i) the stoichiometric model, and on (ii) a model additionally accounting for the solvent capacity constraint. The maximal amino-acid uptake rate was set to the same uptake rate as glucose; the results are shown to be robust to the choice of this value ([Text S1](#pcbi.1002018.s004){ref-type="supplementary-material"}). Supporting Information {#s5} ====================== Dataset S1 ::: {.caption} ###### Human biomass composition. (XLSX) ::: ::: {.caption} ###### Click here for additional data file. ::: Dataset S2 ::: {.caption} ###### Enzyme molecular weight data for the reactions in the model. (XLSX) ::: ::: {.caption} ###### Click here for additional data file. ::: Dataset S3 ::: {.caption} ###### Enzyme turnover number data for the reactions in the model. (XLSX) ::: ::: {.caption} ###### Click here for additional data file. ::: Text S1 ::: {.caption} ###### Validating the robustness of the results to various model parameters and exploring additional changes in the model. (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Special thanks to Christoph Kaleta and Stefan Schuster for their many detailed comments and suggestions that have served to significantly improve our manuscript. The authors have declared that no competing interests exist. This research was supported by an Edmond J Safra Bioinformatics fellowship to T.B. and grants from the Israeli Cancer Research Fund (ICRF, <http://www.icrfonline.org>, PG-10-3099) to T.S and E.R and from the Israeli Science Foundation (ISF, <http://www.isf.org.il>) to T.S. (ISF grant no. 1198/09) and to E.R. (ISF grant no. 1085/09). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: ER TS TB. Performed the experiments: TB. Analyzed the data: ER TS TB. Contributed reagents/materials/analysis tools: ER TS. Wrote the paper: ER TS TB. Analyzed the results: ER TS TB RS EG.

PubMed Central

2024-06-05T04:04:19.687751

2011-3-10

PMC3053320

Introduction {#s1} ============ MicroRNAs (miRNAs) are endogenous small non-coding RNAs which negatively regulate the protein production of their targets in metazoans and plants. They are expected to target a substantial portion of the human genome [@pcbi.1001101-Flynt1] and have been shown to play key roles in several biological processes ranging from development and metabolism to apoptosis and signaling pathways [@pcbi.1001101-Ambros1]--[@pcbi.1001101-Stefani1]. Moreover their profiles are altered in several human diseases [@pcbi.1001101-AlvarezGarcia1], [@pcbi.1001101-EsquelaKerscher1], making miRNAs a major focus of research in nowadays molecular biology. Recent work, reviewed in [@pcbi.1001101-Martinez1], has shown that the actions of miRNAs and transcription factors (TFs) are often highly coordinated, suggesting that the transcriptional and post-transcriptional layers of regulation are strongly correlated and that miRNA functions can be fully understood only by addressing TF and miRNA regulatory interactions together in a single "mixed" network. As in the case of purely transcriptional networks [@pcbi.1001101-Milo1], in this mixed network several recurrent wiring patterns can be detected, called network motifs [@pcbi.1001101-Re1]--[@pcbi.1001101-Yu1]. The common lore is that network motifs were selected by evolution (and are thus overrepresented in the network) to perform elementary regulatory functions. Among these motifs one of the most interesting is the miRNA-mediated feedforward loop (FFL) in which a master TF regulates a miRNA and, together with it, a set of target genes (see [Figure 1](#pcbi-1001101-g001){ref-type="fig"}). This motif, which shall be the main interest of our paper, was recently discussed in [@pcbi.1001101-Re1]--[@pcbi.1001101-Tsang1]. In all these papers, despite the fact that the authors used very different computational approaches, the FFL was shown to be remarkably overrepresented in the network, thus supporting the idea that it should play an important regulatory role. Depending on the sign of the transcriptional regulations, FFLs can be divided into two classes: coherent and incoherent [@pcbi.1001101-Re1], [@pcbi.1001101-Tsang1], [@pcbi.1001101-Hornstein1]. In the coherent FFLs both pathways from the TF to the target have the same effect (both repressing or activating target expression), while in the incoherent ones the two pathways have opposite effects. Correspondingly one finds different expression patterns in the two cases: coexpression of miRNA and its target for incoherent FFLs and mutually exclusive expression for the coherent ones ([Figure 1](#pcbi-1001101-g001){ref-type="fig"}). This dual picture allows to better understand the complex patterns of correlated expression of miRNAs and their targets observed in experiments [@pcbi.1001101-Flynt1], [@pcbi.1001101-Tsang1], [@pcbi.1001101-Shkumatava1]. In many cases the targets show low expression in miRNA-expressing cells, suggesting coherent regulation. On the other hand, several other cases present an opposite trend, showing that miRNA repression can act in opposition to transcriptional regulation. Indeed, different degrees of expression overlap, due to different regulatory circuitries, have been related to different miRNA functions in several recent papers [@pcbi.1001101-Flynt1], [@pcbi.1001101-Bartel1], [@pcbi.1001101-Bushati1], [@pcbi.1001101-Hornstein1], [@pcbi.1001101-Bartel2]. For example, in a coherent FFL as the one in [Figure 1D](#pcbi-1001101-g001){ref-type="fig"}, the miRNA expression is induced by an upstream TF that at the same time represses the target transcription, with the effect of enforcing mutually exclusive domains of expression as the ones observed in the fruit fly [@pcbi.1001101-Stark1] or for miR-196 and its target Hoxb8 in mouse [@pcbi.1001101-Mansfield1] and chicken [@pcbi.1001101-Hornstein2]. In this cases the miRNA can help the transcriptional repression of a target protein that should not be expressed in a particular cell type, acting as a post-transcriptional failsafe control. Instead, an incoherent FFL ([Figure 1C](#pcbi-1001101-g001){ref-type="fig"}) can promote high target expression in miRNA-expressing cells, suggesting that miRNAs may have in this case a fine-tuning function, keeping the protein level in the correct functional range. A typical example is the regulation of the atrophin gene by the miRNA miR-8 in *Drosophila*. It was shown [@pcbi.1001101-Karres1] that both a too high and a too low level of expression of the atrophin gene could be detrimental for the organism and that miR-8 is mandatory to keep the expression level exactly in the correct range. ::: {#pcbi-1001101-g001 .fig} 10.1371/journal.pcbi.1001101.g001 Figure 1 ::: {.caption} ###### Overview of the connections between miRNA-target expression, miRNA function and regulatory circuitry. \(A) MiRNAs and corresponding targets can present different degrees of coexpression between the two extremes of concurrent expression (high correlation) and exclusive domains (high anticorrelation). These two opposite situations are expected to correspond to different functional roles (B) for the miRNA repression. A peculiar expression pattern, evidence of a functional aim, is a consequence of the network structure in which miRNAs are embedded. A high miRNA-target correlation can be achieved through the incoherent FFL (C), where the miRNA repression is opposite to the TF action. Whereas a failsafe control can be performed by a coherent FFL (D), in which the miRNA reinforces the TF action leading to mutually exclusive domains of miRNA-target expression. :::  ::: It is by now well understood that gene espression is inherently a stochastic process [@pcbi.1001101-Kaern1]--[@pcbi.1001101-Maheshri1]. This has particularly relevant effects when the number of proteins and/or messenger RNAs (mRNAs) involved is small and stochastic fluctuations may give sizeable deviations from the mean value of the final protein product. Thus, the question that naturally arises is how the cell can reconcile the fine-tuning function described above with these fluctuations. If there is only a relatively narrow protein level which is optimal, the tuning regulation must also prevent protein fluctuations outside the functional range. In fact, it has been conjectured that the incoherent FFLs that enable tuning interaction, can also have a role in buffering noise in the target expression [@pcbi.1001101-Tsang1], [@pcbi.1001101-Hornstein1], [@pcbi.1001101-Wu1]. The main goal of our paper is to introduce and solve analytically a stochastic model describing these incoherent FFLs in order to give a proof to this conjecture. Our results show that with respect to the simple gene activation by a TF, the introduction of a miRNA-mediated repressing pathway can significantly dampen fluctuations in the target protein output for essentially all the choices of input parameters and initial conditions. As a test of our analysis we also performed extensive numerical simulations which nicely agree with our analytical results. It is important to stress (and we shall discuss this issue in detail in the following) that this noise buffering function is actually a precise consequence of the peculiar topolgy of the FFL. In fact, in order to fine-tune the level of a target protein one would not necessarily need a FFL topology. The same result could well be obtained with an independent miRNA (not under the control of the master TF which activates the target), but this choice would lead to strong fluctuations in the target expression. In the same theoretical framework we can show that the construction of an optimal noise filter does not necessarily imply a strong repression, in agreement with the observation that the miRNA down-regulation of a target is often modest [@pcbi.1001101-Baek1], [@pcbi.1001101-Selbach1]. Results {#s2} ======= The theoretical framework {#s2a} ------------------------- Here we focus on the incoherent FFL in [Figure 2A](#pcbi-1001101-g002){ref-type="fig"} to present our modeling strategy. For each gene in the circuit we take into account the essential features of transcription, translation, degradation and interactions between genes in the regulatory network (detailed scheme in [Figure 2A](#pcbi-1001101-g002){ref-type="fig"}′). Accordingly, the state of the system is described by five variables representing: the number of mRNAs transcribed from the TF gene, the number of TF molecules, the number of miRNAs, the number of mRNAs transcribed from the target gene and the number of target proteins. We want to explore the mean () and the standard deviation () of each molecular species and we will show that these quantities can be obtained analitically at the steady-state. The noise strength of the species will be expressed by the coefficient of variation defined as . As usual in this type of models, transcriptional activation is introduced by choosing the rate of transcription of the regulated gene ( in our case) as a nonlinear increasing function of the number of TFs () present in the cell [@pcbi.1001101-Alon1]--[@pcbi.1001101-Bintu1]: ::: {#pcbi-1001101-g002 .fig} 10.1371/journal.pcbi.1001101.g002 Figure 2 ::: {.caption} ###### Representation of the incoherent FFL and the two circuits used for comparison. \(A) A miRNA-mediated incoherent FFL that can be responsible for miRNA-target coexpression; (B) a gene activated by a TF; (C) an open circuit that leads to the same mean concentrations of the molecular species of the FFL in scheme A. (A′)(B′)(C′) Detailed representation of the modelization of the corresponding circuits. Rectangles represent DNA-genes, from which RNAs () are transcribed and eventually degraded (broken lines). RNAs can be translated into proteins ( is the TF while is the target protein) symbolized by circles, and proteins can be degraded (broken circles). Rates of each process (transcription, translation or degradation) are depicted along the corresponding black arrows. Regulations are represented in red, with arrows in the case of activation by TFs while rounded end lines in the case of miRNA repression. TF regulations act on rates of transcription that become functions of the amount of regulators. MiRNA regulation makes the rate of translation of the target a function of miRNA concentration. :::  ::: where and are dissociation constants, specifying the amount of TFs at which the transcription rate is half of its maximum value ( and respectively). is the Hill coefficient and fixes the steepness of the activation curve. The miRNA action can direct translational repression or destabilization of target mRNAs [@pcbi.1001101-ValenciaSanchez1], i.e. it decreases the rate of translation or increases the rate of degradation of target mRNAs. We choose to model the effect of miRNA regulation by taking the translation rate of the target () to be a repressive Hill function of the number of miRNAs (): The parameter specifies the quantity of miRNAs that determines a rate of translation , and is again the Hill coefficient. For simplicity we use the same Hill coefficient for each Hill function, but the analysis can be straigthforwardly generalized to the case of different steepnesses. The alternative choice of a degradation rate of mRNAs as a function of miRNA concentration does not yield significantly different results, as reported in [Text S1](#pcbi.1001101.s001){ref-type="supplementary-material"}. The use of Hill functions to model regulations by miRNAs is coherent with their established catalytic action in animals [@pcbi.1001101-Alberts1]. A stoichiometric model has been studied in the context of sRNA regulation in bacteria [@pcbi.1001101-Levine1]--[@pcbi.1001101-Shimoni1], in which each sRNA can pair with one messenger and drive its sequestration or degradation in an irreversible fashion. A comparison with a possible stoichiometric action is shown in [Text S1](#pcbi.1001101.s001){ref-type="supplementary-material"}. The probability of finding in our cell exactly molecules at time satisfies the master equation: where are transcription rates, are translation rates, and represents the degradation rate of the species . In order to solve the master equation for and for all at the steady state we have to linearize Hill functions. This is by now a standard procedure [@pcbi.1001101-Komorowski1], [@pcbi.1001101-Thattai1]. The idea is that at the steady state the distributions of regulators (TFs or miRNAs) have a finite width and sample only small regions of the domains of the corresponding Hill functions. We may therefore approximate Hill functions by their linearizations around the mean values of the regulators or (see [Text S1](#pcbi.1001101.s001){ref-type="supplementary-material"} for details of the linearization), ending up with: We would like to emphasize that linearizing the Hill functions does not mean to linearize the model. In fact, even with a linearized dependence on the miRNA copy number, our model keeps a nonlinear contribution in the term encoding the target translation (due to the fact that it depends on both the number of miRNAs and mRNAs). As we will see later, this nonlinearity leads to non trivial consequences. Despite this nonlinearity, the moment generating function approach [@pcbi.1001101-Komorowski1], [@pcbi.1001101-Thattai1], [@pcbi.1001101-Shahrezaei1] can be succesfully used. By defining the generating function:and using the linearization in equation 4 we can convert equation 3 into a second-order partial differential equation: We now use the following properties of the moment generating function: ; ; where means evaluation of at for all . At the steady state () differentiation of equation 6 generates equations for successively higher moments. In particular, we are interested in and and differentiating up to the fourth moments leads to their analytical expressions (see [Text S1](#pcbi.1001101.s001){ref-type="supplementary-material"} for details of the calculation). Noise in protein expression is originated by the combination of two types of fluctuations: intrinsic and extrinsic ones. Intrinsic fluctuations are essentially due to the stochasticity of the gene expression process. Extrinsic ones, instead, are due to the environment. In the latter case a prominent role is played by the noise transmitted by upstream genes [@pcbi.1001101-Pedraza1], [@pcbi.1001101-Volfson1]. As a matter of fact there is a certain degree of arbitrariness in the definition of extrinsic and intrinsic noise [@pcbi.1001101-Paulsson1]. Since we focus on the target production we define "intrinsic" the noise derived from the stochastic steps of its expression (transcription, translation and degradation) and "extrinsic" the noise propagating from its regulators () that makes the parameters () fluctuate through the Hill functions. Therefore in our model we do not have to include extrinsic noise with an arbitrary distribution as it naturally arises from the stochastic steps of production of regulators and propagates to the target gene. Comparison with a TF transcriptional control {#s2b} -------------------------------------------- To show the noise buffering role of the miRNA-mediated incoherent FFL ([Figure 2A](#pcbi-1001101-g002){ref-type="fig"}) we first compare it to a simpler process: a gene activated by a TF ([Figure 2B](#pcbi-1001101-g002){ref-type="fig"}), without any post-transcriptional regulation. The strategy used to model this linear network is equivalent to the one explained in the previous section for the FFL (see [Text S1](#pcbi.1001101.s001){ref-type="supplementary-material"} for more details) and it is presented schematically in [Figure 2B](#pcbi-1001101-g002){ref-type="fig"}′. Starting from a gene activated by a TF, in principle the gain of a new regulator implies also a new source of extrinsic noise for the target, given that the fluctuations in the number of regulators propagate to downstream genes and, as discussed in [@pcbi.1001101-Shahrezaei2], the addition of extrinsic fluctuations generally increases the noise of a system. However, the peculiar structure of the FFL can overcome this problem, actually reducing noise, as was previously shown in the case of negative transcriptional auto-regulation [@pcbi.1001101-Becksei1]. Given that the two circuits lead to different mean values, the comparison of noise strengths in target protein will be done in terms of the coefficient of variation (). With the parameter choice explained in the caption of [Figure 3](#pcbi-1001101-g003){ref-type="fig"}, the predicted are 0.147 and 0.1 for the TF-gene cascade and the FFL respectively. Therefore the introduction of the miRNA pathway not only controls the mean value but also reduces the relative fluctuations. This effect can be clearly seen looking at the shape of the probability distributions in [Figure 3C](#pcbi-1001101-g003){ref-type="fig"}. It is rather easy to understand the origin of this noise buffering effect: any fluctuation in the concentration of TFs affects the rate of mRNA transcription, driving consequently the target protein away from its steady state, but mRNA and miRNA concentrations tend to vary in the same direction in the FFL. In this way, miRNAs can always tune the protein production against TF fluctuations. As can be seen in [Figure 3A and B](#pcbi-1001101-g003){ref-type="fig"}, there is a certain degree of correlation in the time evolution of due to noise propagation, despite the overimposed higher-frequency intrinsic noise of each molecular species, but in the case of the FFL the trajectory is less sensitive to fluctuations thanks to the action of miRNAs (). It is important to stress that this result is not affected by the Hill function linearization discussed above. In fact, the predictions of the model are in good agreement with Gillespie simulations (which keep into account the full nonlinear repressing and activating Hill functions). Moreover our results turn out to be robust with respect to parameter choice, showing a rather stable noise reduction essentially for any choice of expression and degradation constants (see [Text S1](#pcbi.1001101.s001){ref-type="supplementary-material"} for details). ::: {#pcbi-1001101-g003 .fig} 10.1371/journal.pcbi.1001101.g003 Figure 3 ::: {.caption} ###### Noise properties of the FFL compared with a TF-gene linear circuit. \(A) An example of simulation results for the FFL (scheme on the right or more detailed in [Figure 2A](#pcbi-1001101-g002){ref-type="fig"}′). The normalized trajectory of each molecular species is shown as a function of time after reaching the steady state. The rate of transcription of the TF is and of translation . Proteins degrade with a rate , while mRNAs and miRNAs with . The parameters in the Hill functions of regulation (equations 1,2) are the following: the maximum rate of transcription for mRNAs is , while for miRNAs is ; the maximum rate of translation of the target is ; dissociation constants are ; Hill coefficients are all , as typical biological values range from 1 (hyperbolic control) to 30 (sharp switching)[@pcbi.1001101-Thattai1]. (B) Time evolution in a simulation for the molecular players in the linear TF-gene cascade (scheme on the right or more detailed in [Figure 2B](#pcbi-1001101-g002){ref-type="fig"}′). Compared to the FFL case, the evolution is more sensitive to TF fluctuations. (C) The probability distribution of protein number for the two circuits. Histograms are the result of Gillespie simulations while continuous lines are empirical distributions (gaussian for the FFL and gamma for the TF-gene) with mean and variance predicted by the analytical model. :::  ::: Comparison with an open regulatory circuit {#s2c} ------------------------------------------ The same fine-tuning of the mean target concentration achieved with a FFL could be equally obtained with an open circuit like the one in [Figure 2C](#pcbi-1001101-g002){ref-type="fig"}, where the miRNA gene is controlled by an independent TF. If the two TFs, activating the miRNA and target gene expression, have the same rate of transcription, translation and degradation of the single master TF in the FFL -as well as the other model parameters as in [Figure 2A](#pcbi-1001101-g002){ref-type="fig"}′ and C′- the stationary mean levels of the various molecular species are the same in both circuits. In particular, the mean concentration of the target protein can be fine-tuned to the same desired value by both circuits. However, while the deterministic description at the steady state is the same in the two cases (see [Text S1](#pcbi.1001101.s001){ref-type="supplementary-material"} for details) the behaviour of fluctuations is completely different. As we shall see below, the open circuit leads to much larger fluctuations in the final product than the FFL. It is well possible that this is the reason for which FFLs have been positively selected by evolution and are presently overrepresented in the mixed TF-miRNA regulatory network. In fact, fine-tuning can be implemented advantageously only together with a fluctuation control: a precise setting of the mean value of a target protein is useless without a simultaneous damping of the stochastic fluctuations. To assess this result we used the same strategy discussed above: we solved analitically for both circuits the master equation and tested our results with a set of Gillespie simulations. Our results are shown in [Figure 4:](#pcbi-1001101-g004){ref-type="fig"} the lack of correlation between the miRNA and mRNA trajectories in the open circuit ([Figure 4B](#pcbi-1001101-g004){ref-type="fig"}) leads to much larger deviations from the mean number of proteins with respect to the FFL case. Using the same parameter values of [Figure 3](#pcbi-1001101-g003){ref-type="fig"}, the predicted for the open circuit is , to be compared with the value of the FFL. Different cell-to-cell variability can be clearly seen comparing the distributions of the number of target proteins for the two circuits ([Figure 4C](#pcbi-1001101-g004){ref-type="fig"}). Note that a target embedded in an open circuit has an even more noisy expression than a gene simply regulated by a TF, for which . ::: {#pcbi-1001101-g004 .fig} 10.1371/journal.pcbi.1001101.g004 Figure 4 ::: {.caption} ###### Noise properties of the FFL compared with an analogous open circuit. \(A) An example of simulation results for the FFL (scheme on the right or more detailed in [Figure 2A](#pcbi-1001101-g002){ref-type="fig"}′). The parameter values are the same of [Figure 3](#pcbi-1001101-g003){ref-type="fig"}. (B) Time evolution in a simulation for the molecular players in the open circuit (scheme on the right or more detailed in [Figure 2C](#pcbi-1001101-g002){ref-type="fig"}′). The correlation between the and trajectories that is present in the FFL (A) is completely lost in the open circuit. As a consequence while the mean value of is approximately the same, its fluctuations are appreciably greater in the open circuit case. (C) The probability distribution of protein number for the two circuits. Histograms are the result of Gillespie simulations while continuous lines are empirical distributions (gaussian for the FFL and gamma for the open circuit) with mean and variance predicted by the analytical model. :::  ::: ### Deviant effects {#s2c1} Stochastic equations are the natural formalism to describe a set of biochemical reactions when the number of molecules involved is small and thus fluctuations are important. As the number of molecules increases, the stochastic description smoothly converges, at least for linear systems, toward a deterministic one and stochastic equations can be substituited by ordinary differential equations (ODE). It is usually expected that even in the regime in which fluctuations cannot be neglected one could use ODE if interested only in the time evolution of the mean values. This approximation can be thought as a sort of "mean field" approach (by analogy with statistical mechanics where the mean field approximation is implemented by neglecting fluctuations). However, similarly to what happens in statistical mechanics in the proximity of a critical point, it may happen that, even at the level of mean values, the ODE description does not coincide with the (more rigorous) stochastic one. These breakdowns between the two descriptions are known as "deviant effects" [@pcbi.1001101-Samoilov1] and are typically a consequence of nonlinear terms in the equations or of strong extrinsic fluctuations [@pcbi.1001101-Shahrezaei2], [@pcbi.1001101-Shahrezaei3]. In some cases these deviant effects can have relevant phenomenological consequences. This is the case of our system: although the FFL ([Figure 2A,A′](#pcbi-1001101-g002){ref-type="fig"}) and the open circuit ([Figure 2C,C′](#pcbi-1001101-g002){ref-type="fig"}) have the same deterministic description at the steady state (see [Text S1](#pcbi.1001101.s001){ref-type="supplementary-material"} for details), the master equation approach gives a mean value of the target protein systematically lower in the FFL circuit, with respect to the same quantity in the open circuit. This is a non trivial consequence of the correlated fluctuations in the number of mRNAs and miRNAs in the FFL. This correlation between fluctuations obviously cannot be taken into account in the deterministic description and as a consequence the ODE analysis correctly describes the steady state mean number of target proteins only for the open circuit. This result can be understood by looking at the analytical expression of the mean value of : In a FFL, fluctuations of and are highly correlated ([Figure 3A](#pcbi-1001101-g003){ref-type="fig"}), because the transcription rates of messengers and miRNAs depend on a shared TF. The result is that in this case . On the other hand, the production of and is independently regulated in an open circuit, implying that . A deterministic description does not take into account fluctuations so correctly describes only when uncorrelated noise is averaged out without affecting mean values. In conclusion, the correlation in fluctuations introduced by the FFL topology affects the target protein mean value, establishing a systematic decrease with respect to the deterministic description. This deviant effect can be substantial and underlines the necessity of a stochastic nonlinear modeling. A fully linearized description, as for example the one proposed by [@pcbi.1001101-Komorowski1] for post-transcriptional regulation, would not be able to show this type of effects. The incoherent feedforward loop is effective in reducing extrinsic fluctuations {#s2d} ------------------------------------------------------------------------------- In the previous sections we compared different regulatory circuits keeping the same amount of input noise, i.e. the same amount of fluctuations in the copy number of master TFs. Since the topology of a regulatory motif can have stronger effects on extrinsic rather than intrinsic noise [@pcbi.1001101-Shahrezaei2], it would be very interesting to study how the mixed incoherent FFL behaves as a function of such extrinsic noise. As a matter of fact extrinsic and intrinsic fluctuations are generally coupled in a non-trivial way in biochemical networks [@pcbi.1001101-TanaseNicola1], but we developed a strategy to control fluctuations in upstream TF expression, known to be one of the major sources of extrinsic noise in eukaryotes [@pcbi.1001101-Volfson1], without affecting the copy number of the molecular species in the circuit. From equation 6 we can calculate (which denotes the mean number of TFs) and its noise strength :where, as above, the parameters and denote the rate of transcription and translation of the TF respectively, and and the corresponding degradation constants. Setting the rates of degradation ( and ) and the product to constant values, we end up with: and . This is a well known result: fluctuations in the protein level are stronger when the rate of translation is higher [@pcbi.1001101-Raj1] and can be tuned (while keeping the mean value fixed) by changing the ratio with . This represents a perfect theoretical setting to test the dependence of the target noise on the input noise , while maintaining unchanged the mean value of all the molecular species involved in the circuit. We report in [Figure 5](#pcbi-1001101-g005){ref-type="fig"} the results of such analysis for the three circuits discussed in the previous sections. While extrinsic fluctuations increase, so does the FFL\'s performance in filtering out noise, compared to the other circuits. Once again it is easy to understand the reason of this behaviour: the FFL architecture channels fluctuations of an upstream factor into parameters with opposite effect on the target protein, forcing them to combine destructively. Therefore the specific FFL topology seems effective in the maintenance of gene expression robustness despite noisy upstream regulators. The introduction of a miRNA pathway, building a FFL from a TF-gene cascade, really makes the difference in situations of strong upstream noise. This feature can explain why miRNAs can often be deleted without observable consequences [@pcbi.1001101-Wu1], since experiments usually do not measure fluctuations and are typically performed in controlled environments, where noise is kept at minimal levels. ::: {#pcbi-1001101-g005 .fig} 10.1371/journal.pcbi.1001101.g005 Figure 5 ::: {.caption} ###### The effect of fluctuations in an upstream TF. We maintain constant the number of TFs , while we vary its relative fluctuations , tuning the relative contribution of transcription (rate ) and translation (rate ). All the other parameters have the values reported in caption of [Figure 3](#pcbi-1001101-g003){ref-type="fig"}. The incoherent FFL makes the target less sensitive to fluctuations in the upstream TF. The extent of the noise reduction, with respect to the other circuits, depends on the magnitude of the TF noise, pointing out that the FFL topology is particularly effective in filtering out extrinsic fluctuations. Dots are the result of Gillespie simulations with the full nonlinear dynamics while continuous lines are analytical predictions. :::  ::: Noise filtering optimization {#s2e} ---------------------------- The efficiency of the FFL in controlling the fluctuations of the target protein is a function of three main parameters: the number of master TFs (which in turn is a function of and ), the number of miRNA copies (function of and ) and the strength of miRNA repression (defined as ). In this section we shall study the efficiency of the FFL in buffering noise as a function of each one of these three quantities, changing a corresponding parameter while keeping fixed all others. As we shall see, in all three cases the noise reduction with respect to a simple TF-target interaction (i.e. without the miRNA) shows a U-shaped profile with a well defined minimum which allows us to identify the values of the parameters which optimize the noise reduction property of the FFL. This pattern is rather robust, and only marginally depends on the way the variable of interest is tuned (for instance, by changing or in the case of miRNA concentration). It is important to stress that in all three cases optimal noise filtering does not imply strong repression, a result which well agrees with the observation that miRNAs embedded in an incoherent FFL usually have a fine-tuning effect on the targets instead of switching them off completely. It is exactly in the intermediate region of the parameters, in which fine-tuning occurs, that we also have optimal noise reduction. ### Optimal repression strength {#s2e1} The repression strength is defined as (inverse of the dissociation constant in the Hill function of equation 2). As it can be seen in [Figure 6A](#pcbi-1001101-g006){ref-type="fig"}, the FFL exhibits a noise profile with a typical U-shape and reaches an optimal value of noise reduction (measured as the difference in the noise strength with respect to the simple TF-gene circuit) for intermediate values of repression strength. The open circuit, constrained to give the same mean value , always introduces larger target fluctuations. As mentioned above, optimal noise filtering is reached for intermediate values of the repression strength and does not require strong target repression. For instance with the choice of parameter values of [Figure 6](#pcbi-1001101-g006){ref-type="fig"}, optimal noise reduction is obtained for a mean value of the target protein which is about of the value obtained without the miRNA, i.e. with a simple TF-target interaction. This prediction could be experimentally tested via manipulation of the repression strength, in analogy to the work of [@pcbi.1001101-Dublanche1] on the transcriptional autoregulatory motif. It is instructive to notice the analogies between the behaviour of the mixed FFL and that of the negative transcriptional autoregulation loop. This purely transcriptional network motif occurs ubiquitously in transcriptional regulatory networks and was recently studied in great detail [@pcbi.1001101-Shahrezaei2], [@pcbi.1001101-Singh1]. Also in this case, optimal noise filtering is obtained for a well defined value of the repression stength. It is easy to understand the reason of this behaviour. In a negative transcriptional autoregulation, the protein expressed from a gene inhibits its own transcription by increasing expression when protein numbers are low, while decreasing expression when protein numbers are high. Increasing the repression strength improves the circuit potential to reduce stochasticity, but at the same time decreases the copy number of mRNAs and proteins, with a rise in intrinsic fluctuations that can overcome any attenuation. Consistently, experiments show a well defined range of repression strength for which noise minimization is optimal [@pcbi.1001101-Dublanche1]. ::: {#pcbi-1001101-g006 .fig} 10.1371/journal.pcbi.1001101.g006 Figure 6 ::: {.caption} ###### How an optimal noise filter can be built. \(A) The coefficient of variation of the target protein as a function of the repression strength . The Figure shows the presence of an optimal repression strength for which the introduction of a miRNA regulation in a FFL minimizes noise. (B) as a function of the mean number of miRNAs . In this case is changed through the maximum rate of transcription (see equation 1). (C) as a function of , varying the dissociation constant . In both cases (B and C) is evident a U-shaped profile, implying an optimal noise buffering for intermediate miRNA concentrations. (D) as a function of the mean number of TFs . The number of TFs depends on the rate of their transcription and of their translation . The Figure is obtained manipulating , but the alternative choice of leads to equivalent results (see [Text S1](#pcbi.1001101.s001){ref-type="supplementary-material"}). For intermediate concentration of the TF, the noise control by the FFL outperforms the one of the other circuits. In each plot, dots are the result of Gillespie simulations while continuous lines are analytical predictions. The values of parameters kept constant are the same of [Figure 3](#pcbi-1001101-g003){ref-type="fig"}, however the results are quite robust with respect to their choice (see [Text S1](#pcbi.1001101.s001){ref-type="supplementary-material"} for details). :::  ::: ### Optimal miRNA concentration {#s2e2} Another variable which can be tuned in order to achieve optimal noise reduction is the miRNA concentration. If we keep the number of TFs constant then the miRNA concentration can only depend on the maximum rate of transcription of the miRNA gene () and on the affinity of its promoter to the TF (proportional to , where is the dissociation constant in equation 1). In [Figures 6B and 6C](#pcbi-1001101-g006){ref-type="fig"} we analyze the noise strength of the target protein in the FFL for different miRNA concentrations and compare it to the obtained with the simple TF-gene interaction and with the open circuit. Changing the miRNA concentration by varying ([Figure 6B](#pcbi-1001101-g006){ref-type="fig"}) or ([Figure 6C](#pcbi-1001101-g006){ref-type="fig"}) we find very similar U-shaped profiles for . As for the previous analysis, also in this case it is possible to find an optimal miRNA concentration, and again it is such that the effect on the protein target is only a modest reduction (in this case of the value obtained without the miRNA). Apart from the conserved U-shaped profile, there are rather deep differences in the noise behaviour depending on the choice of the tuning parameter. In fact, while an increase of always induces an increase of , this quantity becomes insensitive to above a certain threshold. Since the number of TFs is constant in this analysis, it is clear that increasing ([Figure 6C](#pcbi-1001101-g006){ref-type="fig"}) the system can reach at best the value of consistent with the maximum rate of transcription. At the same time a large value of means that very few TFs are enough to support the maximum transcription rate for the miRNA gene, so fluctuations in the number of TFs become irrelevant despite the topology of the circuit. As a consequence the curves for the FFL and the open circuit converge to a commom value ([Figure 6C](#pcbi-1001101-g006){ref-type="fig"}). A refined experimental control of miRNA concentration through graded miRNA overexpression or silencing would permit a test of the U-shaped profile of in a FFL. ### Optimal TF concentration {#s2e3} The last case that we consider in this section is the effect of different TF concentrations on the noise buffering properties of the FFL. It is expected that for high TF concentrations (i.e. high values of ) the activation functions in equations 1 reach the saturation point, making the system insensitive to variations in TF concentration. As long as the number of TFs does not fluctuates below the saturation point, the miRNA and the target gene are maximally transcribed, with no reference to the exact number of TFs. Accordingly, becomes asymptotically constant for large for each circuit topology ([Figure 6D](#pcbi-1001101-g006){ref-type="fig"}). The gap between the asymptotic values of the direct TF regulation and the two other circuits is due to the fact that the former does not suffer for the additional external noise due to the fluctuations in the miRNA number. On the other hand, for small values of also the number of target proteins is very small as its expression is hardly activated regardless of the circuit type, with a consequent increase of the noise strength ([Figure 6D](#pcbi-1001101-g006){ref-type="fig"}). The central region is the most interesting one: this is the region in which the system is maximally sensitive to changes in TF concentration. In this regime the FFL outperforms both the simple direct regulation and the open circuit in buffering noise. Also in this case the optimal TF concentration is placed in a region corresponding to a modest reduction of , despite the miRNA repression. ### Exploring the parameter space {#s2e4} To give a more comprehensive insight into the relation between noise control and target repression, we finally evaluate the buffering of fluctuations () for each average number of TFs and each degree of target suppression (), where and represent here the constitutive mean expression and fluctuations in absence of miRNA regulation. Results of this analysis are reported in [Figure 7A](#pcbi-1001101-g007){ref-type="fig"}. As discussed above, noise reduction can be implemented successfully in the parameter region where the target activation function (in [Figure 7B](#pcbi-1001101-g007){ref-type="fig"}) is not saturated, since this is the region where the target is sensitive to changes in TF concentration and therefore also to its fluctuations, regardless of the presence or absence of miRNA regulation. It is exactly in this region that noise buffering can be observed (compare [Figures 7A and B](#pcbi-1001101-g007){ref-type="fig"}). In particular, for each TF concentration the best noise reduction appears for a target level around 60% of its constitutive expression. In the optimal setting, noise can be remarkably reduced to about one half of its constitutive value, but the reduction remains substantial also for weaker repressions, further confirming that a strong miRNA repression is not required for noise control. ::: {#pcbi-1001101-g007 .fig} 10.1371/journal.pcbi.1001101.g007 Figure 7 ::: {.caption} ###### Exploring the parameter space. \(A) The target noise , achieved with the FFL, is evaluated with respect to noise deriving from constitutive expression (i.e. in absence of miRNA regulation) for different mean levels of the TF and different degrees of reduction of the target protein level (where is the mean constitutive expression). The TF level is changed through its rate of translation (equivalent results can be obtained changing the rate of transcription ), while the target level is tuned varying the repression strength. All the other parameters have the values reported in caption of [Figure 3](#pcbi-1001101-g003){ref-type="fig"} except (lower than in [Figure 3](#pcbi-1001101-g003){ref-type="fig"} to explore a more noisy situation). The region where miRNA repression leads to larger fluctuations with respect to constitutive ones is shown in white. When a noise reduction is gained the value of is reported with the color code explained in the legend. The best noise control is achieved with a modest suppression of target expression, around the 60% of its constitutive value. (B) The rate of transcription of the target mRNA as a function of the mean number of TFs. The noise reduction shown in the above plot can be obtained outside the saturation regime (where the slope of the activation curve tends to zero). :::  ::: We consider the characterization of the optimal setting of miRNA-mediated incoherent FFLs for noise buffering, and the resulting U-shaped profile predicted for the noise reduction factor, as one of the major results of our analysis which, we expect, should be amenable of direct experimental validation. The fact that an optimal noise filtering is obtained with a rather modest reduction in the amount of the target protein also agrees with the recent experimental observation that miRNA down-regulation of targets is often modest [@pcbi.1001101-Baek1], [@pcbi.1001101-Selbach1] and apparently dispensable from a functional point of view. In this respect it is tempting to conjecture that, at least for some targets of incoherent FFLs, the down-regulation could only be the side effect of an evolutionary design aiming instead to optimize noise reduction. Comparison with purely transcriptional incoherent feedforward loops {#s2f} ------------------------------------------------------------------- The capability of incoherent FFLs to reduce fluctuations was previously studied with simulations in the contest of purely transcriptional networks [@pcbi.1001101-Shahrezaei2]. In this section we present a comparison of the noise properties of microRNA-mediated FFLs (scheme in [Figure 1A](#pcbi-1001101-g001){ref-type="fig"}′) and purely transcriptional ones (detailed scheme of reactions in [Figure 8A](#pcbi-1001101-g008){ref-type="fig"}), where the miRNA is replaced by a protein that inhibits transcription (rather than translation, as miRNAs do). The transcriptional FFL can be modeled with the same strategy explained previously for the miRNA-mediated version and analogously mean values and standard deviations of the various molecular species can be calculated analytically with the moment generating function method (see [Text S1](#pcbi.1001101.s001){ref-type="supplementary-material"} for more details on calculations and model assumptions). In order to make an unbiased comparison of the noise properties of these two circuits, the corresponding models must be constrained to produce the same amount of target proteins. Although there is no unambiguous way to put this constraint, due to the presence of more free parameters ( and ) in the purely transcriptional case, a reasonable choice is to keep the shared parameters to same values (i.e repression/activation efficiencies and production/degradation rates) and choose the two additional ones to make the amount of repressor proteins in the transcriptional case equal to the amount of miRNAs in the mixed circuit. With this choice we can evaluate the target noise as a function of the repression strength () for both circuits ([Figure 8B](#pcbi-1001101-g008){ref-type="fig"}). Even though the transcriptional version can potentially reduce target fluctutions, buffering efficiency appears clearly increased by the use of miRNAs as regulators. Furthermore, a comparison of [Figure 8C](#pcbi-1001101-g008){ref-type="fig"} and [Figure 7B](#pcbi-1001101-g007){ref-type="fig"} points out that a miRNA-mediated FFL can buffer fluctuations over a wider range of conditions as well as to a greater extent. This is mainly due to the additional step of translation required for the production of proteins which unavoidably adds noise to the system. We would like to emphasize that the shown efficiency in noise reduction, achieved with the transcriptional FFL, should be considered as an upper bound. In fact, the constraints imposed on and limit the translational burst size, i.e. the average number of proteins produced from a single mRNA, and this parameter crucially influences the intrinsic fluctuation amplitude of proteins [@pcbi.1001101-Kaern1] (see [Text S1](#pcbi.1001101.s001){ref-type="supplementary-material"} for details on parameter constraints). With the parameter values used in [Figure 8](#pcbi-1001101-g008){ref-type="fig"}, the translational burst size is , while in eukaryotes it is expected to be larger (certainly larger than one) because of the long average half-life of messenger RNAs compared to the time required for one translation round [@pcbi.1001101-Larson1]. Therefore the noise added by the step of translation of proteins should realistically be more substantial than reported for this model setting, with harmful consequences on the noise buffering efficiency of the purely transcriptional circuit. ::: {#pcbi-1001101-g008 .fig} 10.1371/journal.pcbi.1001101.g008 Figure 8 ::: {.caption} ###### Comparison with a purely transcriptional incoherent FFL. \(A) Detailed scheme of a purely transcriptional incoherent FFL. (B) The coefficient of variation of the target protein as a function of the repression strength for a miRNA-mediated FFL and for its transcriptional counterpart. Thanks to the constraints imposed on parameters we can directly compare their noise-buffering efficiency with respect to a gene only activated by a TF. Both circuitries lead to a curve with a minimum for an intermediate repression strength, but the miRNA-mediated circuit appears more efficient in filtering out fluctuations. The values of parameters kept constant are the same of [Figure 3](#pcbi-1001101-g003){ref-type="fig"}. Dots are the result of Gillespie simulations with the full nonlinear dynamics while continuous lines are analytical predictions. Also in this case, analytical solutions fit nicely with simulation results. (C) The noise reduction , achieved with a purely transcriptional incoherent FFL, evaluated for different mean levels of the TF and different degrees of reduction of the target protein level . The parameter values and the color code are the same of [Figure 7](#pcbi-1001101-g007){ref-type="fig"} so as to allow a direct comparison. :::  ::: Moreover some peculiarities (not currently included in our model) of the mixed-motif can further explain why it can be better suited for noise buffering. Firstly, fluctuations in RNA polymerase and ribosome abundance are possible sources of extrinsic noise in gene expression [@pcbi.1001101-Swain1]. These fluctuations are expected to influence unspecifically the rate of transcription and translation respectively of each gene. In a miRNA-mediated FFL the correlation between target mRNA and miRNA production, which is crucial for noise reduction, is robust to these additional sources of noise as mRNAs and miRNAs are identically affected only by global RNA polymerase fluctuations. On the other hand, in purely transcriptional FFLs the number of repressor proteins is exposed to the independent fluctuations in ribosome concentration, so the regulator-regulated correlation can be compromised with potentially negative consequences on the circuit\'s noise reduction efficiency. Secondly, delays in the action of regulators (miRNA or proteins) in the indirect pathway from the master TF to the target can damage the noise buffering function (see [Text S1](#pcbi.1001101.s001){ref-type="supplementary-material"} for a more detailed study of the impact of time delays on noise control). However, the biogenesis of miRNAs is thougth to be faster than the one of proteins, and thus miRNAs may affect the target expression with a shorter delay with respect to factors regulating nuclear events like a TF [@pcbi.1001101-Li1]. This feature should enable miRNAs to produce rapid responses, as required to counteract fluctuations. Finally, the presence of a nucleus makes the eukaryotic cell a two-compartment system with stochastic transport channels, with consequences on gene expression noise [@pcbi.1001101-Xiong1]. In fact, transcriptional regulation requires an additional transport step with respect to the post-transcriptional one. In a transcriptional FFL, the repressor protein (replacing the miRNA) must return into the nucleus to act on the target. This again potentially reduces the correlation of its fluctuations with the target ones, affecting the noise buffering ability. Cross-talk between microRNA targets {#s2g} ----------------------------------- A recent study pointed out that the action of a miRNA on a specific target gene expression is affected by the total number of miRNA targets and their mRNA abundance [@pcbi.1001101-Arvey1], a phenomenon called "dilution effect". This is presumably a consequence of target competition for a finite intracellular pool of miRNAs. In particular, the degree of downregulation of an individual target expression is generally reduced by the presence of other transcribed target genes. A similar cross-talk between targets has been previously shown for sRNA regulation in bacteria [@pcbi.1001101-Levine1] both theoretically and experimentally. Therefore, the functionality of a genetic circuit that involves miRNA regulations, as the one studied in this paper, can be influenced by the expression level of miRNA targets not embedded in the circuit. To address this issue we evaluate in this section the impact of an additional miRNA target independently transcribed (a situation depicted in [Figure 9A](#pcbi-1001101-g009){ref-type="fig"}) on the circuit ability in noise buffering. ::: {#pcbi-1001101-g009 .fig} 10.1371/journal.pcbi.1001101.g009 Figure 9 ::: {.caption} ###### Effects of cross-talk between miRNA targets. \(A) Scheme of a miRNA-mediated FFL with an additional independently transcribed target gene (second target). (B) The degree of protein downregulation is depicted as a function of the ratio of effective transcription rates of the secondary target () and of the FFL joint target (), for different values of . Since the rate of transcription of the joint target is a function of the TF concentration, we consider for this analysis the effective mean rate as a reference (where is constant as we are not tuning the TF concentration). The transcription of the second target is modeled as an independent birth-death process with birth rate . In this plot the coupling constants between targets and miRNAs are assumed equal () and for each value the coupling constant is chosen so as to start with the same amount of target proteins () in absence of secondary targets (the complete set of parameters values is presented in [Text S1](#pcbi.1001101.s001){ref-type="supplementary-material"}). In the limit of infinite out-of-circuit target expression, the joint target protein level approaches its constitutive value if , while remains constant in the ideal case of perfectly catalytic miRNA repression (red curve). Continuous lines are analytical solutions of the deterministic model (Equations 9), while dots are the result of stochastic simulations. (C) With the parameter setting of [Figure 9B](#pcbi-1001101-g009){ref-type="fig"}, the noise reduction is evaluated in the same range. Dots are the result of Gillespie simulations while continuous lines come from trivial interpolations. (D) The noise reduction is evaluated as a function of the out-of-circuit mRNA fluctuations , relative to the joint target fluctuations . The fluctuations of the second target are modulated considering its rate of transcription as a function of an independent TF and changing the TF noise with the same strategy used for [Figure 5](#pcbi-1001101-g005){ref-type="fig"} (see [Text S1](#pcbi.1001101.s001){ref-type="supplementary-material"} for more details). The concentrations of the TFs activating the two targets are constrained to be equal so as to study the situation of two independent targets with the same effective transcription rate. Dots are the result of Gillespie simulations, simply interpolated with continuous lines. :::  ::: ### Stoichiometric versus catalytic models of miRNA action {#s2g1} The model used so far for miRNA regulation was based on the hypothesis of perfectly catalytic action. The rate of translation of target mRNAs was assumed to be a nonlinear decreasing function of miRNA concentration, neglecting the details of mRNA-miRNA physical coupling with the implicit assumption that the downregulation process does not affect the available miRNA pool. A perfectly catalytic action does not predict any competition effect among multiple targets at equilibrium, since each target can only sense the available number of miRNAs without altering it. On the other hand, a stoichiometric model has been proposed in the context of sRNA regulation in bacteria [@pcbi.1001101-Levine1]--[@pcbi.1001101-Shimoni1], in which each sRNA can pair with one messenger leading to mutual degradation. In this latter case the expression of a secondary target can capture a significant portion of the sRNAs, with a resulting decrease in the average repression acting on the first target. The nature of miRNA regulation is presumably somewhere in between these two extreme possibilities, although usually generically referred to as catalytic. In this view, in order to address the effect of target cross-talk on miRNA-mediated FFLs, we consider a deterministic model (introduced previously in [@pcbi.1001101-Levine1]) that explicitely takes into account the physical coupling of miRNAs and target mRNAs and the catalytic/stoichiometric nature of this coupling. While the full detailed model is presented in [Text S1](#pcbi.1001101.s001){ref-type="supplementary-material"}, the effective equations describing the dynamics of the mean number of miRNAs , mRNAs of the target in the FFL and mRNAs of the secondary miRNA target are:where and describe the probability of miRNA-mRNA coupling for the target in the FFL and the secondary target respectively, while is the probability (assumed equal for both targets) that a degradation event of a mRNA, induced by a miRNA, is accompained by the degradation of the miRNA itself. The limit describes a stoichiometric mode of action, while the opposite situation of represents a perfectly catalytic mode in which the rate of mRNA degradation is a linear function of the number of miRNAs. The corresponding stochastic model, of which equations 9 describe the mean-field limit, cannot be solved analytically starting from the master equation, therefore noise properties will be examined in the following with simulations only. ### Dilution effect {#s2g2} In the first place we evaluate the dependence of the target protein downregulation on the expression rate of the secondary target, starting from the model described by Equations 9. The dilution effect is shown in [Figure 9B](#pcbi-1001101-g009){ref-type="fig"} for different values of : the downregulation exerted on the FFL target depends on the rate of expression of the secondary target, in line with the observed inverse correlation between target abundance and mean downregulation in higher eukaryotes [@pcbi.1001101-Arvey1] and in bacteria [@pcbi.1001101-Levine1]. Similar results can be obtained by varying the coupling constant with respect to (as reported in [@pcbi.1001101-Levine1]). Therefore, the noise buffering function and the optimality criteria discussed in previous sections could be compromised in the presence of many or highly transcribed independent miRNA targets. This issue will be addressed in details in the following section. As expected, a perfectly catalytic mode does not feel the effect of secondary mRNA targets (red line in [Figure 9B](#pcbi-1001101-g009){ref-type="fig"}), while the stoichiometric mechanism is the most sensitive (green line in [Figure 9B](#pcbi-1001101-g009){ref-type="fig"}). This result suggests that a catalytic mode (at least approximately), like the miRNA one, can allow a larger proliferation of the number of targets while limiting the effects of their cross-talk. ### Consequences of dilution effect and secondary target fluctuations on noise buffering {#s2g3} Since a high level of expression of secondary targets can determine a decrease of the average downregulation, it can potentially reduce the FFL ability in filtering out target fluctuations. In fact, also the noise reduction (where is the constitutive noise in absence of miRNA) is a function of the additional target expression, as shown in [Figure 9C](#pcbi-1001101-g009){ref-type="fig"}. As the expression of the out-of-circuit target increases, its messengers are able to capture more and more miRNAs and the efficiency in noise reduction is gradually compromised. Finally the FFL target fluctuations approach the constitutive ones when the messengers of the FFL target become a small fraction of the total miRNA targets. The robustness of the circuit functioning with respect to the dilution effect is again dependent on the repression mode (that changes with ). Moreover, as discussed in [Text S1](#pcbi.1001101.s001){ref-type="supplementary-material"}, different modes (stoichiometric/catalytic) of miRNA action have a different potential in reducing fluctuations: even in absence of secondary targets, where models with different have been constrained to produce the same amount of target protein, the noise buffering efficiency decreases with ([Figure 9C](#pcbi-1001101-g009){ref-type="fig"}). This observation highlights that the level of miRNA ability to avoid mutual degradation while targeting a mRNA can play a role in the optimization of fluctuation counteracting, besides conferring stability with respect to target cross-talk. While the corruption of the noise-buffering ability seems mainly due to the increase in the mean level of secondary messengers, there is another more subtle cause that gives a contribution: the uncorrelated fluctuations of secondary messengers. Since the secondary target is independently transcribed (not under the control of the master TF activating the miRNA gene) its fluctuations are expected to be completely uncorrelated with the miRNA ones, implying a random sequestration of miRNAs. To disentagle this contribution from the dilution effect, we studied the case of a secondary target transcribed at the same effective rate of the FFL target, but with different levels of fluctuations (see [Figure 9D](#pcbi-1001101-g009){ref-type="fig"}). In the case of equal transcription rates the dilution effect has a negligible impact on the noise buffering activity of the circuit (see [Figure 9C](#pcbi-1001101-g009){ref-type="fig"}), nevertheless the level of noise reduction () is progressively reduced as the second target concentration becomes more and more noisy, as reported in [Figure 9D](#pcbi-1001101-g009){ref-type="fig"}. This effect seems especially relevant for a hypothetically stoichiometric miRNA repression. Therefore, the noise level of additional targets is a variable that must be taken into account in evaluating the cross-talk effect on the noise-buffering efficiency of the circuit. Although the FFLs are overrepresented in the mixed network [@pcbi.1001101-Re1]--[@pcbi.1001101-Yu1], a single microRNAs can downregulate hundreds of target genes and consequently not every target is expected to be under the control of the same TF regulating the miRNA gene (see [Text S1](#pcbi.1001101.s001){ref-type="supplementary-material"} for a more detailed discussion). Therefore, even though most motif function analysis are carried out looking at the motif operating in isolation, we have shown that the presence of additional miRNA targets in the network can alter the functioning of a miRNA-mediated motif. In fact, the efficiency of miRNA-mediated FFLs as noise controllers should be considered contest-dependent. While this circuit seems properly designed to filter out fluctuations when the miRNA-target interaction is specific or secondary targets are poorly transcribed, cell types or conditions that require a high expression of out-of-circuit miRNA targets can significantly corrupt this circuit property. Besides the understanding of the function of endogenous miRNA-mediated FFLs, this analysis of target cross-talk effects can be a useful warning for the growing field of synthetic biology [@pcbi.1001101-Mukherji1]: the implementation of genetic circuits incorporating small RNA regulations for specific scopes must take into account the sRNA specificity and the level of expression (and fluctuations) of eventual other targets. Discussion {#s3} ========== Experimental and bioinformatic evidences of the relevance of miRNA mediated FFLs in gene regulation {#s3a} --------------------------------------------------------------------------------------------------- Few cases of incoherent miRNA-mediated FFLs have been experimentally verified until now: a case involving c-Myc/E2F1 regulation [@pcbi.1001101-ODonnell1] and more recently a miR-7 mediated FFL in *Drosophila* [@pcbi.1001101-Li1]. As a matter of fact, miR-7 has indeed been found to be essential to buffer external fluctuations, providing robustness to the eye developmental program. The fact that miR-7 is interlocked in an incoherent FFL provides a first hint that our model can be biologically relevant. On the purely computational side, it is interesting to notice that in [@pcbi.1001101-Re1] it was shown that the typical targets of these FFLs are not randomly distributed but are instead remarkably enriched in TFs. These are the typical genes for which a control of stochastic fluctuations should be expected: the noise in a regulator expression propagates to all its targets, affecting the reliability of signal transmission in the downstream network. Finally, a significant enrichment in oncogenes within the components of the FFLs was also observed [@pcbi.1001101-Re1]. The mentioned FFL containing c-Myc/E2F1 is just an example [@pcbi.1001101-EsquelaKerscher1]. In view of the emerging idea that non-genetic heterogenetity, due to stochastic noise, contributes to tumor progression [@pcbi.1001101-Brock1] and affects apoptotic signal response [@pcbi.1001101-Spencer1], the role of miRNA-mediated FFLs in reducing fluctuations can explain why they are often involved in cancer-related pathways. Concluding remarks {#s3b} ------------------ The type of regulatory action which a miRNA exerts on its targets can be rather well understood looking at the degree of coexpression with the targets [@pcbi.1001101-Flynt1], [@pcbi.1001101-Bartel1], [@pcbi.1001101-Bushati1], [@pcbi.1001101-Hornstein1], [@pcbi.1001101-Bartel2]. In particular, an incoherent mixed-FFL implies a high level of miRNA-target coexpression, so it is suitable to implement a fine-tuning interaction. The target is not switched off by miRNA repression, rather its mean level is adjusted post-transcriptionally to the desired value. However, many cells can have a protein concentration far from the finely controlled mean value, if strong fluctuations are allowed. Hence, a noise buffering mechanism can be crucial at the level of single cells, and a fine-tuning interaction will be effective for a large part of the cell population only if coupled with a noise control. Some authors proposed the conjecture that the incoherent mixed-FFL can actually have a role in noise buffering [@pcbi.1001101-Tsang1], [@pcbi.1001101-Hornstein1], [@pcbi.1001101-Wu1] and biological evidences that miRNAs can effectively be used as expression-buffers have been recently found [@pcbi.1001101-Wu1], [@pcbi.1001101-Li1]. From this point of view the miRNA-target interactions classified as neutral [@pcbi.1001101-Bartel2], as the mean level of the target only changes inside its functional range by the presence/absence of miRNAs, actually could have been selected by evolution to prevent potentially harmful fluctuations. In this paper we demonstrated, through stochastic modeling and simulations, that the incoherent mixed-FFL has the right characteristics to reduce fluctuations, giving a proof to the previously proposed intuitive conjecture and supplying the lacking quantitative description. In particular, we showed that this circuit filters out the noise that is propagating from the master TF, giving robustness to the target gene expression in presence of noisy upstream factors. Furthermore, our theoretical description led to the prediction that there is a value of the miRNA repression strength for which the noise filtering is optimal. A maximum of target-noise attenuation appears likewise varying the miRNA concentration or the TF concentration and this robust prediction could be tested experimentally. In all cases the implementation of the best noise filter does not imply a strong suppression of the target protein expression, coherently with a fine-tuning function and in agreement with the observation that the miRNA down-regulation of a target is often modest [@pcbi.1001101-Baek1], [@pcbi.1001101-Selbach1]. Our paper presents the first model explicitly built on the mixed version of the FFL. From a theoretical point of view, we addressed the detailed master equation describing the system (without neglecting the dynamics of mRNA), instead of the approximate Langevin description, and we were able to apply the moment generating function approach despite the presence of nonlinear terms that can give rise to deviant effects. This approach allowed us to take into account extrinsic fluctuations as the noise propagating from upstream genes, without an arbitrary definition of the extrinsic noise distribution. This strategy can be naturally extended to other circuits in the mixed network to test their potential role in the control of stochasticity. Furthermore, we compared, in terms of noise buffering ability, miRNA-mediated FFLs with their purely transcriptional counterparts, where the miRNA is replaced by a protein that inhibits transcription rather than translation. This comparison shows that a miRNA regulator can be better suited for the noise buffering purpose. Finally, we tryed to overcome the limitations in the analysis that can arise from considering a genetic circuit as operating in isolation. In this perspective, we evaluated the impact that the recently discovered dilution effect [@pcbi.1001101-Levine1], [@pcbi.1001101-Arvey1] can have on the noise buffering function of miRNA-mediated incoherent FFLs. More specifically, we showed than an efficient noise control requires the minimization of the number of miRNA target sites on out-of-circuit genes, especially if highly expressed or strongly fluctuating in the mRNA level. The hypothesis of a role of miRNAs in noise buffering can shed new light on peculiar characteristics of miRNA regulation. As discussed in [@pcbi.1001101-Wu1] and [@pcbi.1001101-Li1], it can explain why miRNAs are often highly conserved, controlling key steps in development, but in many cases they can be deleted with little phenotypic consequences. On the evolutionary side, the origin of vertebrate complexity seems to correspond to the huge expansion of non-coding RNA inventory (including miRNAs) [@pcbi.1001101-Heimberg1]. This can suggest a further reasoning: the morphological complexity requires a high degree of signaling precision, with a strict control of stochasticity, and miRNA regulation can satisfy these requirements if embedded in an appropriate circuit, as we showed for the ubiquitous miRNA-mediated FFL. Methods {#s4} ======= Simulations were implemented by using Gillespie\'s first reaction algorithm [@pcbi.1001101-Gillespie1]. The reactions simulated were those presented in schemes 2A′,B′,C′ and 8A. Reactions that depend on a regulator were allowed to have as rates the corresponding full nonlinear Hill functions. All the results are at steady state, which is assumed to be reached when the deterministic evolution of the system in analysis is at a distance from the steady state (its asymptotic value) smaller than its 0.05% (see [Text S1](#pcbi.1001101.s001){ref-type="supplementary-material"} for details). For the parameter set used for [Figures 3](#pcbi-1001101-g003){ref-type="fig"}-[](#pcbi-1001101-g004){ref-type="fig"} [](#pcbi-1001101-g005){ref-type="fig"} [](#pcbi-1001101-g006){ref-type="fig"} [](#pcbi-1001101-g007){ref-type="fig"} [](#pcbi-1001101-g008){ref-type="fig"} [9](#pcbi-1001101-g009){ref-type="fig"} the steady state was assumed at 5000 seconds, around 14 times the protein half-life. Each data point or histogram is the result of 100000 trials. Supporting Information {#s5} ====================== Text S1 ::: {.caption} ###### Details on the theoretical model, supplementary analysis, and simulations. (0.75 MB PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: We would like to thank Mariama El Baroudi and Antonio Celani for useful discussions. The authors have declared that no competing interests exist. No funders have supported this work. [^1]: Conceived and designed the experiments: MO CB MC. Performed the experiments: MO CB. Analyzed the data: MO CB DC. Wrote the paper: MO CB MC.

PubMed Central

2024-06-05T04:04:19.690801

2011-3-10

PMC3053348

Introduction {#s1} ============ Inflammation results from recognition of invading microorganisms through pathogen--associated molecular patterns (PAMPs) and from reaction to tissue damage--associated molecular patterns (DAMPs) [@ppat.1001315-Gallucci1], [@ppat.1001315-Janeway1]. It is known that the innate immune system recognizes both PAMPs and DAMPs through pattern recognition receptors, such as Toll--like receptors (TLRs) and other receptors [@ppat.1001315-Donato1], [@ppat.1001315-Schmidt1], [@ppat.1001315-Sparvero1], [@ppat.1001315-Lin1]. Multiple positive feedback loops between DAMPs and PAMPs and their overlapping receptors temporally and spatially drive immune regulatory functions. Despite the identification of specific signaling pathways negatively regulating responses to PAMPs or DAMPs [@ppat.1001315-ONeill1], [@ppat.1001315-Chen1], the unexpected convergence of molecular pathways responsible for recognition of PAMPs and DAMPs raised the question of whether and how the host discriminates between these two molecular patterns [@ppat.1001315-Liu1], [@ppat.1001315-Sitkovsky1]. DAMPs such as the high mobility group box 1 protein (HMGB1) and S100 proteins represent important danger signals that, although primarily intracellular, may mediate inflammatory responses through autocrine/paracrine interactions with the receptor for advanced glycation end--products (RAGE), a multiligand receptor of the immunoglobulin superfamily [@ppat.1001315-Donato1], [@ppat.1001315-Schmidt1], [@ppat.1001315-Sparvero1], [@ppat.1001315-Bianchi1], [@ppat.1001315-Donato2]. Integral to the biology of RAGE and its ligands is their up--regulation and increased accumulation in multiple biological and disease settings. The ability to activate expression programs that encode innate immune responsive genes confers to RAGE a central role in chronic inflammatory diseases. Engagement of RAGE converts a brief pulse of cellular activation to sustained cellular dysfunction, eventually leading to inflammation [@ppat.1001315-Schmidt1] and tumor promotion [@ppat.1001315-Gebhardt1]. However, because RAGE is expressed in multiple, distinct cell types, including immune cells, and both murine and human RAGE genes undergo extensive splicing with distinct splice isoforms being uniquely distributed in different tissues [@ppat.1001315-Kalea1], it is not surprising that diverse signal transduction and effector pathways may be impacted by RAGE depending on sites, ligands and time course of ligand--RAGE stimulation [@ppat.1001315-Leclerc1], [@ppat.1001315-Ostendorp1], [@ppat.1001315-Leclerc2]. The complexity of the system is enhanced by the findings that the ligands of RAGE may interact with distinct TLR--binding molecules thus amplifying inflammatory and immune responses in infection [@ppat.1001315-Bianchi1], [@ppat.1001315-Ivanov1], [@ppat.1001315-Tian1], [@ppat.1001315-Yanai1]. Thus, although promoting pathology, RAGE signaling also contributes to beneficial, inflammatory mechanisms of repair, in certain settings [@ppat.1001315-Sparvero1]. Ultimately, discerning the primal versus the chronic injury--provoking roles for this ligand--receptor interaction is a challenge in delineating the functions of the ligand/RAGE axis [@ppat.1001315-Clynes1]. Given that RAGE is expressed at the highest levels in the lung compared to other tissues [@ppat.1001315-Sparvero1], [@ppat.1001315-vanZoelen1] and both protects and causes lung injury [@ppat.1001315-Sparvero1], the DAMP/RAGE axis likely integrates with the PAMP/TLR axis in the inflammatory responses in lung infections. We have addressed whether and how the two systems interact in a mouse model of pulmonary infection with a model fungal pathogen as well a common cause of severe infections and diseases, *Aspergillus fumigatus* [@ppat.1001315-Segal1]. Humans inhale hundreds of conidia per day without adverse consequences [@ppat.1001315-Aimanianda1], except for a small minority of persons in whom defense systems fail and a life--threatening angioinvasive form of aspergillosis can develop. Some degree of inflammation is required for protection during the transitional response occurring between the rapid innate and slower adaptive response. However, progressive inflammation worsens disease and ultimately prevents pathogen eradication, a condition in which it is an exaggerated inflammatory response that likely compromises a host\'s ability to eradicate infection and not an "intrinsic" susceptibility to infection that determines a state of chronic or intractable disease [@ppat.1001315-Romani1]. We disclosed the complexity of signalling integration between different innate immune biosensors by showing that the spatiotemporal regulation of TLRs and RAGE by S100B limits pathogen-- as well as danger--induced inflammation and ensures protection in infection. Results {#s2} ======= RAGE and DAMPs expression in pulmonary aspergillosis {#s2a} ---------------------------------------------------- We assessed the expression of RAGE in the lungs of mice infected with *Aspergillus* conidia by immunohistochemical staining, protein and gene expression analysis. RAGE expression was observed at mRNA ([**Fig. 1A**](#ppat-1001315-g001){ref-type="fig"} and **[Fig. S1A](#ppat.1001315.s001){ref-type="supplementary-material"}**) and protein ([**Fig. 1B**](#ppat-1001315-g001){ref-type="fig"}) levels and maximally occurred in alveolar epithelial cells, as revealed by immunofluorescence staining ([**Fig. 1D**](#ppat-1001315-g001){ref-type="fig"}). On assessing which putative ligands of RAGE were concomitantly expressed in infection, we found that HMGB1 was not increased either at the level of gene ([**Fig. 1A**](#ppat-1001315-g001){ref-type="fig"} and **[Fig. S1A](#ppat.1001315.s001){ref-type="supplementary-material"}**) or protein ([**Fig. 1B**](#ppat-1001315-g001){ref-type="fig"}) expression. In contrast, S100B was promptly induced in infection, and declined thereafter to return to basal levels a week later, as revealed by gene and protein expression analysis in the lung ([**Fig. 1A,B**](#ppat-1001315-g001){ref-type="fig"} and **[Fig. S1A](#ppat.1001315.s001){ref-type="supplementary-material"}**) and protein secretion in the bronchoalveolar lavage fluid (BAL) ([**Fig. 1C**](#ppat-1001315-g001){ref-type="fig"}). S100B immunoreactivity was high in bronchiolar epithelial cells as revealed in wild--type mice (WT) (immunofluorescence staining in [**Fig. 1D**](#ppat-1001315-g001){ref-type="fig"}) or in transgenic mice expressing *s100b*--EGFP+ ([**Fig. 1E**](#ppat-1001315-g001){ref-type="fig"}). Further analysis on purified lung cells from transgenic mice confirmed that epithelial cells were major sources of S100B in infection (**[Fig. S1B](#ppat.1001315.s001){ref-type="supplementary-material"}**) while *Ager* was expressed on epithelial cells, macrophages, dendritic cells (DCs) (**[Text S1](#ppat.1001315.s005){ref-type="supplementary-material"}**) and polymorphonuclear neutrophils (PMNs) (**[Fig. S1C](#ppat.1001315.s001){ref-type="supplementary-material"}**), as described [@ppat.1001315-Sparvero1]. Presumably associated with PMNs\' infiltration, *s100a8* and *s100a9* expressions in the lung mainly occurred at 3 days post--infection ([**Fig. 1A**](#ppat-1001315-g001){ref-type="fig"}). These data suggest that S100B pairs with RAGE very early in infection before its transcriptional downregulation. The prompt induction of S100B followed by its downregulation suggests that S100B may serve as a danger signal to control inflammation. ::: {#ppat-1001315-g001 .fig} 10.1371/journal.ppat.1001315.g001 Figure 1 ::: {.caption} ###### RAGE and DAMPs expression in pulmonary aspergillosis. Expression of RAGE, HMGB1, S100B, S100A8 and S100A9 by RT--PCR (**A**), western blotting (**B, C**) and immunoistochemical staining (**D**) on the lungs of C57BL6 mice infected with *Aspergillus* conidia intranasally at different days postinfection (dpi). In **C**, S100B was assessed on BAL by western blotting. For immunohistochemistry, lung sections were incubated overnight with anti--S100B or anti--RAGE antibody followed by the secondary antibodies. Nuclei were counter--stained with DAPI. (**E**) Lung sections from transgenic mice expressing *s100b*--EGFP+ at different days postinfection. Bars, 100 µm. Microscopy was performed on a DM Rb epifluorescence microscope equipped with a digital camera. Representative of 2 experiments. :::  ::: RAGE--deficient mice develop pathogen--induced inflammation {#s2b} ----------------------------------------------------------- To determine the role of the S100B/RAGE axis in response to the fungus, we evaluated parameters of infection, inflammation and adaptive immunity in RAGE KO mice with pulmonary aspergillosis. Despite an initial higher fungal growth in the lung and brain of KO than WT mice, the fungal growth was eventually restrained in both types of mice ([**Fig. 2A**](#ppat-1001315-g002){ref-type="fig"}). Inflammation and signs of parenchyma damage, in contrast, were greatly exacerbated in RAGE KO mice and failed to resolve as opposed to WT mice ([**Fig. 2B**](#ppat-1001315-g002){ref-type="fig"}, upper panels with fungi magnified in the inset). The number of PMNs increased and maintained elevated in the lung parenchyma and the BAL fluids ([**Fig. 2B**](#ppat-1001315-g002){ref-type="fig"}, lower panels and inset) of RAGE KO mice. Gene expression analysis of the lung confirmed the higher and persistent inflammatory response in KO than WT mice, as revealed by the higher mRNA expression of *Cxcl1*, *Cxcl2* and *Mpo* genes as well as genes for inflammatory cytokines, such as IL--1β and IL--6 ([**Fig. 2C**](#ppat-1001315-g002){ref-type="fig"}). ::: {#ppat-1001315-g002 .fig} 10.1371/journal.ppat.1001315.g002 Figure 2 ::: {.caption} ###### RAGE--deficient mice develop pathogen--induced inflammation. Fungal growth (CFU±SE) (**A**), lung histology (PAS, H&E and Gomori stainings) and BAL morphometry (**B**), inflammatory chemokines and cytokine gene expression in the lungs by real--time RT--PCR (**C**), conidiocidal activity (**D**), oxidant production (**E**) in mice infected with live *Aspergillus* conidia intranasally. Note the sustained parenchymal damage (PAS staining in the upper panel of **Fig. 2B**), fungal growth (Gomori staining in the inset), and inflammatory cell recruitment in lungs (H&E staining in the lower panel) and BAL (May--Grünwald Giemsa--staining in the inset) in RAGE KO mice. Bars indicated magnifications. Dpi, days postinfection. BAL morphometry (numbers refer to % polymorphonuclear (PMN) or mononuclear (MNC) cells), lung RT--PCR, conidiocidal activity and oxidant production were assessed 3 days after the infection. Total lung cells, purified alveolar macrophages and PMNs were incubated with unopsonized resting conidia at 37°C for conidiocidal activity \[percentage of colony forming units\' inhibition (mean ± SE) at 60 min\] or oxidant production by DHR. PMA, phorbol 12--myristate 13--acetate. \**P*, KO vs WT mice. Data are pooled from 4 experiments or representative of 2 experiments (for histology). :::  ::: Despite the fact that an inflammatory response was not observed upon challenge with inactivated conidia (data not shown), the failure to resolve inflammation was not secondary to either a deficient conidiocidal activity of lung cells, including macrophages ([**Fig. 2D**](#ppat-1001315-g002){ref-type="fig"}), or a defective oxidant production ([**Fig. 2E**](#ppat-1001315-g002){ref-type="fig"}). PMNs only from KO mice showed between 25 to 35% reduction of their conidiocidal activity as compared to WT mice, a finding pointing to a requirement for RAGE in the execution of PMNs\' effector activity. These data indicate that RAGE, known to mediate PMN recruitment through interaction with beta 2 integrins [@ppat.1001315-Chavakis1], is neither required for lung inflammatory cell recruitment or oxidant production in aspergillosis but unexpectedly protects from unintended inflammation. Both the subverted innate inflammatory response to the fungus [@ppat.1001315-Romani1] and the requirement for RAGE in DC and T cell functions [@ppat.1001315-Manfredi1], [@ppat.1001315-Moser1] would predict altered adaptive Th responses to the fungus. This was indeed the case as shown by the results of lung DC and Th cell activation in response to the fungus. Purified DCs from RAGE KO mice responded to *Aspergillus* conidia or hyphae with higher expression level of mRNA for IL--1β, IL--6, IL--23 (p19), and similar levels of IL--12 (p35) or IL--10 compared to WT DCs (**[Fig. S2A](#ppat.1001315.s002){ref-type="supplementary-material"}**). Of interest, similar to the response to the fungus, higher levels of inflammatory cytokines were also observed in KO vs WT DCs in response to the TLR2/TLR6 ligand bacterial lipopeptide macrophage--activating lipopeptide (MALP--2) but not to LPS or ODN--CpG (**[Fig. S2B](#ppat.1001315.s002){ref-type="supplementary-material"}**). In terms of Th cell activation, although cytokine and transcription factor mRNAs were higher in unstimulated CD4+T cells from KO than WT mice, a further increased was observed for Th2 (*Gata3/Il4*) or Th17 b(*Rorc/Il17a*) but not for Th1 (*Tbet/Ifnγ*) or Treg (*Foxp3/Il10*) specific transcripts (**[Fig. S2C](#ppat.1001315.s002){ref-type="supplementary-material"}**). Thus, RAGE deficiency is associated with deregulated innate and adaptive antifungal immunity and the inflammatory program activated in DCs in response to the fungus/TLR ligands is compatible with the impairment of antifungal Th1/Treg protective responses and upregulation of inflammatory Th2/Th17 cell responses [@ppat.1001315-Zelante1], [@ppat.1001315-Bonifazi1]. RAGE pairs with S100B for anti--inflammatory and pro--inflammatory signals {#s2c} -------------------------------------------------------------------------- To formally prove that RAGE pairs with S100B in infection, experiments of S100B neutralization or administration were performed. We found that S100B neutralization decreased resistance to infection in WT mice as indicated by the increased fungal growth ([**Fig. 3A**](#ppat-1001315-g003){ref-type="fig"}), PMN recruitment and inflammation in the lung ([**Fig. 3B**](#ppat-1001315-g003){ref-type="fig"}), an effect that was mimicked by treatment with antibodies neutralizing RAGE engagement ([**Fig. 3A,B**](#ppat-1001315-g003){ref-type="fig"}). Accordingly, exogenously administered S100B decreased the fungal growth and the inflammatory pathology, but this occurred at nanomolar doses ranging from 5 to 50 ng/kg but not at doses up to 5000 ng/kg ([**Fig. 3A,B**](#ppat-1001315-g003){ref-type="fig"}). Both effects were RAGE--dependent being abrogated, albeit partially for low--dose S100B, in RAGE KO mice ([**Fig. 3A**](#ppat-1001315-g003){ref-type="fig"}). Similar experiments done for HMGB1 (**[Text S1](#ppat.1001315.s005){ref-type="supplementary-material"}**) showed that the fungal burden and inflammation were both increased upon its administration and involved RAGE (**[Fig. S3](#ppat.1001315.s003){ref-type="supplementary-material"}**). Because S100B itself didn\'t show a direct activity on fungal growth and morphology (**[Text S1](#ppat.1001315.s005){ref-type="supplementary-material"}** and **[Fig. S4](#ppat.1001315.s004){ref-type="supplementary-material"}**) and was ineffective if given before the infection (data not shown), these data suggest that S100B pairs with RAGE for anti-- and pro--inflammatory activities, a feature consistent with the unique ability of S100B to exhibit opposite effects depending on doses [@ppat.1001315-Donato2], [@ppat.1001315-Leclerc2], [@ppat.1001315-Donato3]. ::: {#ppat-1001315-g003 .fig} 10.1371/journal.ppat.1001315.g003 Figure 3 ::: {.caption} ###### Effects of S100B administration or neutralization on aspergillosis. Fungal growth (CFU±SE) (**A**), lung histology (PAS staining) and BAL morphometry (May--Grünwald Giemsa--staining in the inset) (**B**) in C57BL6 or RAGE KO mice infected with *Aspergillus* live conidia intranasally and treated intraperitoneally for 3 consecutive days with different doses (ng/kg) S100B, 1 mg/kg anti--S100B or 0.5 mg/kg anti--RAGE antibodies. *P*, treated *vs* untreated (--) mice. NF--κB activation was assessed by western blotting (**C**) on lung cells from WT and KO mice, untreated (--) or treated with 50 or 500 ng/kg S100B, a day after the last treatment, or by nuclear translocation (indicated by arrowheads in **D**) on purified PMNs, unexposed (Ct, control) or exposed in vitro for 60 min to *Aspergillus* conidia alone (--) or in the presence of S100B at 4 nM or 4 µM. Western blotting data are presented as immunoblots of cell lysates with phosphorylation--specific antibodies and fold increases (pixel density) in the phosphorylated to total protein ratios. (**E**) Oxidant production was assessed PMNs exposed as in D (by DHR at 60 min). (**F**) *Fas* and *Bcl2* expressions were assessed by real--time RT--PCR after 6 h--exposure. Representative of 3 experiments. \**P*\<0.05, treated *vs* untreated (--) cells. :::  ::: Mechanistically, we assessed whether S100B affected the activation of nuclear factor κB (NF--κB) and oxidant production, important inflammatory pathways downstream RAGE activation [@ppat.1001315-Donato1], [@ppat.1001315-Schmidt1] in vivo and in vitro on purified PMNs, known to respond to S100B [@ppat.1001315-Donato2]. In vivo, 500, but not 50, ng/kg S100B promoted RAGE--dependent NF--κB activation in the lung ([**Fig. 3C**](#ppat-1001315-g003){ref-type="fig"}). In vitro, micromolar but not nanomolar S100B activated NF--κB ([**Fig. 3D**](#ppat-1001315-g003){ref-type="fig"}) and increased oxidant production in response to the fungus ([**Fig. 3E**](#ppat-1001315-g003){ref-type="fig"}). Interestingly, and consistent with the dose--dependent prosurvival/prodeath effects of S100B on cells [@ppat.1001315-Huttunen1], *Fas* expression was decreased and antiapoptotic *Bcl2* expression increased in WT and KO PMNs exposed to nanomolar S100B and the opposite was true with micromolar S100B acting via RAGE ([**Fig. 3F**](#ppat-1001315-g003){ref-type="fig"}). These data confirm that RAGE activation by S100B is dependent on doses and also suggest that S100B may possess characteristics beyond the RAGE activating function which mediate its anti--inflammatory effects. The S100B/RAGE axis restrains TLR2/MyD88--dependent inflammation {#s2d} ---------------------------------------------------------------- Danger--sensing mechanisms are known to participate in the TLR responses to PAMPs [@ppat.1001315-Bianchi1], [@ppat.1001315-Yanai1] and to negatively regulate excessive inflammation during infection [@ppat.1001315-Sitkovsky1]. Given the ability of HMGB1 to bind and act in synergy with endogenous and exogenous TLR ligands [@ppat.1001315-Bianchi1], [@ppat.1001315-Hreggvidsdottir1], we assessed whether S100B also binds TLR ligands in solid phase by ELISA. We found that S100B highly binds exogenous and endogenous TLR ligands, such as MALP--2, HSP70, class B ODN--CpG (ODN 1982), mammalian DNA, fungal RNA and, partly, DNA, in a Ca^2+^--and dose--dependent manner, with the maximum binding activity observed at the nanomolar dose ([**Fig. 4A**](#ppat-1001315-g004){ref-type="fig"}). No binding was observed to Zymosan, LPS, double--stranded RNA \[polyinosinic--polycytidylic acid, Poly(I:C)\], non-CpG ODN (ODN 1982) or single--stranded RNA (the imidazoquinoline resiquimod R848). These data suggest that S100B may interact with TLR2 (HSP70) but not with Dectin--1 (Zymosan), with the heterodimer TLR2/TLR6 (MALP--2) and with intracellular nucleic acid--sensing TLRs. ::: {#ppat-1001315-g004 .fig} 10.1371/journal.ppat.1001315.g004 Figure 4 ::: {.caption} ###### The S100B/RAGE axis restrains TLR2/MyD88--dependent inflammation. (**A**) Binding of S100B to TLR ligands by ELISA. S100B was incubated in microtiter plates coated with 10 µg/ml (as by preliminary experiments) of MALP--2, Zymosan, HSP70, LPS, Poly(I:C), fungal DNA or RNA, mammalian genomic DNA, Class B ODN--CpG (2006), nonstimulatory ODN--CpG (2310) and R--848 with and without 1 mM EGTA. Data indicate the mean ±SE of triplicates from three independent experiments. \**P*\<0.05, S100B vs no S100B (--). (**B**) Levels of ERK1/2 or p38 MAPK (30 min) in purified PMNs exposed to 5 µg/ml MALP--2, nanomolar or micromolar S100B, 20 µg/ml anti--S100B antibody, 5 µM SB202190, 300 nM HMGB1, alone or in combination, for 30 min as described. Data are presented as immunoblots of cell lysates with phosphorylation--specific antibodies and fold increases (pixel density) in the phosphorylated to total protein ratios. Representative of 2 experiments. (**C**) TLR2-transfected HEK293 cells were stimulated with MALP--2 for 30 min with and without 4 nM or µM S100B or 20 µg/ml anti--S100B antibody. Cell lysates were subjected to immunoprecipitation after overnight incubation with 2 µg/ml polyclonal anti--S100B or anti--RAGE antibody. Immunoprecipitates were probed with antibodies to the corresponding antigens. RT-PCR analysis confirmed that both RAGE and S100B were expressed on tranfected HEK293 cells and experiments in KO cells confirmed the specificity of the anti-RAGE antibody (data not shown). Representative of 2 experiments. (**D**) Direct visualization of RAGE interaction with TLR2 in the presence of nanomolar S100B by in situ proximity ligation assay (PLA). TLR2-transfected HEK293 cells were transiently transfected with a RAGE expression vector (pcDNA3/RAGE) or empty vector (pcDNA3) and stimulated with MALP--2 for 30 min with or without 4 nM or µM S100B or 20 µg/ml anti--S100B antibody. Cells were stained with anti-RAGE and anti-TLR2 antibodies, and subjected to PLA. (**E**) *s100b* gene--expression by real--time RT--PCR in different TLR-deficient mice at different days postinfection (dpi) with *Aspergillus* conidia intranasally. \**P*\<0.05, dpi 1 and 3 *vs* 0. (**F**) *S100b* gene--expression by real--time RT--PCR in lung of C57BL6 mice injected with 2.5 µg MALP--2, 10 µg LPS, 50 µg Poly(I:C), 50 µg Class B ODN--CpG 3 days before. *P*, treated *vs* untreated (--) mice. (**G**) H&E--stained sections from C57BL6 or RAGE KO mice injected as in **F**, 3 days before. Representative of 2 experiments. :::  ::: Because TLR2 activates the inflammatory state of PMNs in infection [@ppat.1001315-Bellocchio1], [@ppat.1001315-Moretti1] and unrestrained inflammation occurred in condition of defective S100B/RAGE axis, we hypothesized that the S100B/RAGE axis may inhibit TLR2/MyD88--driven inflammation to the fungus. We assessed therefore whether and how nanomolar or micromolar S100B would affect TLR2--mediated activation of PMNs. We found that ERK phosphorylation in response to MALP--2 was inhibited by nanomolar S100B and potentiated by micromolar S100B or by blocking serum S100B. This occurs through a p38--dependent mechanism, as shown by the ability of nanomolar S100B to induce p38--phosphorylation as well as ERK phosphorylation in the presence of the specific p38 inhibitor SB202190. Like micromolar S100B, HMGB1 activated ERK more than p38 phosphorylation ([**Fig. 4B**](#ppat-1001315-g004){ref-type="fig"}). These data indicate that S100B, like HMGB1 [@ppat.1001315-Bianchi1], potentiates the biological activity of TLR2 ligands upon forming complexes with them. However, they also unexpectedly revealed that forming complexes with nanomolar S100B negatively regulates their functions. That p38 is a negative regulator of TLR2 expression has already been described [@ppat.1001315-Sahay1]. We assessed here whether inhibition of TLR2 occurred through physical association by performing immunoprecipitation studies with TLR2--transfected HEK cells stimulated with MALP--2, in the presence or not of S100B. Both RAGE and TLR2 were found to associate with nanomolar or micromolar S100B upon TLR2 stimulation, but TLR2 physically associated with RAGE only in the presence of nanomolar S100B ([**Fig. 4C**](#ppat-1001315-g004){ref-type="fig"}). Although RAGE was found to be expressed in TLR2--transfected HEK 293 cells (unpublished observations), we transiently transfected TLR2--HEK 293 cells with RAGE to visualize the RAGE/TLR2 interaction using the in situ proximity ligation assay. We confirmed that RAGE strongly interacts with TLR2 in the presence of nanomolar but not micromolar S100B. In addition, the lack of interaction observed upon S100B neutralization suggests that endogenous S100B likely mediates TLR2/RAGE physical interaction in steady-state conditions ([**Fig. 4D**](#ppat-1001315-g004){ref-type="fig"}). Thus, S100B interacts with RAGE and TLR2 and mediates the physical association of the two at nanomolar doses. Because S100B production mainly occurred in infection via the TLR2/MyD88 pathway ([**Fig. 4E, F**](#ppat-1001315-g004){ref-type="fig"}), our findings indicate the existence of an autocrine/paracrine loop by which TLR2--induced S100B binds to extracellular RAGE to inhibit TLR2 upon physical association. This scenario would suggest an increased responsiveness to TLR2--mediated inflammation of RAGE KO mice. Consistent with the high reactivity of DCs to MALP--2 (**[Fig. S2B](#ppat.1001315.s002){ref-type="supplementary-material"}**), the inflammatory response to intranasally delivered MALP--2 was higher in RAGE KO than WT mice as compared to other TLR agonists the sensitivity to which was not different between KO and WT mice ([**Fig. 4G**](#ppat-1001315-g004){ref-type="fig"}). TLR3/9 signaling inhibits S100B expression via TRIF/noncanonical NF--κB {#s2e} ----------------------------------------------------------------------- The finding that S100B production in vivo was upregulated in the absence of TLR3 and TRIF, conditions in which we noticed a defective transcriptional downregulation of S100B ([**Fig. 4E**](#ppat-1001315-g004){ref-type="fig"}), led us to suppose that binding to intracellular nucleic acids is a mechanism by which S100B is down-regulated and its pro--inflammatory activity restrained in infection. We resorted to lung epithelial cells as major sources of S100B in infection. We assessed p38 phosphorylation in cells from WT and selected TLR--KO mice exposed to *Aspergillus* resting (RC) or swollen (SC) conidia, MALP--2, Poly(I:C) or ODN--CpG and the relative contribution of endogenous S100B. We found that p38 phosphorylation occurred maximally in response to *Aspergillus* RC and Poly(I:C), to a lesser extent in response to ODN--CpG and SC, did not occur in response to MALP--2, and was largely TLR3/TLR9/TRIF--dependent, but MyD88--independent ([**Fig. 5A**](#ppat-1001315-g005){ref-type="fig"}). Both *Ifnb1* and *Ifna1* gene expression in response to Poly(I:C) were unaffected upon the addition of S100B but decreased upon neutralizing S100B by siRNA ([**Fig. 5B**](#ppat-1001315-g005){ref-type="fig"}), a finding indicating that S100B participates in the functional sensing of intracellular nucleic acids by TLR3. Although similar results were obtained in response to ODN--CpG, the overall responsiveness of epithelial cells to TLR9 was lower (data not shown), as already reported [@ppat.1001315-Sha1]. In terms of source of intracellular nucleic acids, consistent with the binding activity of S100B in vitro, we found that fungal RNA not only complexes with S100B in infection ([**Fig. 5C**](#ppat-1001315-g005){ref-type="fig"}) but also activates epithelial cells in a TLR3--dependent manner, as indicated by IRF3 phosphorylation ([**Fig. 5D**](#ppat-1001315-g005){ref-type="fig"}). ::: {#ppat-1001315-g005 .fig} 10.1371/journal.ppat.1001315.g005 Figure 5 ::: {.caption} ###### TLR3 and TLR9 inhibit *s100b* expression via TRIF/noncanonical NF--κB. (**A**) Levels of p38 phosphorylation on purified epithelial cells from C57BL6 mice or TLR KO mice, unexposed (--) or exposed to *Aspergillus* resting (RC) or swollen (SC) conidia, 5 µg/ml MALP--2, 10 µg/ml Poly(I:C) or 10 µg/ml ODN--CpG for 8 h. (**B**) *Ifnb1* or *Ifna1* gene expression by real time RT--PCR on epithelial cells from C57BL6 mice exposed to Poly(I:C) in the presence of siS100B or 4 nM S100B. \**P*\<0.05, siRNA --treated *vs* untreated cells. Representative of 2 experiments. (**C**) S100B co-localization with fungal RNA in HEK293 cells pulsed with fungal RNA and stained with Syto17 red fluorescent nucleic acid stain and anti-S100B antibody followed by FITC-conjugated goat anti--rabbit. Mock-pulsed (control) and pulsed cells were analyzed on confocal microscope. (**D**) Fungal RNA stimulates epithelial cells via TLR3. Levels of IRF3 phosphorylation on lung epithelial cells exposed to Poly(I:C) or fungal RNA. Data are presented as immunoblots of cell lysates and fold increases (pixel density) in the phosphorylated to total protein ratios. (**E**) Immunoblots and pixel density of IKKβ and IKKα phosphorylation in epithelial cells from C57BL6 mice exposed as in (**A**). (**F**) Results from an ELISA procedure to monitor activation of p65, p52, and RelB in nuclear extract from epithelial cells exposed as in (**A**) for different lengths of time. Time 0 indicates untreated cells. Relative activities (A~450~) are mean±SE of two experiments, each in triplicate. \**P*\<0.05, treated vs untreated cells. (**G**) *s100b* gene expression by real time RT--PCR upon canonical/noncanonical NF--κB inhibition by siRNA in epithelial cells exposed to MALP--2 or Poly(I:C), as above. SB202190 (5 µM) was used as p38 inhibitor. \**P*\<0.05, siRNA--treated or p38--inhibited cells *vs* MALP--2 or Poly(I:C)--exposed cells. Representative of 2 experiments. :::  ::: The transcriptional downregulation of *s100b* in infection led us to hypothesize that transcription factors downstream p38/TRIF would mediate this effect. Given the existence of specific binding sites for NF--κB family members in the promoter of both human (GenBank: M59486) and murine (GenBank: NC\_000076.5) *S100b*, we assessed whether NF--κB transcription factors regulate *s100b* gene expression. For this purpose, we evaluated the activation of canonical/noncanonical NF--κB pathways downstream TLR2/MyD88 and TLR3/TLR9/TRIF and their contribution to *s100b* gene expression. Of the two IkB kinase complex catalytic subunits, known to have opposing roles in inflammation [@ppat.1001315-Bonizzi1], IKKβ more than IKKα was phosphorylated in response to SC, MALP--2, the opposite was true in response to Poly(I:C), while both pathways were activated by ODN--CpG ([**Fig. 5E**](#ppat-1001315-g005){ref-type="fig"}). We also quantified the activation of the NF--κB family members, using an ELISA kit specific for mouse p65, p52 and RelB. Significant nuclear translocation occurred for p52 and RelB, but not for p65, following 15--60 min of exposure to Poly(I:C). Significant translocation of all members was observed in response to ODN--CpG, while only p65 translocation occurred in response to MALP--2 ([**Fig. 5F**](#ppat-1001315-g005){ref-type="fig"}). The two pathways had opposite effects on *S100b* gene expression, as shown by experiments in which either pathway was silenced by siRNA. *S100b* expression was inhibited upon blocking the canonical pathway or promoted upon blocking the noncanonical, p38-dependent, pathway ([**Fig. 5G**](#ppat-1001315-g005){ref-type="fig"}). These data suggest that *s100b* expression is transcriptionally regulated by the sequential action of downstream MyD88-- and TRIF--dependent NF--κB signalling pathways. S100B activity in vivo is contingent upon TLRs {#s2f} ---------------------------------------------- Experiments in vivo confirmed that the pro- and anti-inflammatory activity of S100B is contingent upon these TLRs. The anti--inflammatory activity of nanomolar S100B, as revealed by the fungal growth restriction and PMN recruitment, occurred independently of TLR4 but required the presence of TLR2, TLR6 and the MyD88 adaptor. Consistent with the ability of the TLR3/TLR9/TRIF pathway to downregulate *s100b*, S100B became pro--inflammatory at the nanomolar dose in the relative absence of TLR9, TLR3 and the adaptor TRIF ([**Fig. 6A, B**](#ppat-1001315-g006){ref-type="fig"}). These in vivo findings confirm that the spatiotemporal integration of signals from TLRs and RAGE by S100B limits pathogen-- and danger--induced inflammation in murine aspergillosis ([**Fig. 7**](#ppat-1001315-g007){ref-type="fig"}). ::: {#ppat-1001315-g006 .fig} 10.1371/journal.ppat.1001315.g006 Figure 6 ::: {.caption} ###### S100B activity *in vivo* is contingent upon TLR signaling. Fungal growth (CFU±SE) (**A**) and (**B**) *Mpo* expression by real time RT--PCR in the lung of TLR--deficient mice infected with *Aspergillus* conidia and treated with 50 ng/ml S100B as in legend to [**Fig. 3**](#ppat-1001315-g003){ref-type="fig"}. Data were obtained at 3 days postinfection. Representative of 4 experiments. *P*, treated vs untreated (None) mice. :::  ::: ::: {#ppat-1001315-g007 .fig} 10.1371/journal.ppat.1001315.g007 Figure 7 ::: {.caption} ###### Spatiotemporal integration of signals from TLRs and RAGE by S100B limits pathogen-- and danger--induced inflammation. The figure shows that TLR2 activation on epithelial cells by the fungus resulted in the release of the Ca^2+^--binding protein, S100B, that paracrinally binds to RAGE, on polymorphonuclear neutrophils, and mediates its association with TLR2 with subsequent inhibition. However, S100B, upon binding to nucleic acids, also activates an intracellular TLR3/TLR9/TRIF--dependent pathway culminating in the transcriptional down-regulation of S100B. The transcriptional regulation of S100B by the sequential action of downstream MyD88-- and TRIF--dependent NF--κB signalling pathways provides the molecular basis for an evolving braking circuit in infection whereby the endogenous danger protects the host against pathogen--induced inflammation and a nucleic acid--sensing mechanism resolves danger--induced chronic inflammation. :::  ::: Discussion {#s3} ========== To initiate an appropriate inflammatory response, organisms have developed ways to recognize potentially life--threatening signals. Our study reveals that sequential signaling between different innate immune biosensors serves to limit pathogen-- and danger-- induced collateral inflammation in infection. This occurs through a previously undescribed TLR/RAGE interaction via S100B, an EF--hand calcium--binding protein, with both intra-- and extracellular activities, that acts in either an autocrine or paracrine manner through RAGE [@ppat.1001315-Donato2], [@ppat.1001315-Leclerc1]. RAGE is known to interact with TLR9 via HMGB1 which results in either a potentiation [@ppat.1001315-Tian1] or suppression [@ppat.1001315-Popovic1] of TLR9 function. We found that, upon engagement, RAGE associated with and inhibited TLR2. This occurred through epithelial cell--released S100B that paracrinally inhibited the TLR2--dependent activity of recruited PMNs, a finding consistent with the ability of RAGE to impair neutrophil functions [@ppat.1001315-Collison1] as well as with the down--regulated TLR2 activity in pulmonary aspergillosis [@ppat.1001315-Chai1]. PMNs\' recruitment is a characteristic feature of pulmonary aspergillosis [@ppat.1001315-Segal1] and PMNs\' activity is tightly regulated by TLRs [@ppat.1001315-Bellocchio1]. RAGE was dispensable for PMNs\' recruitment but potently regulated TLR2--induced MAPK kinase activation, NF--κB phosphorylation and survival, via low, but not high, doses of S100B. Notably, the ability of S100B to bind TLR2 also predicts an activity on TLR2 in a RAGE-autonomous fashion. Therefore, consistent with the biology of RAGE and its ligands [@ppat.1001315-Donato1], [@ppat.1001315-Bierhaus1], their up--regulation exerted a proximal role in the inflammatory cascade. However, at least for extracellular S100B, the interaction with RAGE also served to limit pathogen--induced inflammation. Thus, S100B plays a dual role in infection, restraining the inflammatory response in the early response to pathogen, through a paracrine epithelial cells/PMNs braking circuit, but also contributing, similar to HMGB1, to excessive inflammation through feed--forward RAGE activation [@ppat.1001315-Chavakis1] and likely through additional TLR interactions. The opposite effects on cells observed with low and high doses of S100B could be mechanistically explained by considering that calcium binding triggers structural changes in the S100 protein that allow interaction with target proteins as an octamer or a higher-order multimer form [@ppat.1001315-Ostendorp1]. Trophic *vs* toxic effects are observed on neuronal cells in which nanomolar S100B stimulate neurite growth and promote survival, while micromolar levels result in inflammatory effects [@ppat.1001315-Huttunen1], [@ppat.1001315-Huttunen2]. Structural and biochemical data have provided evidence that octameric S100B is highly stable and triggers RAGE activation by receptor dimerisation resulting in high--affinity binding [@ppat.1001315-Leclerc1], [@ppat.1001315-Ostendorp1] to the RAGE V and C(1) domains activating NF--κB [@ppat.1001315-Leclerc1]. Both domains are important for ligand binding and for intracellular signaling, respectively. In contrast, nanomolar S100B required RAGE to inhibit TLR2 for the elaboration of its anti--inflammatory activity. We have already shown that some S100B--induced cellular effects may not depend on RAGE signalling yet requiring the receptor [@ppat.1001315-Adami1]. This appears to be the case in our model in which the ability of S100B to bind endogenous and exogenous TLR2 ligands may offer a plausible molecular explanation for RAGE/TLR2 physical association. How this may prevent TLR2 signalling is not obviously clear, although signalling by TLR2 upon binding of ligands possessing fatty acyl moieties suggests a dynamic model of interaction, in which only a specific orientation of the ligand favors formation of a signal inducing ternary complex [@ppat.1001315-Kajava1]. Thus, similar to HMGB1 [@ppat.1001315-Bianchi1], S100B, by forming complexes with various TLR ligands, may present the partner molecule to its normal receptor in a way in which the conformation of the partner molecule is changed or in an allosteric interaction with the receptor or both. The ability of S100B to bind nucleic acids, while qualifying S100B as possible sentinel for nucleic acid--mediated immune activation [@ppat.1001315-Yanai1], also serves to explain the intracellular function of S100B in epithelial cells in infection. It is known that, upon calcium binding, the change of conformation in the C--terminal domain of S100B allows the exposure of hydrophobic residues critical for the binding to a variety of target proteins [@ppat.1001315-Heizmann1], thereby affecting their activities and allowing the elaboration of a variety of intracellular functions [@ppat.1001315-Donato2]. We found that S100B was able to bind, in a calcium--dependent manner, Class B ODN--CpG, mammalian DNA and fungal RNA and DNA, resulting in the activation of a p38/TRIF--dependent signalling, downstream TLR3 and TLR9. Thus, intracellular S100B may signal trough both TLR3 and TLR9 to scavenge pathogen-- and host--derived nucleic acids. Of interest, at variance with HMGB1 [@ppat.1001315-Tian1], S100B also discloses a TLR9--depending signalling pathway that converges on TRIF rather than on MyD88 [@ppat.1001315-ONeill1]. The molecular basis for this result in epithelial cells is presently under investigation, but is consistent with the finding that modification of the structure of the DNA ligand affects its sub--cellular localization and this may impact on sorting and signaling adapters as well as the biological response to TLR9 activation in DCs [@ppat.1001315-Guiducci1], [@ppat.1001315-Honda1]. That S100B may affect the intracellular compartmentalization of DNA upon binding pathogen and self DNA is, ultimately, a likely expectation for a chaperon molecule that localizes to the cytoplasm in a soluble form and in complexes with cytoskeletal and filament--associated target proteins [@ppat.1001315-Donato3]. This may also predict an inherent risk of autoimmunity associated with S100B. Incidentally, elevated levels of S100B have been observed in certain immuno--mediated diseases [@ppat.1001315-Donato2]. In addition to binding fungus DNA, whose unmethylated CpG motifs activates TLR9 [@ppat.1001315-RamirezOrtiz1], S100B also bound fungal RNA, a PAMP able to activate DCs for antifungal priming [@ppat.1001315-Bozza1]. That endogenous mRNA [@ppat.1001315-Kariko1] and pathogen RNA [@ppat.1001315-Aksoy1] activate TLR3 is an established finding. We found that endogenous S100B binds fungal RNA and activation of epithelial cells by fungal RNA is TLR3--dependent. Thus, in addition to sensing tissue necrosis [@ppat.1001315-Cavassani1], TLR3, abundantly expressed on epithelial cells [@ppat.1001315-Sha1], functions as an endogenous sensor of fungal RNA. Even more interesting is the finding that the activation of the TRIF--dependent, nucleic acid sensing pathway, mainly considered an inducer of antimicrobial innate immune responses, contributes to resolution of inflammation in infection. This occurs by downregulating *s100b* gene expression transcriptionally via noncanonical NF--κB signalling. Although *s100b* gene expression is tightly regulated in human cells [@ppat.1001315-Castets1], little is known about mechanisms regulating its transcription. The transcriptional regulation of *s100b* expression by the sequential action of downstream MyD88-- and TRIF--dependent NF--κB signalling pathways is thus a novel finding that not only establishes a link between pathogen-- and danger--sensing signaling pathways but also confirms the inhibitory role of TLR3 on the S100B/RAGE axis [@ppat.1001315-Cai1]. *In toto*, we have identified a mechanism that discriminates between pathogen-- and danger--induced immune responses via the spatiotemporal integration of signals from different innate immune biosensors. Conceptually, our study details an evolving braking circuit in infection whereby an endogenous danger protects the host against pathogen--induced inflammation and a nucleic acid--sensing mechanism terminates danger--induced inflammation. Thus, in addition to the notion that danger signal may terminate overactive immune responses [@ppat.1001315-Sitkovsky1], our study reveals that a pathogen--induced signal may also terminate unnecessary danger--induced injury. This raises the intriguing possibility that the host may have developed mechanisms to ameliorate the response to DAMPs via PAMPs. The scenario is dominated by the highly adaptive S100B/RAGE axis that, in sensing danger, plays a critical and unanticipated role as a fine modulator of inflammation via the promiscuous activity of S100B at the extracellular and intracellular levels. On a translational level, our findings suggest that a defective danger sensing associated with the different isoforms of the RAGE receptor may underlie individual differences in the clinical course of invasive aspergillosis and the inherent patient\'s susceptibility to infection. Materials and Methods {#s4} ===================== Ethics statement {#s4a} ---------------- Experiments were performed according to the Italian Approved Animal Welfare Assurance A--3143--01. Legislative decree 157/2008-B regarding the animal licence obtained by the Italian Ministry of Health lasting for three years (2008--2011). Infections were performed under avertin anesthesia and all efforts were made to minimize suffering. Mice {#s4b} ---- Female C57BL6 mice, 8--10 wk old, mice were purchased from Charles River (Calco, Italy). Homozygous *Tlr2--/--, Tlr3--/--, Tlr4--/--, Tlr9--/--, Myd88--/-- and Trif--/--* mice on a C57BL6 background were bred under specific pathogen--free conditions at the Animal Facility of Perugia University, Perugia, Italy. RAGE*--/--* mice were obtained from Dr. Angelika Bierhaus (Heidelberg, Germany). *s100b*--EGFP+ transgenic mice [@ppat.1001315-Vives1] were obtained from Dr. Catherine Legraverend (Montpellier, France). Fungal strains, infections, and treatments {#s4c} ------------------------------------------ The strain of *A. fumigatus* was obtained as described [@ppat.1001315-Romani1]. Viable resting, swollen *Aspergillus* conidia and hyphae were obtained as described [@ppat.1001315-Bonifazi1]. For infection, mice were anesthetized by intraperitoneal (i.p.) injection of 2.5% avertin (Sigma Chemical Co, St. Louis, MO) before instillation of a suspension of 2×10^7^ conidia/20 µl saline intranasally (i.n.). Fungi were suspended in endotoxin--free (Detoxi--gel; Pierce, Rockford, IL) solutions (\<1.0 EU/mL, as determined by the LAL method). Mice were monitored for fungal growth \[Colony forming units (CFU 9/organ, mean ± SE\]. BAL was performed by cannulating the trachea and washing the airways with 3 ml of PBS to collect the BAL fluid. Total and differential cell counts were done by staining BAL smears with May--Grünwald Giemsa reagents (Sigma) before analysis. At least 200 cells per cytospin preparation were counted and the absolute number of each cell type was calculated. Photographs were taken using a high--resolution Microscopy Olympus DP71 (Olympus, Milan, Italy). Mice were treated daily i.p. for 3 consecutive days starting the day of the infection with different doses of purified S100B (see below), 1 mg/kg polyclonal rabbit anti--S100B antibodies (Swant, CH--6501 Bellinzona, Switzerland) or 0.5 mg/kg anti--RAGE goat polyclonal IgG (Santa Cruz Biotechnology, inc. DBA, Milan, Italy). Control received PBS or isotype controls (Sigma--Aldrich). In vivo treatments with TLR agonists {#s4d} ------------------------------------ MALP--2 (2.5 µg), Poly(I:C) (50 µg), ultrapure LPS from *Salmonella minnesota* Re 595 (10 µg) (all from Sigma Chemical Co) and Class B ODN--CpG (50 µg) [@ppat.1001315-Aimanianda1] were given once intranasally to mice infected as above. Control received PBS or isotype controls (Sigma--Aldrich). Mice were sacrificed three days after treatment for histology (H&E staining) and *s100b* expression by real--time RT--PCR. Control received PBS. Histology, fluorescence and immunohistochemistry {#s4e} ------------------------------------------------ For histology, sections of paraffin--embedded tissues were stained with the periodic acid--Schiff (PAS), hematoxylin and eosin (H&E) or Gomori\'s methenamine Silver procedures [@ppat.1001315-Romani1]. For detecting S100B--expressing cells, lungs in OCT or purified cells from *s100b*--EGFP mice were analyzed. For immunohistochemistry, lung sections were incubated overnight with polyclonal anti--S100B antibody (1∶100) or polyclonal anti--RAGE antibody (1∶20) followed by the secondary antibodies, i.e., tetramethyl rhodamine isocyanate--conjugated goat anti--rabbit IgG (Sigma--Aldrich) for S100B, and AlexaFluor 594 donkey anti--goat IgG (Invitrogen), for RAGE. Nuclei were counter--stained with 4′,6-diamidino-2-phenylindole (DAPI). Endogenous peroxidase activity was quenched using 3% H2O2 in PBS. Immunostaining of lungs from RAGE KO mice was used as negative controls. Fluorescence and immunofluorescence microscopy was performed on a DM Rb epifluorescence microscope equipped with a digital camera (Leica, Wetzlar, Germany). Cell preparation, cultures and treatments {#s4f} ----------------------------------------- Purified lung CD11b^+^Gr--1^+^ PMNs (\>98% pure on FACS analysis) were obtained as described [@ppat.1001315-Bellocchio1]. Lung epithelial cells, at ∼99% expressing cytokeratin, on pan--cytokeratin antibody staining of cytocentrifuge preparations, and \>90% viable on trypan blue exclusion assay, were isolated as described [@ppat.1001315-You1].The average yield of tracheal cells was 1.7×10^5^ cells/trachea \[±0.58×10^5^ (SD)\]. Alveolar macrophages were purified by plastic adherence. Total lung cells, purified alveolar macrophages and PMNs were incubated with unopsonized resting conidia at 1∶1 ratio at 37°C for conidiocidal activity \[percentage of colony forming units inhibition (mean ± SE) at 60 min\] or oxidant production \[oxidation of dihydrorhodamine 123 (DHR), Molecular Probes (Invitrogen S.R.L. San Giuliano Milanese, Milan, Italy, measured by fluorimetry with the multifunctional microplate reader Tecan Infinite 200, Tecan Austria GmbH, Salzurg, Austria) at different time points. PMNs or epithelial cells were exposed to nanomolar or micromolar S100B as described [@ppat.1001315-Sorci1], 20 µg/ml anti--S100B antibody (SWant), 300 nM HMGB1, 5 µg/ml MALP--2, 10 µg/ml Poly(I:C), 10 µg/ml ultrapure LPS from *Salmonella minnesota* Re 595 and 10 µg/ml ODN--CpG. In vitro experiments were done in the presence of 2% FBS. Control cells were treated with PBS, DMSO or control antibody. S100B binding assays {#s4g} -------------------- S100B binding to TLR ligands was assessed in solid phase by ELISA. Briefly, plates were coated overnight at 4°C with 10 µg/ml (based on preliminary experiments) of MALP--2, Zymosan (Sigma Aldrich), HSP70 (StressMarq Biosciences Inc, Victoria Canada), Poly(I:C) or LPS in carbonate buffer (pH 9.55) or total fungal DNA or RNA, human DNA, Class B ODN--CpG (2006), non-CpG ODN (ODN 1982) [@ppat.1001315-Aimanianda1], resiquimod (R--848, Invivogen, Labogen S.r.l. Rho, Italy) in Reacti--Bind^TM^ DNA Coating Solution (Pierce). Fungal DNA and RNA were obtained as described [@ppat.1001315-Bozza1]. Total fungal RNA was routinely pretreated with RNase--free DNase I (50 units of DNase I/100 mg RNA) (Sigma Aldrich) at 25°C for 2 h. Nanomolar or micromolar S100B was added in blocking buffer (TBS 1%BSA) for 2 h at room temperature followed by the addition of rabbit anti--S100B antibodies (1∶1000) and HRP--conjugated rabbit secondary antibody (R&D Systems, Space Import--Export srl Milano, Italy). EGTA was used at 1 mM. The plates were developed using TMB Microwell Peroxidase Substrate system (BioFX Laboratories, Inc MD, U/SA). ODs were read at 450 nm. Data indicate the mean ±SE of triplicates from three independent experiments. Syto17 red fluorescent nucleic acid stain {#s4h} ----------------------------------------- To detect S100B co-localization with fungal RNA, TLR2--transfected HEK293 cells were pulsed with fungal RNA by means of N-\[1-(2,3-dioleoyloxypropyl\]-N,N,N,-trimethylammonium methylsulfate (DOTAP; Roche), as described [@ppat.1001315-Bozza1]. After pulsing cells were fixed in 3.7% formaldehyde, Triton--X100 permeabilized and incubated with Syto17 red fluorescent universal nucleic acid stain (Molecular Probe; 2.5 µM, 5 min) and anti-S100B antibody (1∶20 dilution) followed by FITC-conjugated goat anti--rabbit IgG (Vector Laboratories). Mock-pulsed (control) and pulsed cells were analyzed on confocal microscope Nikon Eclipse TE-2000U (Tokyo, Japan). Western blotting {#s4i} ---------------- Blots of cells lysates were incubated with monoclonal rabbit monoclonal anti--S100B IgG (clone EP1576Y, Epitomics, CA), goat polyclonal anti--RAGE IgG (Santa Cruz Biotechnology, Inc.), rabbit anti--HMGB1 IgG2a (Calbiochem, Milan, Italy), mouse monoclonal anti--TLR2 IgG2a, Santa Cruz Biotechnology, Inc.), rabbit polyclonal Abs recognizing the unphosphorylated form of ERK and p38 followed by horseradish peroxidase--conjugated anti--goat, mouse or rabbit IgG (Cell Signaling Technology) or biotin--conjugated (Vectastain Elite ABC system; Vector Laboratories, Burlingame, CA, USA) secondary antibodies. Blots were developed with the Enhanced Chemiluminescence detection kit (Amersham Pharmacia Biotech, Milan, Italy) and SuperSignal West Pico (Pierce). Scanning densitometry was done on a Scion Image apparatus. The pixel density of bands was normalized against total proteins or tubulin. The inhibitor p38 (5 µM, SB202190) was purchased from Calbiochem (San Diego, CA) and dissolved at 1000× the final concentration in DMSO (Sigma). Control experiments included staining without the primary antibody. Co--immunoprecipitation {#s4j} ----------------------- The human HEK293 embryonic kidney cell lines stably transfected with human TLR2 were maintained as described [@ppat.1001315-Moretti1]. Cells were stimulated with MALP--2 for 30 min with and without 4 nM or µM S100B or 20 µg/ml anti--S100B antibodies (SWant). Cell lysates were subjected to immunoprecipitation after overnight incubation with 2 µg/ml polyclonal anti--S100B (SWant) or anti--RAGE (Santa Cruz Biotechnology, Inc) antibody. Immunoprecipitates were probed with antibodies to the corresponding antigens. Control experiments included western blottings on immunoprecipated with an irrelevant antibody. In situ proximity ligation assay (PLA) {#s4k} -------------------------------------- We resorted to PLA [@ppat.1001315-Soderberg1] to directly visualize the RAGE interaction with TLR2 in an S100B--dependent manner. TLR2--transfected HEK293 cells were transiently transfected with a RAGE expression vector (pcDNA3/RAGE) or empty vector (pcDNA3) and stimulated with MALP--2 for 30 min with or without 4 nM or µM S100B or 20 µg/ml anti--S100B antibody (SWant). Cells were then fixed in cold methanol and treated with a rabbit anti--RAGE (H300, Santa Cruz Biotechnology, Inc) and a goat anti--TLR2 antibody, and subjected to PLA (OLINK Bioscience, Uppsala) according to the manufacturer\'s instructions. Cells were visualized on the DM Rb epifluorescence microscope. Canonical and noncanonical NF--κB {#s4l} --------------------------------- To detect NF--κB (p65) nuclear translocation, purified PMNs were fixed in cold methanol, permeabilized with Triton-X100 0.1% in PBS, incubated with blocking solution (PBS containing 3% BSA and 1% glycine), and incubated overnight at 4°C with rabbit anti-p65 (C-20) antibody (sc-372, Santa Cruz Biotechnology; 1∶50 dilution) followed by tetramethyl rhodamine isocyanate-conjugated goat anti-rabbit IgG (Sigma-Aldrich; 1∶50 dilution) as secondary antibody. Nuclei were counter-stained with DAPI. Cells were visualized on the epifluorescence microscope. We used an ELISAbased TransAM Flexi NFkB Family Kit (Active Motif) to monitor activity of NF--κB family members. Anti--phospho--IKKα (Ser180)/IKKβ (Ser181) rabbit Abs (Cell Signaling Technology) were used for western blotting of phospho IKKα and IKKβ. Western blotting with specific polyclonal antibodies (Santa Cruz Biotechnology) was done to assess level of p65. Exposure of epithelial cells to fungal RNA {#s4m} ------------------------------------------ Epithelial cells were exposed to fungal RNA (25 µg/ml) [@ppat.1001315-Bozza1] for 8 h before determination of levels of IRF3 phosphorylation by immunoblotting with rabbit polyclonal anti--IRF3 antibodies and anti--rabbit--horseradish peroxidase (Santa Cruz Biotechnology Inc.). Data are presented as immunoblots of cell lysates and fold increases (pixel density) in the phosphorylated to total protein ratios. Expression and purification of S100B {#s4n} ------------------------------------ Recombinant bovine S100B, 97% identical to mouse S100B, was expressed and purified as reported [@ppat.1001315-Donato3], [@ppat.1001315-Huttunen1]. Purified S100B was passed through END--X B15 Endotoxin Affinity Resin column to remove contaminating bacterial endotoxin. The S100B concentration was calculated using the M~r~ of the S100B dimer (21 kDa). SiRNA synthesis and transfection {#s4o} -------------------------------- SiRNA to target IKKα, IKKβ and S100B were done as described [@ppat.1001315-Bonifazi1]. The siRNA specific sequences were selected, synthesized and annealed by the manufacturer, and were used in combination with nontargeted control siRNA (Ambion, Applied Biosystem International, Monza Italy). Transfections of siRNA (at 1 nM/well) were performed by using the INTERFERin^TM^Transfection reagent, as per manufacturer\'s instructions (PEQLAB Biotechnologie GmbH, Erlangen, Germany). Cells were stimulated 48 h after transfection at 37°C. Expression of IKKα, IKKβ and S100B transcripts in transfected cells was evaluated by RT--PCR or western blotting. Reverse transcriptase--PCR and real--time PCR {#s4p} --------------------------------------------- Real--time RT--PCR was performed using the iCycler iQ detection system (Bio--Rad) and SYBR Green chemistry (Finnzymes Oy, Espoo, Finland). Cells were lysed and total RNA was extracted using RNeasy Mini Kit (QIAGEN, Milan, Italy) and was reverse transcribed with Sensiscript Reverse Transcriptase (QIAGEN) according to the manufacturer\'s directions. The sense/antisense primers were as follows: *Ager* sense 5′--GCCCTCATTGATGTCTTCCACC--3′; antisense (5′--GAACTCATGGCAGGCCGTGGTC--3′); *s100b* sense 5′--GCCCTCATTGATGTCTTCCACC--3′; antisense 5′--GAACTCATGGCAGGCCGTGGTC--3′; *s100a8* sense 5′--TCGTGACAATGCCGTCTGAACTG--3′; antisense 5′--TGCTACTCCTTGTGGCTGTCTTTG--3′; *s100a9*, sense 5′-- CGCAGCATAACCACCATCATC--3′; antisense 5′--GCCATCAGCATCATACACTCC--3′; *Hmgb1* sense, 5′--GGCTGACAAGGCTCGTTATG--3′; antisense 5′--GCAACATCACCAATGGATAAGC--3′;*Fas* sense 5′--CTACTGCGATTCTCCTGGCTGTG--3′; antisense 5′--AGTTTGTATTGCTGGTTGCCTGTGC--3′; *Bcl2* sense 5′--ACGAGTGGGATGCTGGAGATG--3′; antisense 5′--TCAGGCTGGAAGGAGAGATGC--3′. Other primers were as described [@ppat.1001315-Bonifazi1] Amplification efficiencies were validated and normalized against *Gapdh*. The thermal profile for SYBR Green real time PCR was at 95°C for 3 min, followed by 40 cycles of denaturation for 30 s at 95°C and an annealing/extension step of 30 sec at 60°C. Each data point was examined for integrity by analysis of the amplification plot. The mRNA--normalized data were expressed as relative cytokine mRNA in stimulated cells compared to that of mock--infected cells. Statistical analysis {#s4q} -------------------- Data were analyzed by GraphPad Prism 4.03 program (GraphPad Software, San Diego, CA). Student\'s *t* test or analysis of variance (ANOVA) and Bonferroni\'s test were used to determine the statistical significance (*P*) of differences in organ clearance and in vitro assays. The data reported are either from one representative experiment out of three to five independent experiments (western blotting and RT--PCR) or pooled from three to five experiments, otherwise. The in vivo groups consisted of 6--8 mice/group. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### RAGE and DAMPs expression in pulmonary aspergillosis. (A) Expression of Ager, hmgb1, s100b, s100a8 and s100a9 by real time RT-PCR on lung of C57BL6 mice at different days postinfection (dpi) with Aspergillus conidia intranasally. Representative of 2 experiments. (B) S100B-expression on purified cells from transgenic mice expressing s100b-EGFP+ infected with Aspergillus conidia 3 days before. M∅ alveolar macrophages, DC, dendritic cells, PMN, polymorphonuclear neutrophils. (C) Expression of Ager by RT-PCR on purified lung cells from uninfected C57BL6 mice. Microscopy was performed on a DM Rb epifluorescence microscope equipped with a digital camera. Representative of 2 experiments. (0.41 MB TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### RAGE-deficient mice develop pathogen-induced Th inflammation. (A) Purified DCs from uninfected mice were exposed to live resting conidia or hyphae as described for 18 h before real time RT-PCR. (B) Cytokine gene-expression by real-time RT-PCR in DCs from RAGE KO or WT uninfected mice exposed to MALP-2, LPS or ODN-CpG for 18 h. (C) Freshly isolated CD4+T cells from TLN were assessed for transcription factor expression by RT-PCR. P, KO vs WT mice. P, KO vs WT mice. Data are pooled from 4 experiments or representative of 2 experiments (for histology). Representative of 2 experiments. P, KO vs WT DCs. N.D., not determined. (0.57 MB TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S3 ::: {.caption} ###### Effects of HMGB1 administration in mice with aspergillosis. Fungal growth (CFU±SE) (A) and lung histology (PAS staining) (B) in C57BL6 or RAGE KO mice infected with Aspergillus live conidia intranasally and treated intraperitoneally for 3 consecutive days with 50 µg/kg HMGB1. P, treated vs untreated (-) mice. Representative of 3 experiments. \*P\<0.05, treated vs untreated (-) cells. (0.65 MB TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S4 ::: {.caption} ###### Effect of S100B on A. fumigatus morphology and germination. Germination refers to the percentages (mean ± SE) of germinating cells over a total of 400 cells counted. Magnification x 40. Shown are the pooled results from 2 experiments. Photographs were taken using a high Resolution Microscopy Color Camera AxioCam, using the AxioVision Software Rel. 3.1 (Carl Zeiss S.p.A., Milano, Italy). (1.25 MB TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Text S1 ::: {.caption} ###### Supplemental methods referred to Supplemental [Figures S1](#ppat.1001315.s001){ref-type="supplementary-material"}, [S3](#ppat.1001315.s003){ref-type="supplementary-material"}, [S4](#ppat.1001315.s004){ref-type="supplementary-material"}. (0.04 MB DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: We thank Dr. Cristina Massi Benedetti for digital art and editing. The authors have declared that no competing interests exist. The studies were supported by the Specific Targeted Research Project \'ALLFUN\' (FP7--HEALTH--2009 Contract number 260338 to LR), \"SYBARIS\" (FP7--HEALTH--2009 Contract number 242220 to LR), the Italian Grant Application 2010 Fondazione per la ricerca sulla Fibrosi Cistica (Research Project number FFC\#21/2010 to LR) and by the Italian Projects of the Fondazione Cassa di Risparmio di Perugia (2007.0218.020 and 2009.020.0021 to RD). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: GS FB RD LR. Performed the experiments: GG FR PB TZ SZ. Analyzed the data: GS RD LR. Wrote the paper: GS RD LR.

PubMed Central

2024-06-05T04:04:19.697272

2011-3-10

PMC3053349

Introduction {#s1} ============ *Staphylococcus aureus* belongs to the normal human flora. About one in three healthy individuals are colonized asymptomatically with *S. aureus* in the nostrils without any associated disease. However, *S. aureus* is also a leading cause of hospital- and community-acquired infections worldwide [@ppat.1002006-Deleo1]. This potent Gram-positive pathogen can grow in any part of the human body, and also propagates in other animals. The severity and locations of infections vary widely, from minor skin infections to deep-seated infections such as endocarditis, bone and joint infections, or severe pneumonia. Concern about *S. aureu*s infections is heightened because of the emergence and spread of hypervirulent, drug-resistant, and community-acquired strains [@ppat.1002006-Chambers1]. The pathogenesis of *S. aureus* is intricate and relies on an arsenal of virulence-associated factors including toxins, adhesins, enzymes, and immune-modulators [@ppat.1002006-Plata1]. These proteins are delivered in a coordinated manner by sophisticated regulatory networks. To this end, multiple *trans*-acting modulators, including regulatory proteins, secondary metabolites, small peptides, and RNAs, are brought into play [@ppat.1002006-Novick1], [@ppat.1002006-Wyatt1]. The universality of small, usually non-coding, RNAs (sRNAs) playing a role in gene regulation in bacteria is now well established [@ppat.1002006-Vogel1], [@ppat.1002006-Waters1]. The number of sRNA identified in bacteria has considerably increased this past decade [@ppat.1002006-Liu1]. Most of them exert regulatory functions by interacting with proteins and by pairing with mRNAs. Besides these *trans*-acting sRNAs, a variety of large mRNA leaders sense environmental cues or intracellular concentrations of small metabolites to adopt structures that prevent/activate their extended transcription or translation. Examples of sRNA-dependent regulations are given in [Figure 1](#ppat-1002006-g001){ref-type="fig"}. Recent studies on various bacteria indicated that pervasive transcription generates massive antisense transcription [@ppat.1002006-ToledoArana1]--[@ppat.1002006-Dornenburg1]. All these sRNAs are members of regulatory circuits involved in metabolism, stress adaptation, and virulence. Although still a recent field, the study of sRNAs has already extended our knowledge of regulatory circuits in bacteria, in relation to pathogenesis. In *S. aureus*, the multifunctional regulatory RNA, called RNAIII, is a paradigm for RNA-mediated regulation of virulence genes [@ppat.1002006-Novick2], [@ppat.1002006-Boisset1]. It is the effector of the accessory gene regulator (*agr*) system [@ppat.1002006-Novick1], which controls the switch between the expression of surface proteins and excreted toxins. Within the last few years, several reports highlighted the importance and diversity of staphylococcal sRNAs [@ppat.1002006-Pichon1]--[@ppat.1002006-Beaume1]. This review focuses on *S. aureus* regulatory RNAs including RNAIII, newly discovered island-encoded sRNAs, *cis*-encoded antisense RNAs (asRNA), and *cis*-acting regulatory regions of mRNAs. For all these RNAs, their structural diversity and phylogenetic distribution is documented and discussed, with emphasis on those for which their targets were identified and regulatory mechanisms elucidated. Some of these sRNAs have been demonstrated as tractable targets for compounds inhibiting *S. aureus* pathogenesis. ::: {#ppat-1002006-g001 .fig} 10.1371/journal.ppat.1002006.g001 Figure 1 ::: {.caption} ###### General mechanisms given for several *S. aureus* regulatory RNAs. \(A) S-adenosyl methionine (SAM) riboswitch regulates several operons encoding enzymes and transporter proteins. SAM binds to an aptamer domain and stabilizes the formation of a terminator hairpin (the alternate pairings are in red) to arrest transcription [@ppat.1002006-Blount1]. (B) Schematic representation of a T-Box involved in the regulation of aminoacyl-tRNA synthetases (aaRS). Non aminoacylated tRNA binds to the leader region at two different sites: the anticodon sequence of the tRNA base paired with a codon-like triplet present in the "specifier loop", and the ACCA end of the tRNA binds to a complementary sequence located in the T-Box motif [@ppat.1002006-GutierrezPreciado1]. This interaction stabilizes an anti-terminator structure allowing transcription of the downstream genes. \*The aminoacyl-tRNA synthetases regulated by this mechanism are the ValRS, MetRS, IleRS, PheRS, GlyRS, SerRS, HisRS, and the AspRS. (C) The SprD pathogenicity island RNA (in red) binds to the ribosome binding site (The SD is green) of *sbi* mRNA to repress translation initiation [@ppat.1002006-Chabelskaya1]. (D) The RsaOX (in red) (or Teg14as) *cis*-acting asRNA [@ppat.1002006-Bohn1], [@ppat.1002006-Beaume1] is complementary to the coding sequence of *tnp* mRNA and is predicted to induce rapid degradation of *tnp* mRNA. Both the asRNA and the mRNA target site are highly folded, suggesting that the pairing is initiated by a loop--loop interaction. :::  ::: Diversity of sRNAs Expressed from the *S. aureus* Genome {#s2} ======================================================== The complex structure of RNAIII, the first sRNA reported in *S. aureus*, and the intriguing pleiotropic phenotypes associated with its inactivation, led to the proposal and subsequent demonstration that RNAIII was a regulatory RNA [@ppat.1002006-Novick2] (cf. below). In 2001, with the exception of tmRNA, the sRNAs were ignored from the analysis of the first staphylococcal genome sequences [@ppat.1002006-Kuroda1]. As sRNAs initially emerged as major regulators for bacterial physiology in *Escherichia coli*, several laboratories engaged in a quest to identify sRNAs in various *S. aureus* strains. In 2005, Pichon and Felden demonstrated for the first time the existence of sRNAs produced by horizontally acquired genomic islands by identifying seven sRNAs encoded on pathogenicity islands (PIs) in *S. aureus* [@ppat.1002006-Pichon1]. Recently, several publications on this bacterium have contributed to an impressive catalog of putative and experimentally validated sRNAs that place *S. aureus* as a new model organism for sRNA studies. Approaches for identifications were based on dedicated computing software [@ppat.1002006-Pichon1], [@ppat.1002006-Geissmann1], [@ppat.1002006-Marchais1], Affymetrix microarrays [@ppat.1002006-Anderson1], [@ppat.1002006-Roberts1], conventional cloning/sequencing of small sized cDNAs [@ppat.1002006-AbuQatouseh1], and 454 [@ppat.1002006-Bohn1] and Illumina [@ppat.1002006-Beaume1] high throughput sequencing (HTS). The sRNA genes are located randomly in the core genome and mobile accessory elements, and some of them are present in multiple copies. Besides the housekeeping RNAs (such as 4.5S, RNase P, and tmRNA), 6S RNA, and *cis*-acting regulatory sequences, conservation of most sRNAs is restricted to the genus *Staphylococcus*, and among them, about 50% are found so far only within the *S. aureus* species. Approximately 100 *trans*-encoded sRNAs, 100 *cis*-encoded asRNAs, and more than 30 *cis*-acting regulatory regions of mRNAs (e.g., riboswitch, T-Box, protein-binding motif) were discovered to be encoded on the *S. aureus* chromosome, and nine sRNAs on the pN315 plasmid. The expression of more than 90 of these was confirmed by alternative methods such as northern blots, RNA extremity mapping, or RT-qPCR ([Table S1](#ppat.1002006.s001){ref-type="supplementary-material"}). The HTS study performed by Beaume et al. [@ppat.1002006-Beaume1] confirmed almost all sRNAs from other studies [@ppat.1002006-Pichon1], [@ppat.1002006-Geissmann1], [@ppat.1002006-Bohn1], [@ppat.1002006-Marchais1], with the exception of 12 sRNAs that were reported solely by Abu-Qatouseh et al. [@ppat.1002006-AbuQatouseh1]. This singularity might reflect the distance between the unsequenced clinical isolates and the *S. aureus* strains in which sRNAs are mainly studied. This observation may suggest that the sRNA profile is a signature of a given strain; if in the case of N315 we are approaching a full inventory, it is not the case for the other *S. aureus* strains. Numerous *cis*-Encoded Antisense RNAs (asRNAs) {#s2a} ---------------------------------------------- These RNAs pair with an extended perfect match to RNAs expressed from their complementary gene strand ([Figure 1D](#ppat-1002006-g001){ref-type="fig"}). The first one identified in *S. aureus* was shown to control the rolling-circle replication of plasmid pT181 by transcriptional attenuation [@ppat.1002006-Novick3]; the striking discovery of the recent studies is the large proportion of asRNAs among the inventoried sRNAs [@ppat.1002006-Pichon1], [@ppat.1002006-AbuQatouseh1]--[@ppat.1002006-Beaume1]. Many asRNAs are expressed from PIs and mobile elements (plasmids or transposons). Transposable genetic elements are important motors of genetic variability but can also compromise genome integrity. Hence, transposition would expectedly be tightly regulated. The control of transposase synthesis occurs through different mechanisms, one being by asRNAs [@ppat.1002006-Nagy1]. Among them, RsaOX is complementary to the coding sequence of SA0062 mRNA encoding a transposase [@ppat.1002006-Bohn1] ([Figure 1D](#ppat-1002006-g001){ref-type="fig"}). Another interesting case is the control of the IS1181 transposase, which has its gene repeated eight times in the *S. aureus* N315 genome. Two small RNAs, Teg17/RsaOW and Teg24as complementary to the 5′ and 3′ IS1181 UTRs, respectively, were detected. The expression of Teg17/RsaOW is constitutive during growth [@ppat.1002006-Bohn1], and is strongly enhanced in response to pH and temperature changes [@ppat.1002006-Beaume1]. Interestingly, these asRNAs ([Figures 1D](#ppat-1002006-g001){ref-type="fig"} and [2](#ppat-1002006-g002){ref-type="fig"}) show predicted structural similarity to the *E. coli* "RNA-OUT" asRNA, which regulates *tnp* translation of the IS10 insertion element, suggesting that these asRNAs have been tuned for fast binding to mRNAs [@ppat.1002006-Wagner1]--[@ppat.1002006-Ross1]. Some of these asRNAs are surprisingly long; for example, one of them, which is complementary to SA0620 encoding a secretory antigen, SsaA-like, exceeds 1 kb [@ppat.1002006-Beaume1]. AsRNAs may participate in the differential expression of genes belonging to the same operon; this could be the case for two asRNAs that are complementary to *capF* and *capM* mRNA regions of the large *cap* operon mRNA encoding enzyme for capsular polysaccharide synthesis [@ppat.1002006-AbuQatouseh1], [@ppat.1002006-Beaume1]. Several overlapping 3′UTRs of convergent mRNAs were also detected in staphylococci, in which the 3′UTR of one mRNA overlaps the mRNA on the opposite strand, and convergent genes share the same terminator hairpin. Since they pair between each other with extended perfect matches, these 3′UTRs could be categorized as asRNAs. sRNAs that likely issue from the processing of extended 3′ UTRs were also found. How these overlapping and processed UTRs affect gene expression is unknown [@ppat.1002006-Beaume1]. ::: {#ppat-1002006-g002 .fig} 10.1371/journal.ppat.1002006.g002 Figure 2 ::: {.caption} ###### Secondary structures of selected *S. aureus* sRNAs. The secondary structures of RsaA and RsaE [@ppat.1002006-Geissmann1] and of SprD [@ppat.1002006-Chabelskaya1] were experimentally determined. The secondary structure of the antisense RNA RsaOX (or Teg14as) is proposed based on computer predictions [@ppat.1002006-Bohn1], [@ppat.1002006-Beaume1]. The genomic locations and flanking genes of the sRNAs are indicated. The known regulatory domains of the sRNA, which bind to mRNA targets, are in red. The black circled nucleotides are accessible C-rich motifs that are proposed to be crucial for the initial binding with mRNA [@ppat.1002006-Geissmann1]. :::  ::: A Global sRNA Expression Variation Associated with a Phenotype {#s2b} -------------------------------------------------------------- Small colony variant (SCV) isolates exhibit particular properties such as host intracellular persistence and the ability to cause antibiotic-refractory, latent, or recurrent infections [@ppat.1002006-Proctor1]. Several asRNAs and sRNAs showed differential expression in the SCV compared to a wild type strain [@ppat.1002006-AbuQatouseh1]. In addition, SCVs does not express RNAIII [@ppat.1002006-AbuQatouseh1], [@ppat.1002006-Kahl1]. SCVs also repressed an asRNA that presumably regulates expression of PhoB, an alkaline phosphatase involved in inorganic ion transport [@ppat.1002006-AbuQatouseh1]. These variations of regulatory RNA expression may correlate with electron transport deficiencies associated with SCVs [@ppat.1002006-Proctor1]. Hence, perturbation of genetic regulatory circuits and their associated effects on virulence may be a consequence of and/or contribute to the SCV phenotype. RNome-Related *S. aureus* Specificities {#s2c} --------------------------------------- *S. aureus* has a small genome (2.8 M bp) with a low GC composition (32.8%); these features affect its RNome and it is likely that the features of sRNA as learned from studies in enteric bacteria will differ in *S. aureus*. Specifically, in *E. coli* and many other bacteria, RNase E is implicated in the sRNA-dependent mRNA turnover, and the RNA chaperone Hfq is required for the activity of most *trans*-encoded sRNAs and, as shown recently, for a *cis*-encoded asRNA [@ppat.1002006-Ross1]. However, *S. aureus* do not have an RNase E, but instead possesses RNases J1, J2, and Y functional homologs [@ppat.1002006-Anderson2]. In *Bacillus subtilis*, a complex resembling the *E. coli* degradosome, including the three RNases J1/J2/Y, glycolytic enzymes (enolase, phosphofructokinase PfkA) and the RNA helicase CshA were recently reported [@ppat.1002006-Commichau1]--[@ppat.1002006-Shahbabian1]. Homologs of all these enzymes are present in the staphylococcal strains, but whether these enzymes form a "degradosome" remains to be addressed [@ppat.1002006-Anderson2]. Concerning Hfq, this protein is not expressed in several tested *S. aureus* strains [@ppat.1002006-Geisinger1], [@ppat.1002006-Liu2], and the deletion of its corresponding gene was thought to have no effect on bacterial physiology [@ppat.1002006-Bohn2]. However, a recent report indicates that in strains where Hfq is detected, the deletion of its coding gene could result in decreased toxicity and virulence of *S. aureus*, leading to the conclusion that Hfq is a global regulator that controls pathogenicity [@ppat.1002006-Liu2]. In these strains, analyzing more precisely the regulatory functions of Hfq as well as its involvement on the characterized sRNA-dependent regulations would be required. The commitment of Hfq is not straightforward since several strains in which Hfq is not detected produce toxins and are virulent. Moreover, Hfq is not required for the reported cases of sRNA-induced translational repression [@ppat.1002006-Boisset1], [@ppat.1002006-Geissmann1], [@ppat.1002006-Geisinger1], [@ppat.1002006-Chabelskaya1]; in several of these examples, the sRNAs efficiently bind to the mRNA targets and form extended pairings that are specifically cleaved by the double-strand-specific RNase III [@ppat.1002006-Boisset1], [@ppat.1002006-Huntzinger1], [@ppat.1002006-Chevalier1]. The dispensability of Hfq in several *S. aureus* strains could be due, as compared to *E. coli*, to the presence of longer "sRNA--mRNA" hybrids that compensate for low GC content of the pairings [@ppat.1002006-Jousselin1]. General strategies for the use of RNA-dependent regulation by bacteria vary according to species as the result of environmental and evolutionary constraints. For instance, in *B. subtilis* and *S. aureus*, the autocatalytic site-specific cleavage in the 5′ UTR of *glmS* mRNA, encoding glucosamine 6 phosphate synthase, is stimulated by the binding of glucosamine-6-phosphate; in *E. coli* and *Salmonella* the *glmS* gene is regulated by two sRNAs [@ppat.1002006-Winkler1]--[@ppat.1002006-Urban1]. Mechanistic and functional analyses performed on several *S. aureus* sRNAs revealed its RNome specificity by pointing out its differential roles in the regulation of mobile elements, metabolism, stress adaptation, and virulence. Pathogenicity Island--Encoded RNAs {#s3} ================================== Mobile genetic elements have essential roles in genome evolution. In facultative pathogens such as *S. aureus*, they mediate acquisition of antibiotic resistance genes, including the highly problematic methicillin resistance via the staphylococcal chromosome *mec* cassette (SCCmec), and have conferred a wide range of adaptive processes for survival in their hosts. Among these elements, which include phages, genomic islands, transposons, and plasmids, the horizontally acquired PIs are the repository of many toxins, adherence and invasion factors, superantigens, and secretion systems [@ppat.1002006-Novick4], [@ppat.1002006-Novick5]. In addition to the protein-coding genes, several SaPIs including phage-related chromosomal islands encode and express several sRNAs [@ppat.1002006-Pichon1], [@ppat.1002006-AbuQatouseh1]--[@ppat.1002006-Beaume1] ([Table 1](#ppat-1002006-t001){ref-type="table"}). Some sRNAs are present in multiple copies scattered around the *S. aureus* genome (up to eight genomic copies; repeated events of horizontal transfer as well as gene duplications may account for the presence of multiple copies), and additional copies are on plasmids [@ppat.1002006-Pichon1]. The location of sRNAs on SaPIs suggests that these sRNAs would play important roles during *S. aureus* infections. ::: {#ppat-1002006-t001 .table-wrap} 10.1371/journal.ppat.1002006.t001 Table 1 ::: {.caption} ###### RNAs expressed from N315 *S. aureus* PIs. ::: {#ppat-1002006-t001-1} RNAs Flanking Genes Strand Orientation Locations Lengths[a](#nt101){ref-type="table-fn"} Ends Mapping[a](#nt101){ref-type="table-fn"} Exp. Validation References ------------------------------------------------------- --------------------------------------- -------------------- ------------------ ----------------------------------------- ---------------------------------------------- ----------------- ------------------------------ **SaPIn1** Teg21as *Antisens to hypot. protein SA1825* \> 2064507/2064570 ∼63 No RNA sequencing [@ppat.1002006-Beaume1] Teg22as *Antisens to hypot. protein SA1830* \> 2069004/2069067 ∼63 No RNA sequencing [@ppat.1002006-Beaume1] Teg124 *Probable β-lactamase/Enterotoxin C3* \< 2059473/2059365 ∼108 No RNA sequencing [@ppat.1002006-Beaume1] **SaPIn2** Sau-63 *Hypot. protein/hypot. protein* \< 436958--437055 ∼100 No Northern [@ppat.1002006-AbuQatouseh1] **SaPIn3** [sprA1]{.underline} [b](#nt102){ref-type="table-fn"} *Trunc. transposase/transposase* \> 1856223--1856978 208 Yes Northern [@ppat.1002006-Pichon1] [Teg152]{.underline} [b](#nt102){ref-type="table-fn"} *transposase/trunc.transposase* \< 1856712--1856658 ∼54 No RNA sequencing [@ppat.1002006-Beaume1] sprB Probable*β*-lactamase/truncated HP \< 1866661--1867134 114--118 Yes Northern [@ppat.1002006-Beaume1] sprC *Leukotoxin LukE/Hypot. protein* \< 1871167--1872531 170 Yes Northern [@ppat.1002006-Pichon1] **φ (Bacteriophage)** sprD *Hypot. protein/hypot. protein* \< 2006878--2007561 142 Yes Northern [@ppat.1002006-Chabelskaya1] sprX(RsaOR) *Trunc. amidase/staphylokinase* \< 2008572--2009085 147 (processed) Yes Northern [@ppat.1002006-Bohn1] [sprF]{.underline} [b](#nt102){ref-type="table-fn"} *Holin homolog/hypot.protein* \> 2010789--2011001 186 Yes Northern [@ppat.1002006-Pichon1] [sprG]{.underline} [b](#nt102){ref-type="table-fn"} *Hypot. protein/holin homolog* \< 2011001--2010789 300 Yes Northern [@ppat.1002006-Pichon1] a Except for SprD [@ppat.1002006-Chabelskaya1], experimental determinations of the 5′ and 3′ ends were also performed for sprA, sprB, sprC, sprF, sprG, and sprX (unpublished data); the lengths of the RNAs with no ends mapping are approximate; b RNA couples (underlined) predicted to encode for type I "toxin-antitoxin" modules [@ppat.1002006-Beaume1]. RNA candidates Sau-18 and Sau-6079 within SaPIn2, and Sau-6361, Sau-6473, Sau-7007 from φ were predicted but not confirmed by Northern blots [@ppat.1002006-AbuQatouseh1]. Apart from PIs, *S. aureus* RNAs are also expressed from Genomic Islands (Teg31as and Teg147) and from the staphylococcal chromosome cassette "SCC Mec" (Teg4, Teg5as, Teg6as, Teg7as, Teg8as, Teg10as, Teg14as, Teg118 [@ppat.1002006-Beaume1]). ::: Although the sRNAs expressed from SaPIs are expected to regulate expression of genes located on the cognate PI, they could also establish functional links between the PIs and the core genome. An example is provided by the SprD RNA, expressed from PIφ ([Table 1](#ppat-1002006-t001){ref-type="table"}), which represses translation initiation of the *sbi* mRNA encoding an immune-evasion molecule located at a core genomic locus distant from SprD [@ppat.1002006-Chabelskaya1]. A central hairpin from SprD pairs to the *sbi* mRNA ribosome binding sites and prevents translation initiation ([Figures 1C](#ppat-1002006-g001){ref-type="fig"} and [2](#ppat-1002006-g002){ref-type="fig"}). Interestingly, SprD contributes significantly to causing disease in a mouse infection model, although this effect is not only linked to the down-regulation of Sbi production. Moreover, overproducing SprD in vivo is toxic for the cells and reduces bacterial growth (S. Chabelskaya, N. Sayed, and B. Felden, unpublished data), possibly suggesting that SprD targets essential function(s). Since SprD has a significant impact on virulence, it implies possible strategies in controlling staphylococcal infections by modulating SprD expression levels. Additional sRNAs expressed from the PIs might also be involved in *S. aureus* pathogenicity, either directly or via intricate regulatory networks including transcriptional regulatory factors. Among the currently characterized asRNAs expressed in *S. aureus*, four are located in PIs and six in the SCCmec mobile genetic element, all ranging in sizes from 54 to 400 nucleotides [@ppat.1002006-Beaume1]. Most of them have perfect base complementarities with regions of mRNAs encoding hypothetical protein genes, and are likely to act as regulators. Two of these sRNAs, Teg152 and SprF, are fully complementary to SprA1 and SprG sRNAs, respectively ([Table 1](#ppat-1002006-t001){ref-type="table"}): the "SprA1/Teg152" and "SprG/SprF" RNA pairs are predicted to form type I "toxin-antitoxin" modules in which SprA and SprG would encode hydrophobic small peptides [@ppat.1002006-Fozo1]. These modules are found in multiple copies in several *S. aureus* strains, and several copies are expressed (A. Jousselin, M. L. Pinel, and B. Felden, unpublished data). The independent transcriptional activation of the copies allows the production of accurate sRNA levels for precise functions. SprA1 is a multifunctional RNA with putative antisense properties since its 3-end can pair with the 3′-UTRs of three mRNA targets [@ppat.1002006-Pichon1]. Interestingly, SCCmec carries determinants other than antibiotic resistance genes, which confer selective advantages to methicillin-resistant *S. aureus* (MRSA) in the host. An sRNA carrying a small ORF was recently identified within the SCCmec. This ORF encodes a peptide that has pro-inflammatory and cytolytic characteristics typical of phenol-soluble modulins (PSM). The PSM-mec peptide had significant impact on immune evasion and disease, thus revealing a role of methicillin resistance clusters in staphylococcal pathogenesis [@ppat.1002006-Queck1]. The expression of the PSM-sRNA (Teg4) was strongly enhanced in response to oxidative stress [@ppat.1002006-Beaume1]. A Multifunctional RNA Couples Quorum Sensing to Virulence {#s4} ========================================================= RNAIII is the effector of the *agr* system, which functions as a sensor of population density. The complex cascade of events orchestrated by *agr* has been extensively studied (see [@ppat.1002006-Novick1], [@ppat.1002006-Novick2] for review). Briefly, it comprises a density-sensing cassette (*agrD* and *agrB*) and a two-component sensory transduction system (*agrA* and *agrC*) in which autoinducing peptide (AIP), the *agrD* gene product, is continuously released in the extracellular environment. Upon reaching a critical concentration, AIP activates the two-component *agrA*-*agrC* system, which triggers transcription of RNAIII, of its own operon, and of genes encoding metabolic factors and PSM peptides [@ppat.1002006-Novick2], [@ppat.1002006-Queck2]. In this cascade, expression of RNAIII is maximal in the late logarithmic and stationary phase of growth. RNAIII has the unique property of acting both as an mRNA that encodes the 26-aa delta hemolysin (PSM) peptide, and as a critical regulator that represses early virulence factors and activates post-exponentially expressed exotoxins ([Figure 3](#ppat-1002006-g003){ref-type="fig"}). Further genetic, transcriptomic, and proteomic studies revealed that *agr* belongs to a rich and complex network of regulatory genes in which *agr* is both a target and an effector of regulation (reviewed in [@ppat.1002006-Novick1]). As an effector, RNAIII governs not only the expression of key virulence factors including cell wall--associated proteins and exotoxins, but also numerous two-component systems and global regulators (*arl*, *sae*, *srr*, *rot*) and an impressive list of other processes impacting biofilm formation, peptidoglycan and amino acid metabolism, and transport pathways [@ppat.1002006-Dunman1]--[@ppat.1002006-Jones1]. These effects are quantitatively but not qualitatively variable depending on the staphylococcal strain. For instance, the effect of *a*gr inactivation was more marked on the transcriptome of NCTC 8325 derivatives as compared to the UAMS-1 strain [@ppat.1002006-Cassat1]. The question as to whether these effects result from direct or indirect mechanisms has been only solved for a limited number of genes and benefited from the experimental determination of RNAIII structure [@ppat.1002006-Benito1]. RNAIII, a highly stable molecule (half-life \>45 min) [@ppat.1002006-Huntzinger1] is characterized by 14 stem-loop structures and two long helices formed by long-range base pairings that close off independent structural domains [@ppat.1002006-Benito1]. Specific domains of RNAIII control the expression of different targets ([Figure 3](#ppat-1002006-g003){ref-type="fig"}). The 5′ end of RNAIII positively regulates *hla* translation (encoding alpha hemolysin) by competing directly with an intramolecular RNA secondary structure that sequesters the *hla* ribosomal binding site (RBS) ([@ppat.1002006-Novick6], [@ppat.1002006-Morfeldt1]; [Figure 3A](#ppat-1002006-g003){ref-type="fig"}). The RNAIII hairpin H13 and terminator hairpin H14 of the 3′ domain, and hairpin H7 of the central domain, act separately or coordinately to repress the synthesis of early expressed virulence factors (i.e., coagulase, protein A, and the repressor of toxins, Rot) at the post-transcriptional level by a conserved mechanism with slight variations ([Figure 3C](#ppat-1002006-g003){ref-type="fig"}). The common theme is that RNAIII functions as an asRNA that anneals to target mRNAs, and the formed complexes result in the repression of translation initiation and in rapid mRNA degradation triggered by RNase III. Structures of the complexes depend on the target mRNA, and may comprise i) an extended duplex between RNAIII and the RBS of mRNAs (e.g., *spa*, the peptidoglycan hydrolase LytM, and *SA1000* encoding a fibrinogen-binding protein), or ii) an imperfect duplex that sequesters the RBS completed by a loop--loop interaction in the coding region (for *coa* encoding coagulase), or two loop--loop interactions, one involving the 5′UTR and the other the RBS (for *rot* mRNA) ([@ppat.1002006-Boisset1], [@ppat.1002006-Chevalier1]; [Figure 3](#ppat-1002006-g003){ref-type="fig"}). In these three cases, a single loop--loop interaction is not sufficient for efficient repression, thus limiting the capacity of RNAIII to act as a repressor to the mRNA targets that not only possess a Shine and Dalgarno (SD) sequence complementary to H7, H13, or H14 of RNAIII, but that also display an additional region of interaction or the capacity to form an extended duplex. As discussed above, the RNA-binding protein Hfq, which is an important RNA chaperone in several species [@ppat.1002006-Chao1], is not involved in the RNAIII-dependent regulatory processes, although Hfq binds to RNAIII in vitro [@ppat.1002006-Huntzinger1]. ::: {#ppat-1002006-g003 .fig} 10.1371/journal.ppat.1002006.g003 Figure 3 ::: {.caption} ###### The functional RNAIII structure and its mRNA targets. \(A) Schematic view of the RNAIII-mediated antisense activation mechanism. Hairpin loops H2 and H3 of RNAIII (red) bind to the *hla* mRNA (black) to activate translation initiation. The ribosomal 30S subunit is schematized. The SD sequence is green. (B) RNAIII secondary structure (adapted from [@ppat.1002006-Benito1]) and its genomic location within the *agr* locus (bottom). RNAIII encodes the delta-hemolysin (*hld*, in green). The 5′UTR activates alpha-hemolysin translation [@ppat.1002006-Morfeldt1] and the 3′ domain represses the translation of virulence factors and of the transcriptional repressor of toxins (*rot*) [@ppat.1002006-Boisset1], [@ppat.1002006-Geisinger1]. The conserved C-rich sequences detected in H7, H13, and H14 is indicated. (C) Schematic views of the RNAIII-mediated antisense translation initiation repression mechanisms. RNAIII structural domains H7, H13, and H14 (in red) are involved in interactions with target mRNAs (in black). The AUG codon and SD sequence are in green. The broken black arrows are the RNase III--induced cleavages. :::  ::: With the exception of *hla* translational activation, all the direct effects of RNAIII lead to repression of mRNA targets. However, as Rot is a transcriptional regulatory protein, its repression by RNAIII results in the indirect transcriptional regulation of many genes, including activation of alpha-toxin and repression of protein A, both of which are also directly regulated by RNAIII [@ppat.1002006-SaidSalim1]. These complex regulatory circuits involve several feed-forward loops ([Figure 4](#ppat-1002006-g004){ref-type="fig"}) that regulate expression via RNAIII and Rot at both the transcriptional and post-transcriptional levels. For repression, these double controls prevent leakage at the transcriptional level, which could be particularly suitable for stable mRNAs like *spa*. Therefore, the involvement of RNAIII in such regulatory circuits not only guarantees tight regulation but also might ensure fast recovery after the external stimulus (quorum sensing) is over [@ppat.1002006-Shimoni1]. Hence, a number of genes---and the list will likely grow---are regulated by RNAIII at multiple levels (indirectly on promoter activity, directly on translation and mRNA degradation), suggesting that the amount and timing of production of certain virulence factors is precisely controlled during the course of infection. The in vivo requirement for such strict regulation of virulence protein expression is particularly plausible in the case of the protein A, which harbors multiple functions from anti-opsonic activity to the induction of tumor necrosis factor receptor 1 and B cell superantigenic properties [@ppat.1002006-Gomez1], [@ppat.1002006-Goodyear1]. ::: {#ppat-1002006-g004 .fig} 10.1371/journal.ppat.1002006.g004 Figure 4 ::: {.caption} ###### Integrating the *S. aureus* sRNAs into gene regulation cascades. The "*agr*-RNAIII" auto-activation circuit is indicated with the two feed-forward loops involving RNAIII. When reaching optimal density, the autoinducing peptide (AIP) activates the *agr* autocatalytic circuit, leading to RNAIII transcription. RNAIII represses the expression of *rot*, which activates *spa* transcription and represses that of *hla*. In the meantime, RNAIII also activates *hla* mRNA translation and represses *spa* mRNA translation. The plain and broken lines indicate the direct or indirect gene activations, respectively. The red lines indicate the down-regulations mediated by the various RNAs. The black question marks above the "see-sawing" triangles point to the unknown triggering factors. The transcriptional regulatory proteins are in blue. The complexity of this scheme will certainly increase as we learn more on the sRNA functions. :::  ::: The importance of *agr* for *S. aureus* pathogenesis is the subject of an apparent paradox. In contrast with most other staphylococcal sRNAs, which have been found by bioinformatic approaches or deep sequencing, RNAIII was first identified in a transposon mutagenesis that revealed pleiotropic effects of a single-site insertion [@ppat.1002006-Morfeldt2], [@ppat.1002006-Recsei1]. Because of its impact on virulence factors, the *agr* system and its effector molecule RNAIII were thought to be of major importance for virulence. Indeed, the majorities of clinical isolates from acute infections have a functional *agr* system and produce RNAIII both in vitro and in vivo [@ppat.1002006-Traber1]. However, *agr*-defective mutant strains, which arose during infection, were isolated from patients [@ppat.1002006-Traber1]. Some of these strains have been associated with persistent bacteremia, notably in patients with intravascular devices and with reduced susceptibility to glycopeptides [@ppat.1002006-Sakoulas1], [@ppat.1002006-Fowler1]. Agr defects are also detected in colonizing isolates of patients, and a mixture of *agr*-positive and *agr*-defective *S. aureus* strains were described in healthy humans [@ppat.1002006-Shopsin1]. This supports the model of *agr* being important for full expression of virulence, notably during acute infection, whilst *agr* mutants would be positively selected in chronic infections and dormant states. However, the observation that RNAIII is also present in all staphylococcal species, including *S. epidermidis*, which is typically a nosocomial pathogen, highlights the fact that RNAIII is not solely a determinant of acute virulence but harbors various functions depending on the bacterial species background in which it evolved. Interestingly, the most conserved domain of RNAIII among staphylococcal species is the 3′ domain (H13 and H14), which is involved in the regulation of several *S. aureus*--specific virulence factors (see above). Its conservation in *S. epidermidis, S. lugdunensis*, and other species suggests that some of the target genes, such as those involved in peptidoglycan metabolism, require the presence of this regulatory domain [@ppat.1002006-Benito1], [@ppat.1002006-Tegmark1]. *trans*-Acting RNAs in Stress Response and Metabolism {#s5} ===================================================== Ongoing functional characterization of *S. aureus* sRNAs links them to various environmental and stress-related responses like pH and temperature variations, nutrient starvation, oxidative stress, and quorum sensing, all of which can be encountered during host infection [@ppat.1002006-Geissmann1]--[@ppat.1002006-Beaume1], [@ppat.1002006-Anderson1]. Such environmental stresses and growth conditions largely influence the toxin synthesis [@ppat.1002006-Somerville1] and require several global transcriptional regulators, such as the alternative sigma B factor (σ^B^). The σ^B^ regulon consists of numerous genes involved in metabolism, stress-related responses, membrane transport system, biofilm formation, antibiotic resistance, and virulence [@ppat.1002006-Bischoff1], [@ppat.1002006-Ziebandt1]. Among these genes, several were repressed by σ^B^ via an indirect mechanism most probably involving a σ^B^-induced regulatory protein or sRNA. Along those lines, recent studies showed that the expression of several sRNAs was induced by σ^B^ [@ppat.1002006-Geissmann1], [@ppat.1002006-Christiansen1]; among them, RsaA, which has a typical σ^B^-promoter detected upstream of its corresponding gene [@ppat.1002006-Geissmann1]. RsaA, conserved among staphylococci, can potentially base pair with mRNAs repressed by σ^B^ like *citM* encoding an Mg-citrate transporter ([Figure 2](#ppat-1002006-g002){ref-type="fig"}; [@ppat.1002006-Bischoff1]). Prediction of σ^B^-promoter within intergenic regions of the *S. aureus* genome suggests the existence of additional σ^B^-dependent sRNAs, which awaits experimental validation [@ppat.1002006-Geissmann1], [@ppat.1002006-Beaume1]. Three σ^B^-dependent sRNAs that are highly conserved in *S. aureus* have been recently described [@ppat.1002006-Nielsen1], and two of them are predicted to encode small highly basic peptides [@ppat.1002006-Beaume1], [@ppat.1002006-Nielsen1]. Most of the newly identified sRNAs are conserved among *S. aureus* clinical isolates or are expressed in various staphylococcal strains [@ppat.1002006-Pichon1], [@ppat.1002006-Geissmann1], [@ppat.1002006-Bohn1], [@ppat.1002006-Beaume1]. One exception is the sRNA RsaE for which the sequence and structure have been found strictly conserved in the *Staphylococcus*, *Macrococcus*, and *Bacillus* genera, all of which share a common Gram-positive ancestor ([Figure 2](#ppat-1002006-g002){ref-type="fig"}; [@ppat.1002006-Geissmann1]). The overproduction of RsaE causes a growth defect that is partially alleviated by the non-preferred carbon source, acetate, suggesting that RsaE accumulation alters essential metabolic functions [@ppat.1002006-Bohn1]. Using comparative transcriptomic and proteomic analysis, RsaE was shown to co-regulate the synthesis of several metabolic pathways involved in amino acid and peptide transport (*opp-3* operon), cofactor synthesis, folate-dependent one-carbon metabolism, lipid and carbohydrate metabolism, and the tricarboxylic acid (TCA) cycle [@ppat.1002006-Geissmann1], [@ppat.1002006-Bohn1]. Like RNAIII, a conserved and unpaired C-rich motif within RsaE pairs with the SD sequence of several target mRNAs, including *opp-3B/opp-3A* (amino acid and peptide transporter), *sucC* (succinyl-CoA synthetase of the TCA cycle), and SA0873 (unknown function), all preventing ribosomal initiation complex formation ([Figure 2](#ppat-1002006-g002){ref-type="fig"}; [@ppat.1002006-Geissmann1], [@ppat.1002006-Bohn1]). Thus, RsaE would coordinate down-regulation of energy metabolism (via the TCA cycle) and purine biosynthesis when carbon sources become scarce, facilitating adaptation to the entry into stationary phase ([Figure 4](#ppat-1002006-g004){ref-type="fig"}). The TCA cycle is integrally involved in the regulation of virulence factor synthesis, biofilm formation and antibiotic resistance (for a review, see [@ppat.1002006-Somerville1]). Other sRNAs are probably involved in metabolic regulation ([Figure 4](#ppat-1002006-g004){ref-type="fig"}). For instance, carbohydrate-dependent repression and oxygen availability correlate with altered expression of RNAIII [@ppat.1002006-Somerville1]. Although no iron-dependent sRNA was so far identified in *S. aureus* as in Gram-negative bacteria [@ppat.1002006-Shopsin1], *S. aureus* sense the alteration of iron status via the ferric uptake regulator (Fur), initiating a regulatory program that modifies expression of a large number of virulence factors [@ppat.1002006-Torres1]. The ongoing functional analysis of the *S. aureus* sRNAs will provide a clearer picture of the links between sRNAs, metabolism, stress adaptation, and virulence programming. RNAs as Antimicrobial Drug Targets {#s6} ================================== The continued evolution of anti-microbial resistance in hospitals and the emergence of community-associated MRSA strains are major threats to patient care. Current antibiotic drugs target a narrow spectrum of bacterial functions including peptidoglycan biogenesis, DNA replication, and protein synthesis. Therefore, there is a growing need for selecting new drugs that target other cellular pathways that should, in principle, result in a weaker selective pressure for the appearance of antibiotic resistance, and that can preserve the host endogenous microbiome. As alternative strategies that affect bacterial viability, anti-virulence strategies have been developed to target mechanisms leading to successful infections, such as virulence factors causing host damage and disease [@ppat.1002006-Lynch1]. Among all the antibiotics currently used to treat clinical infections, more than half bind to the ribosomal RNAs [@ppat.1002006-Wilson1]. Their success as antibacterial targets encourages the development of new antibacterial drugs based on regulatory structured sRNAs. Metabolite-sensing mRNAs, the so-called riboswitches, have been recently exploited as drug targets since they have evolved structured receptors to bind small metabolites with high selectivity and to control the expression of downstream essential genes [@ppat.1002006-Blount1]. Riboswitches are located in the 5′UTRs of some mRNAs and exhibit a structured receptor domain specifically recognized by a small compound. Metabolite binding induces a conformational change of the downstream mRNA that provokes either premature transcription, translation repression, or RNA degradation. In *S. aureus*, seven operons and 33 genes are under the control of riboswitches that respond to the intracellular concentration of S-adenosylmethionine (SAM) ([Figure 1A](#ppat-1002006-g001){ref-type="fig"}), thiamine pyrophosphate (TPP), flavin mononucleotide (FMN), lysine, glycine, guanine, 7-aminomethyl-7-deazaguanine (preQ1), and glucosamine-6-phosphate (Glc-6P) [@ppat.1002006-Geissmann1], [@ppat.1002006-Bohn1], [@ppat.1002006-Beaume1], [@ppat.1002006-Marchais1], [@ppat.1002006-Blount1]. Any agonistic molecule targeting one of these riboswitches would likely impact gene regulation even if cells are devoid of the natural metabolite. As a proof of principle, and based on the crystal structure of the guanine receptor binding domain [@ppat.1002006-Serganov1], [@ppat.1002006-Batey1], several rationally designed guanine analogues that bind the purine riboswitch with affinities comparable to that of the natural ligand were shown to inhibit *B. subtilis* growth [@ppat.1002006-Kim1]. In *S. aureus*, the guanine riboswitch regulates expression of the operon, including *xpt*, *pbuX*, *guaB*, and *guaA*. Using the same strategy as Kim et al. [@ppat.1002006-Kim1], a novel pyrimidine derivative, 2,5,6-triaminopyrimidin-4-one (PC1), was designed to bind the purine-sensing riboswitch to repress the downstream genes [@ppat.1002006-Mulhbacher1]. For the first time, this work shows that PC1 has a selective bactericidal activity restricted to a sub-group of bacteria including *S. aureus*, which contains *guaA* under the control of the purine riboswitch. Although the GMP synthase GuaA is not essential for growth in rich media, the enzyme is nevertheless an important contributor to *S. aureus* survival during infection [@ppat.1002006-Mulhbacher1]. The administration of PC1 significantly reduced *S. aureus* infection in a murine model [@ppat.1002006-Mulhbacher1]. The narrow spectrum of bactericidal activity of PC1 also has the advantage of reducing the selective pressure for resistance. This work and the fact that *S. aureus* contains other types of riboswitches offer novel opportunities for the design of drugs that inhibit the function of structured regulatory RNAs. The increasing rate of *S. aureus* sRNA discovery, together with the intensified search for their mechanisms of action, should pave the way to exploit chemical strategies to interfere with sRNA functions and to fight against bacterial infections in a more specific way. Concluding Remarks {#s7} ================== This review provides a first hint at sRNA functions in *S. aureus* and shows that we are just beginning to fully appreciate their roles in gene regulation. The combined use of high throughput genomic methods and phenotypic analyses of *S. aureus* strains mutated for the sRNA genes, regulatory proteins, ribonucleases, and RNA-binding proteins will generate knowledge on how the regulatory RNAs and proteins are integrated into intertwined regulatory networks in stress adaptation and virulence ([Figure 4](#ppat-1002006-g004){ref-type="fig"}). However, complications are expected due to the substantial genetic variability between *S. aureus* strains, which express a subset of regulatory RNAs, or unique RNAs, and are thus far from universal [@ppat.1002006-Pichon1], [@ppat.1002006-Geissmann1]. Future research is also necessary to identify the signals that regulate sRNA transcription and the mechanisms by which sRNAs act on their targets. To date, most identified mechanisms have involved *trans*-acting sRNAs that bind to the RBS of mRNA targets, and only SprA was predicted to bind to the 3′UTRs. Binding to the coding sequence has not yet been observed. While scientific interest has been mainly focused on antisense regulation, regulatory RNAs are also expected to target proteins. For instance, direct interaction of *S. aureus* 6S RNA with the polymerase bound to σ^A^ and its implication in virulence needs to be analyzed. Multifunctional RNAs, like RNAIII, are most probably the rule rather than the exception, and this field is at present completely unexplored. We also need to consider other unexpected possibilities such as RNA-activating virulence factors, or bacterial sRNAs targeting host genes. Recent analysis of the MRSA operon also shows that many mRNAs have long UTRs, more frequently found at the 3′ ends [@ppat.1002006-tenBroekeSmits1]. These regions might have implications in regulation by promoting specific binding sites for *trans*-acting ligands or by their processing to generate sRNAs [@ppat.1002006-Beaume1]. Mechanisms of RNA processing and turnover are not well studied in *S. aureus* and little is known about *S. aureus* RNA-binding proteins associated with sRNAs. This review illustrates the great diversity in sizes, structures, and mechanisms of sRNAs, and shows that determinants required for regulation could sometimes be predicted from the RNA structure. For instance, several sRNAs and RNAIII carry a C-rich motif, located in hairpin loops or in accessible single strands, which is a specific recognition signature to target the mRNA RBS [@ppat.1002006-Boisset1], [@ppat.1002006-Geissmann1]. Determination of the structures of regulatory complexes has paved the way to identify novel drugs that could interfere with RNA functions [@ppat.1002006-Mulhbacher1]. Also, the significant contribution of RNAIII and SprD to cause diseases in animal models of infection implies that these RNAs could be promising drug targets. Another aspect, which might be important for virulence and host adaptation, would be to consider the cell differentiation within a population, as well as cell-to-cell communication. The expression of individual sRNAs might be variable within a population and these differences could confer the ability of a bacterial subpopulation to respond to stress or environmental changes. This is particularly true for biofilm formation. Furthermore, we have no idea how the host "microbiome" influences sRNA expression and their regulatory networks within *S. aureus*, and vice versa. Metagenomics and deep sequencing could address these questions and would contribute to an understanding of how commensal bacteria can cause diseases. The recently identified *S. aureus* RNome (listed in [Table S1](#ppat.1002006.s001){ref-type="supplementary-material"}) reveals additional layers of complexity to gene regulation mechanisms. sRNA-based regulation increases the number of possible regulatory sites and expectedly provides several advantages compared to protein-based regulation [@ppat.1002006-Beisel1]. Since many sRNAs act at transcription termination or when translation starts, fast and efficient responses on protein levels can be achieved. Furthermore, it is also easier to control RNA turnover when compared to protein degradation. To date, the known *S. aureus* regulatory RNAs provide functional links between metabolism, quorum sensing, and virulence ([Figure 4](#ppat-1002006-g004){ref-type="fig"}). As we learn more about sRNA functions, we expect to find more sRNAs involved in *S. aureus* pathogenesis as well as other connections between virulence and housekeeping networks. Supporting Information {#s8} ====================== Table S1 ::: {.caption} ###### Compilation of 91 expressed RNAs forming the *S. aureus* RNome. It includes 18 *cis*-acting regulators including riboswitches (green), 7 *cis*-encoded sRNAs (purple) 47 *trans*-encoded RNAs (pink), 10 repeated sequences (yellow), and 9 potential 5′UTRs, 3′UTRs, or small CDSs. Both the gene and RNA names, their experimental confirmations, strand expression, genomic locations, predicted lengths, flanking and/or antisense genes, expression profiles, and conservations are indicated for each RNA with specific comments about their functions, when available. (XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: We are thankful to all the members of our respective labs and collaborators for their scientific and intellectual input as well as for the stimulating discussions. We are thankful to Sandy Gruss for critical reading of the manuscript and helpful comments. The authors have declared that no competing interests exist. This work was supported by the Institut National de la Santé et de la Recherche Médicale (INSERM; BF & FV), the Centre National de la Recherche Scientifique (CNRS; PB & PR), the Agence Nationale pour la Recherche (ANR09-BLAN-0024-01: PR & FV; ANR10-BLAN-Duplex-Omic: PB, and ANR-09-MIEN-030: BF), the Région Bretagne (BF), and the European Community (ANR ARMSA Era-Net Pathogenomics, PR & FV; CONCORD-F3-2008-222718: FV). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

PubMed Central

2024-06-05T04:04:19.701618

2011-3-10

PMC3053350

Introduction {#s1} ============ Host cell invasion by apicomplexan parasites is a highly specialized feature of this phylum that is crucial to their survival in humans (*Toxoplasma gondii, Plasmodium spp.*, *Cryptosporidium spp.*) and animals (*Neospora caninum*, *Eimeria spp*.) [@ppat.1002007-Baum1], [@ppat.1002007-Cowman1]. Following initial contact with a target host cell via GPI-anchored parasite surface proteins, the pathogen reorients its apical end toward the host cell membrane and secretes the contents of microneme, rhoptry, and dense granule secretory organelles in an intricate, coordinated manner to enable parasite invasion and concomitant formation of a unique vacuole within the host [@ppat.1002007-Carruthers1]. A key stage in invasion is the formation of a tight apposition between parasite and host cell membranes known as the MJ, first visualized in electron micrographs of invading *Plasmodium* [@ppat.1002007-Aikawa1]. Initially a punctate focus, this interface rapidly resolves into a ring that slides posteriorly over the parasite in conjunction with host membrane invagination and eventual engulfment of the invading pathogen. The MJ is essential for this process, as it anchors the parasite to the host surface while the parasite\'s actin-myosin motor (the "glideosome" [@ppat.1002007-SoldatiFavre1]) provides forward motion into the host cell. In addition, studies of invading *Toxoplasma* parasites using fluorescently labeled lipids and host membrane proteins have demonstrated molecular sieving at the junction, presumably responsible for the non-fusogenic nature of the PV that precludes destruction by host lysosomes [@ppat.1002007-Mordue1], [@ppat.1002007-Charron1]. Although the mechanism by which the MJ carries out these functions is unknown, the identification of a complex of rhoptry neck proteins that specifically localize to the *Toxoplasma* MJ provided a significant advance in the characterization of the unique invasion process used by apicomplexan parasites [@ppat.1002007-Alexander1], [@ppat.1002007-Besteiro1], [@ppat.1002007-Lebrun1], [@ppat.1002007-Straub1]. Rhoptries are subdivided into mottled bulbous bodies and tapered electron-dense necks, corresponding to storehouses of proteins that play distinct roles in invasion [@ppat.1002007-Bradley1]. Rhoptry bulb proteins (ROPs) are injected into the host cell where they contribute to the formation of the parasitophorous vacuole [@ppat.1002007-Boothroyd1] and co-opt host cell processes to create a favorable environment for the parasite [@ppat.1002007-Gilbert1], [@ppat.1002007-Saeij1]. Several neck proteins, by contrast, assemble with the microneme protein AMA1 to constitute the *Toxoplasma* MJ [@ppat.1002007-Alexander1], [@ppat.1002007-Lebrun1]. Orthologs of most known *Toxoplasma* MJ/RON components localize to the *Plasmodium* rhoptry neck [@ppat.1002007-Alexander2], [@ppat.1002007-Cao1], [@ppat.1002007-Narum1], [@ppat.1002007-Richard1] and traffic to the *Plasmodium* moving junction [@ppat.1002007-Triglia1]. While this establishes the generally conserved nature of the complex across the Apicomplexa, MJ proteins lack identifiable domains or motifs that could provide clues to their function in this enigmatic invasion machinery. A novel rhoptry neck protein, RON8, was recently identified in *Toxoplasma* and *Neospora* as a coccidia-specific component of the MJ [@ppat.1002007-Besteiro1], [@ppat.1002007-Straub1]. RON8 associates with RONs 2/4/5 in a preformed complex within the rhoptry necks that is injected into the MJ with RONs 4/5/8 exposed to the cytoplasmic face of the host cell membrane tethered to RON2, which is thought to span the host plasma membrane via its transmembrane domains [@ppat.1002007-Besteiro1]. This topology could be facilitated by RON2′s integration within multi-lamellar whorls detected in *Plasmodium* rhoptries by electron microscopy, which could insert into the host plasma membrane during invasion, enabling soluble RONs bound to RON2 to be exposed to the host cytoplasm [@ppat.1002007-Proellocks1]. Here they are ideally poised to carry out filtration or anchoring roles for the parasite, as molecular sieving by the moving junction is restricted to this face of the host membrane [@ppat.1002007-Mordue1], [@ppat.1002007-Charron1]. These findings are supported by exogenous expression of RON8 within mammalian cells, as RON8 in the absence of other complex members traffics to its site of action at the cell periphery via its C-terminal region [@ppat.1002007-Straub1]. In addition to cleavage of the signal peptide, RON8 is processed at the N-terminus, like other MJ/RONs, and the proforms of these proteins can associate *in vitro* with each other and with pro-AMA1 [@ppat.1002007-Besteiro1]. AMA1 coprecipitates RON2 under harsh conditions and likely has a favored association with this RON, supporting a model where the universal "receptor" RON2 grants *Toxoplasma* access to a wide variety of host cells through binding its universal "ligand," AMA1, lodged in the parasite membrane [@ppat.1002007-Besteiro1]. Aside from establishing contact between RON2 and AMA1, however, precise roles for the MJ/RON proteins during invasion remain elusive. In this study, we present RON8 as the first RON junction component to be disrupted in *Toxoplasma*, and show that RON8-deficient parasites have severe defects in host cell entry. Junction proteins secreted from invading *Δron8* parasites are often disorganized, suggesting an imperfect separation of the PV from the host cell membrane after invasion. We further show through complementation of knockout parasites that the RON8 C-terminus is processed, and also identify functional domains of RON8 that are sufficient for rhoptry neck targeting and MJ complex association. RON8-ablated parasites are greater than five logs less virulent in mice, reinforcing the devastating consequences of lacking RON8 for infection *in vivo* and subsequent formation of disease. Taken together, this work identifies crucial adherence and organizational roles for RON8, demonstrating the importance of this protein within the junction for committing *Toxoplasma* to host cell entry. Results {#s2} ======= Deletion of RON8 from *Toxoplasma* {#s2a} ---------------------------------- Our initial attempts to disrupt *RON8* from the RH strain of *T. gondii* were not successful [@ppat.1002007-Straub1], consistent with the necessity of the MJ complex for parasite invasion [@ppat.1002007-Alexander1], [@ppat.1002007-Mital1]. The recent development of *Toxoplasma* strains lacking the non-homologous end-joining protein KU80 virtually eliminates heterologous DNA insertion and enables highly efficient gene knockouts [@ppat.1002007-Fox1], [@ppat.1002007-Huynh1]. We therefore investigated whether this strain would be more receptive to a direct deletion of RON8. The RON8 deletion construct was transfected into *Δku80Δhpt* parasites [@ppat.1002007-Beck1], (referred to as wildtype strain) to replace the sequence encoding residues 1-1716 with the selectable marker hypoxanthine-xanthine-guanine phosphoribosyl transferase (HPT) ([Figure 1A](#ppat-1002007-g001){ref-type="fig"}). We observed that parasites lacking RON8 in transfected populations were fully outcompeted by drug-resistant *RON8+* parasites in less than four passages, consistent with a defect in invasion. We were able to isolate a clonal line of knockout parasites by cloning early following transfection, named *Δron8 (1-1716, +HPT*) strain parasites. To remove the remainder of the *RON8* locus and the *HPT* selectable marker, a second construct (*HPT KO*, [Figure 1A](#ppat-1002007-g001){ref-type="fig"}) was transfected into *Δron8 (1-1716, +HPT)* parasites. Following negative selection with 6-thioxanthine and cloning, PCR confirmed both the absence of RON8 coding sequences and the concomitant shortening of the distance between *RON8* 5\' and 3′UTRs normally separated by ∼20 kb ([Figure 1B](#ppat-1002007-g001){ref-type="fig"}). We thus generated *Δku80ΔhptΔron8* parasites, referred to as *Δron8* ([Figure 1A](#ppat-1002007-g001){ref-type="fig"}). ::: {#ppat-1002007-g001 .fig} 10.1371/journal.ppat.1002007.g001 Figure 1 ::: {.caption} ###### RON8 is the only RON/MJ protein to be disrupted in the Apicomplexa. **A**) Schematic of the two-step targeting strategy used to ablate *RON8* from *Δku80*Δ*hpt* parasites. Homologous recombination at the *RON8* locus replaced the portion of the gene encoding residues 1-1716 with selectable marker *HPT* and concurrently removed the negative screening marker *GFP*. This generated the intermediate strain *Δron8 (1-1716, +HPT)*. To remove the remaining *RON8* genomic sequence and the *HPT* cassette, a 2^nd^ targeting construct without *HPT* and a 3′ flank downstream of the *RON8* stop codon was homologously inserted into the *Δron8 (1-1716, +HPT)* genome, producing *Δron8* parasites. PCR1 amplifies sequence inclusive of C-terminal exons, while PCR2 amplifies sequence between the regions upstream and downstream of the *RON8* locus. (**B**) PCR1 confirms the loss of *RON8* coding sequence from the *Δron8* strain, and PCR2 establishes the bridging of RON8 flanking regions through removal of *RON8*. **C**) IFAs of wildtype (*Δku80Δhpt* parental) and *Δron8* intracellular parasites show the absence of RON8 staining in the rhoptry necks of knockout parasites. RON4 is used for colocalization to the rhoptry necks for each strain. **D**) Western analysis of lysates made from wildtype and *Δron8* parasites demonstrates the loss of RON8 expression in the knockout strain, where RON2 serves as a loading control. An ∼230 kDa major RON8 breakdown product previously seen [@ppat.1002007-Straub1] is also visible. E) The moving junction can still be detected in invading *Δron8* parasites, as seen by RON4 staining (white arrow). The cartoon illustrates the direction of invasion for this *Δron8* parasite (black arrow). :::  ::: Immunofluorescence assays (IFA) of intracellular *Δron8* parasites demonstrated that RON8 was not present in the rhoptry necks ([Figure 1C](#ppat-1002007-g001){ref-type="fig"}), and Western blot analysis confirmed this strain to be devoid of RON8 ([Figure 1D](#ppat-1002007-g001){ref-type="fig"}, note that in lysates of wildtype parasites an ∼230 kDa breakdown product of RON8 is frequently seen as previously described [@ppat.1002007-Straub1]). RON4 could still be detected in the moving junction of invading knockout parasites ([Figure 1E](#ppat-1002007-g001){ref-type="fig"}, arrow), demonstrating that the ability of these parasites to form junctions was not completely compromised by the loss of RON8. We observed that higher inoculums were necessary for obtaining infection levels similar to wildtype parasites; however, no gross change in the time needed for intracellular parasites to replicate was observed (data not shown). While these results establish RON8 is not absolutely required for propagation of *Toxoplasma*, the rapid loss of *Δron8* parasites from transfected populations suggested a severe invasion defect in parasites lacking this junction component. Complementation of *Δron8* parasites by targeted insertion at the *KU80* locus {#s2b} ------------------------------------------------------------------------------ Complementing knockout parasites at the *RON8* locus would not strictly eliminate the possibility that the observed defect in invasion was due to polar effects resulting from the ablation of the gene. Given that *Δku80* parasites greatly favor homologous integration of transfected DNA, we executed a novel strategy for complementing *Δron8* by targeting a RON8 expression cassette containing a C-terminally tagged version of RON8 driven from its endogenous promoter to the ablated *KU80* locus ([Figure 2A](#ppat-1002007-g002){ref-type="fig"}). A clonal line of complemented parasites (referred to as R8c) showed exogenous RON8 both trafficked correctly to the rhoptry necks of intracellular parasites ([Figure 2B](#ppat-1002007-g002){ref-type="fig"}) and localized to the moving junction during invasion ([Figure 2B](#ppat-1002007-g002){ref-type="fig"}, arrow). Western blot analysis demonstrated slightly higher levels of RON8 expression in this line compared to wildtype parasites ([Figure 2C](#ppat-1002007-g002){ref-type="fig"}). PCR confirmed integration of the RON8 expression cassette at the *KU80* and not the *RON8* locus (not shown). Thus, the *KU80* locus appears to be a suitable site for complementation in this recently-developed strain of *T. gondii* parasites. ::: {#ppat-1002007-g002 .fig} 10.1371/journal.ppat.1002007.g002 Figure 2 ::: {.caption} ###### Complementation of *RON8* by targeting a RON8 expression cassette to the *KU80* locus. **A**) Diagram of the complementation construct used to generate R8c parasites. The HA-epitope-tagged RON8 coding sequence is driven by ∼1.8 kb of RON8 promoter sequence (dark green) and GRA2 3\'UTR (not shown) with the downstream selectable marker *HPT.* The construct was targeted to the ablated *KU80* locus by homologous recombination using KU80 flanking regions (blue). **B**) R8c parasites demonstrate restored RON8 expression in the rhoptry necks of intracellular parasites (as shown by colocalization with RON4 by IFA, top panel) and restored RON8 traffic to the moving junction during invasion (arrowhead, bottom panel). **C**) Western analysis of wildtype, *Δron8*, and R8c parasite lysates indicates slightly greater levels of RON8 expression in complemented parasites than the wildtype strain; equal lysate loads are shown by the ROP13 control (bottom panel). **D**) Immunofluorescence showing that the C-terminal HA tag of R8c parasites is often not detected in intracellular parasites (top panel). In parasites where HA is detected, RON8 colocalization shows two more posterior spots that are suggestive of pro-rhoptries (arrows, bottom panel). **E**) Detection of the C-terminal HA tag in the pro-rhoptries of R8c parasites is confirmed by IFA colocalization (arrows) by colocalization with anti pro-ROP4 antibodies. **F**) Comparison of the size of RON8 (detecting primarily the mature form) and HA (detecting the unprocessed form in the pro-rhoptries) via Western blot reveals no detectable shift in size between the unprocessed form and the mature form of the protein. This indicates that the C-terminal cleavage event does not remove a large portion of this region of the protein. :::  ::: RON8 undergoes C-terminal processing {#s2c} ------------------------------------ We observed that the C-terminal HA tag on RON8 was frequently not detectable in intracellular R8c parasites by immunofluorescence ([Figure 2D](#ppat-1002007-g002){ref-type="fig"}, *top panels*). When HA was observed, it did not colocalize with RON8 in the parasite apex but instead stained two slightly more posterior dots indicative of pro-rhoptries formed in dividing parasites ([Figure 2D](#ppat-1002007-g002){ref-type="fig"}, *bottom panels,* arrows). The HA staining within these posterior dots colocalizes with the pro-form of ROP4 in R8c parasites ([Figure 2E](#ppat-1002007-g002){ref-type="fig"}, arrows) [@ppat.1002007-Carey1], confirming pro-rhoptry localization for HA-tagged RON8 and indicating C-terminal processing of the protein during rhoptry development. To investigate the extent of proteolytic cleavage, we compared the size of the protein detected by anti-RON8 antibodies with that given by anti-HA ([Figure 2F](#ppat-1002007-g002){ref-type="fig"}). No distinguishable difference in size was observed between these two forms of the protein, suggesting that the cleavage site does not lie far from the extreme C-terminus, although determination of the extent of cleavage is difficult given RON8\'s large size (∼330 kDa). RON8 is an important mediator of firm attachment and commitment to invasion of host cells {#s2d} ----------------------------------------------------------------------------------------- To examine defects in invasion displayed by *Δron8* parasites in detail, we performed red/green invasion assays [@ppat.1002007-Huynh2], in which wildtype, *Δron8*, or R8c parasites were allowed to infect for a one-hour time period and extracellular and intracellular parasites were detected by staining and microscopic counting. A dramatic 70% reduction in penetration was observed in knockout parasites compared to the wildtype control, a defect which is completely reversed in R8c parasites ([Figure 3A](#ppat-1002007-g003){ref-type="fig"}). While we anticipated seeing a concomitant increase in attached parasites that failed to invade in the *Δron8* strain, we intriguingly saw only a mild increase in the number of extracellular parasites ([Figure 3A](#ppat-1002007-g003){ref-type="fig"}). This could be due to a failure of the knockout parasites to attach to host cells, a function previously associated solely with microneme proteins. To assess whether gross perturbation of microneme function was detectable in *Δron8* parasites, we examined the localization of MIC2 and also assessed the levels of both parasite-associated and secreted MIC2 and saw no apparent differences ([Figure S1A](#ppat.1002007.s001){ref-type="supplementary-material"}--C). We additionally assessed gliding motility by examining deposits of parasite surface antigens onto FBS-coated slides from wildtype versus knockout parasites and again saw no noticeable differences ([Figure S1D](#ppat.1002007.s001){ref-type="supplementary-material"}). While we cannot exclude the possibility that loss of RON8 impacts other microneme functions, these data suggest that microneme protein generation and function of the actin-myosin motor central to driving *T. gondii* invasion are not grossly affected in *Δron8* parasites. ::: {#ppat-1002007-g003 .fig} 10.1371/journal.ppat.1002007.g003 Figure 3 ::: {.caption} ###### Parasites lacking RON8 are deficient in invasion likely through increased detachment from host cells. **A**) Quantification of invasion using red/green assays demonstrates a substantial invasion defect for *Δron8* parasites that is rescued upon complementation. Green bars represent internal/penetrated parasites, while red bars depict attached/extracellular parasites for wildtype, *Δron8*, or R8c strains allowed to invade fibroblast monolayers for 1 hour. For each strain, at least 250 total parasites were counted from nine random fields per sample, and values are presented as internal (Int) or external (Ext) parasites per field. Data are mean values +/− SEM (error bars) for two independent experiments performed in triplicate. The asterisk indicates that parasite penetration is significantly lower (p value = 0.0245 using a Student\'s two-tailed t-test) in Δ*ron8* parasites compared to wildtype. **B**) Initial stages of attachment are unaffected in Δ*ron8* parasites. Equal numbers of wildtype, *Δron8*, or R8c parasites were preincubated with 1 µM cytochalasin D for 15 min prior to incubation with host fibroblasts in the presence of cytochalasin D, then fixed and stained in detergent-free conditions with rabbit antisera against SAG1. For each strain, values are displayed as total numbers of parasites counted divided over nine random fields. The data is expressed as mean values +/− SEM for two independent experiments performed in triplicate. :::  ::: Although RON8 contributions to the initial stages of host cell attachment would explain lower than expected counts of extracellular *Δron8* parasites, this protein could alternatively be important for maintaining a secure contact with host cells, the absence of which leads to abortive invasion and subsequent parasite detachment. To differentiate between these possibilities, we incubated wildtype, *Δron8*, or R8c parasites in cytochalasin D (cytD) to permit only the initial stages of attachment, thereby isolating these steps from the rest of invasion ([Figure 3B](#ppat-1002007-g003){ref-type="fig"}) [@ppat.1002007-Hakansson1]. Similar levels of attachment were observed for all three strains, suggesting that RON8-deficient parasites can properly attach, but do not form a *stable* grip and eventually release the host cell. As treatment with cytD does not inhibit the secretion of rhoptry bulb proteins in evacuoles [@ppat.1002007-Hakansson1], we also stained for these vesicles using anti-ROP2/3/4 antisera. We counted noticeably fewer evacuoles secreted from knockout parasites than either wildtype or R8c strains suggesting that the knockout parasites are less able to advance to the stage at which they are committed to invasion, although variability in numbers between experiments precluded assessing statistical significance (not shown). Together, these experiments demonstrate that invasion is impacted both at the steps of parasite attachment and entry, and that the attachment defect is likely due to parasites failing to obtain the firm grip on host components that is necessary for a commitment to host cell entry. We have previously shown that exogenously expressed RON8 traffics to the periphery of host cells, its predicted site of action during invasion. We tested whether host-expressed RON8 could complement the knockout by comparing infections of wildtype and Δ*ron8* parasites in cells expressing RON8, but saw no apparent rescue of the invasion defect (data not shown). This is most likely due to the inability to incorporate RON8 into the rest of the complex during the rapid process of parasite entry, but also could be due to differences in processing or other modifications of the parasite-derived form of RON8 that are not present in the exogenously expressed protein. Loss of RON8 produces abnormal secretion trails of junction proteins {#s2e} -------------------------------------------------------------------- Having established the diminished ability of *Δron8* parasites to invade host cells, we used immunofluorescence to further examine the junction morphology in those knockout parasites that could successfully invade. Whereas the junction appears as a punctate residual focus on the PV membrane of newly invaded wildtype and R8c parasites ([Figure 4A](#ppat-1002007-g004){ref-type="fig"}), ∼15% (15% and 16% in two independent experiments) of *Δron8* parasites displayed short trails of RON4 extending from the posterior end upon entering the host cell (arrows in [Figure 4B](#ppat-1002007-g004){ref-type="fig"}). In addition to RON4, the other MJ/RONs RON2 and RON5C also specifically localized to these trails ([Figure 4C](#ppat-1002007-g004){ref-type="fig"}). The trails of MJ components are distinct from so called "slime trails" deposited by gliding parasites on FBS coated slides as the MJ components are not present in slime trails or in staining of extracellular parasites. The trails also appear to be specific to RON/MJ proteins as the non-junction rhoptry proteins (ROP2/3/4) are not present within these structures (data not shown). The appearance of these trails in the absence of RON8 suggests this protein is important for maintaining the integrity of the MJ complex or for pinching the nascent PV off from the plasma membrane at the end of invasion. ::: {#ppat-1002007-g004 .fig} 10.1371/journal.ppat.1002007.g004 Figure 4 ::: {.caption} ###### Disorganized secretion of junction components in Δ*ron8* parasites. **A**) Pulse invasion assays using wildtype parasites were carried out for 5 min on host fibroblasts prior to IFA with antibodies against RON4 and RON2, showing the typical punctate spot of junction proteins (arrow) after the parasite has fully invaded the host cell but the vacuole has likely not yet detached from the host membrane (see cartoon). **B**) Similar assays conducted with *Δron8* parasites show a trail of RON4 secreted from the posterior end of the newly invaded parasite (arrow). The orientation of the parasite is shown by costaining with anti-ISP1 which detects the apical cap of the IMC. **C**) Other MJ/RONs detected by anti-RON2 (top panel) or anti-RON5C antibodies which (bottom panel) colocalize with RON4 in secreted trails deposited by *Δron8* parasites (arrows). :::  ::: Selective complementation reveals regions of RON8 sufficient for rhoptry neck targeting and MJ complex formation {#s2f} ---------------------------------------------------------------------------------------------------------------- Our success in rescuing *Δron8* parasites encouraged us to use selective complementation of the knockout to discern RON8 functional domains as well as regions necessary for trafficking to the neck subcompartment. As RON8 has been shown to have a N-terminal prodomain and such prodomains are known to function in rhoptry targeting, we first tested whether the first 262 amino acids are sufficient for targeting to the rhoptry necks by fusing this region to the reporter protein mCherry (R8~pro~mCherry). We expressed the fusion and a full-length control by targeting each back to the ablated RON8 locus in an expression cassette driven from the RON8 promoter ([Figure 5A, B](#ppat-1002007-g005){ref-type="fig"}). In stably transformed parasite clones, the R8~pro~mCherry fusion only partially targeted to the rhoptry necks as assessed by colocalization with the rhoptry neck protein RON1 whereas the full-length control targeted perfectly. mCherry that was not localized to the rhoptry necks was found in punctate spots that localized in both apical and basal regions of the parasite. This data suggests that the RON8 prodomain does play some role in trafficking, but additional sequences are also needed for efficient targeting of the protein to the rhoptry necks. ::: {#ppat-1002007-g005 .fig} 10.1371/journal.ppat.1002007.g005 Figure 5 ::: {.caption} ###### Selective complementation of *Δron8* parasites reveals regions of RON8 necessary for RON8 targeting, complex formation, and function. **A**) Schematic and IFA of complementation with full-length 1--2980 HA-tagged RON8 targeted to the RON8 locus. Shown are the RON8 flanks used for targeting by homologous recombination, and the coding sequence, signal peptide (SP), N-terminal prodomain (pro) and C-terminal HA tag (HA). The selectable marker HPT is also shown. IFA of stable parasite clones shows correct targeting to the rhoptry necks as shown by colocalization with RON5N. **B**) The N-terminal 262 amino acids of RON8 are only partially sufficient for targeting RON8 to the rhoptry necks. The construct contains the first 262 amino acids fused to the mCherry reporter and is targeted to the RON8 locus using the same method as the full length control in "A". IFA shows that some of the fusion is trafficked properly to the rhoptry necks (arrows) as assessed by colocalization with RON1. However, a significant amount of mistargeted material is also seen in punctate spots in the apical end of the parasite and also in more diffuse patches in the posterior portion of the parasite. **C**) Addition of the C-terminal portion of RON8 (residues 1318--2980) restores rhoptry neck targeting. Diagram of the construct and IFA showing restoration of rhoptry neck targeting upon inclusion of the C-terminal region of the protein. Colocalization is shown using antisera against RON1. **D**) Immunoprecipitation of the MJ complex from parasites expressing the full length RON8 in "A" shows efficient purification of all members of the MJ complex. Western blots show an enrichment of all members compared to whole cell lysates (equivalent amounts of lysate and elution are loaded for each). Anti-RON2 antibodies are used for the immunoprecipitation. **E**) The R8~pro~mCherryR8C fusion is incorporated into the MJ complex. Immunoprecipitation of the MJ complex with anti-RON2 precipitates the R8~pro~mCherryR8C fusion as well as other members of the MJ complex. Thus, the RON8 prodomain plus C-terminal region are sufficient for rhoptry neck targeting and complex association. :::  ::: We then assessed whether addition of the C-terminal half of the protein would improve targeting by fusing this portion to the C-terminus of the R8~pro~mCherry protein ([Figure 5C](#ppat-1002007-g005){ref-type="fig"}). Addition of the C-terminal region of RON8 to the fusion construct (termed R8~pro~mCherryR8C) restored efficient rhoptry neck targeting as seen by colocalization with RON1. Having completely restored rhoptry neck targeting, we determined whether the R8~pro~mCherryR8C fusion was incorporated into the MJ complex by immunoprecipitating the complex with RON2 antisera from parasites expressing the fusion protein, using parasites rescued with the full length protein as a control ([Figure 5D](#ppat-1002007-g005){ref-type="fig"}/E). Both the control and the R8~pro~mCherryR8C fusion were efficiently co-precipitated with RON2 and the other RONs in the MJ complex. Thus, the RON8 N-terminal prodomain combined with the C-terminal region is sufficient for rhoptry neck targeting and MJ complex formation. While trafficking and complex association were restored, parasites expressing R8~pro~mCherryR8C showed the same defect in invasion as Δ*ron8* parasites (not shown), demonstrating that the N-terminal region of RON8 is necessary for function. RON8-deficient parasites are dramatically weakened in virulence *in vivo* {#s2g} ------------------------------------------------------------------------- With the significant defects in host cell entry observed *in vitro*, we assessed the degree to which virulence was affected in parasites lacking RON8. To that end, CD1 mice were infected with either a sufficiently lethal dose of wildtype parasites (LD~100~ = 1) or increasing doses of either *Δron8* or R8c ([Figure 6](#ppat-1002007-g006){ref-type="fig"}). While all mice infected with a low dose of 50 wildtype parasites succumbed to infection by day 9, mice subjected to doses of \<5×10^4^ knockout parasites did not show any visible signs of infection. Mice infected with 5×10^4^ parasites did show symptoms of infection, but three of the four mice survived the acute infection. The mice generally became moribund with 5×10^5^ parasites injected; however, even at this high dose, one animal out of the group of four recovered. The virulence phenotype was completely reversed in the complemented strain, demonstrating that the defect was specifically due to the lack of RON8. Mice infected with even the lowest doses of *Δron8* parasites developed an immune response against the parasite as seen by seroconversion, and all mice infected with the knockout survived a challenge with 1×10^4^ wildtype tachyzoites (data not shown). Statistical analysis of these virulence experiments yields an LD~50~ for *Δron8* parasites = ∼2.6×10^5^ parasites, a decrease of more than five logs compared to the wildtype strain. These experiments demonstrate that the inability of *Δron8* parasites to firmly attach and thereby commit to invading host cells *in vitro* produces a dramatically impoverished ability to cause disease *in vivo*. ::: {#ppat-1002007-g006 .fig} 10.1371/journal.ppat.1002007.g006 Figure 6 ::: {.caption} ###### RON8-deficient parasites are severely compromised in establishing disease *in vivo*. **A--C**) Groups of 4 CD1 mice were infected with (**A**) 50 wildtype, (**B**) 50, 500, 5×10^3^, 5×10^4^, or 5×10^5^ *Δron8*, or (**C**) 50, 500, or 5×10^3^ R8c parasites and monitored for 28 days. All surviving mice from *Δron8* parasites were protected against a lethal challenge of 10^4^ wildtype parasites (not shown). :::  ::: Discussion {#s3} ========== RON8 is the first moving junction rhoptry neck protein to be functionally analyzed by direct knockout in an apicomplexan parasite. This overturns the previous paradigm that all MJ components are essential for parasite survival, which had been postulated with the need for conditional approaches to study AMA1 [@ppat.1002007-Mital1] and technical difficulties in ablating other junction partners [@ppat.1002007-Alexander1], [@ppat.1002007-Straub1]. Deleting RON8 in the *Δku80* strain reinforces both the major technical advance in promoting homologous recombination in parasites lacking KU80 and the potential produced by this advance to examine the moving junction with greater capacity than previously thought. Our success in ablating RON8 is a first step toward ultimately establishing the "minimal" junction complex required by *T. gondii* for host cell entry. AMA1 is already known to be essential for tachyzoite invasion [@ppat.1002007-Mital1]. RON2, which displays a privileged association with AMA1 and likely spans the host membrane [@ppat.1002007-Besteiro1], will probably also prove to be essential in *Toxoplasma*. This leaves the soluble junction components RON4 and RON5 (which is processed into N and C-terminal fragments), which we are currently attempting to knockout from *Δku80* and *Δron8* parasites to explore their roles in invasion. RON8\'s dispensability in *Toxoplasma* agrees with increasing evidence suggesting the moving junction is constructed using different complements of junction proteins deployed during particular life cycle stages in apicomplexan parasites. RON4 can be detected in the ring-shaped junctions of egressing parasites, but AMA1 is absent from these structures, and parasites depleted of AMA1 expression are crippled in invasion, but not affected in egress [@ppat.1002007-Alexander1]. Proteomic analysis of *Eimeria tenella* suggests that its ortholog of RON5 is expressed in sporozoites but not in merozoites, while sporozoites appear to lack AMA1 and RON4 [@ppat.1002007-Lal1]. Similarly, PfRON2 is expressed in all *Plasmodium* invasive forms except for ookinetes, which do not form moving junctions and lack rhoptries [@ppat.1002007-TufetBayona1]. Distinct arsenals of junction proteins tailored for specific life cycle stages within each apicomplexan may reflect the evolutionary fine-tuning of this structure that enables phylum members to exploit unique host niches. Our attachment and invasion experiments ([Figure 3](#ppat-1002007-g003){ref-type="fig"}) suggest that RON8 may have evolved within the coccidia to anchor the invading parasite to common host cytoskeletal proteins [@ppat.1002007-Straub1], although a biochemical demonstration of RON8\'s link with the host cell will firmly establish this scenario. The current model of *Toxoplasma* invasion presents attachment as a series of steps increasing in strength, beginning with low affinity interactions between GPI-anchored parasite surface antigens and unknown host contacts, followed by host surface protein association with transmembrane microneme proteins on the parasite surface, and finally moving junction formation through AMA1\'s interaction with RON2 embedded in the host membrane [@ppat.1002007-Besteiro1], [@ppat.1002007-Richard1]. The work presented here adds a further critical step to invasion: the secure connection of the parasite to the host through RON8 contacts *within* the host cell, supported by the topology of the RON proteins in the MJ, the exogenous expression of RON8 in mammalian cells [@ppat.1002007-Besteiro1], [@ppat.1002007-Richard1], and the localization of host F-actin rings observed at *Toxoplasma* and *Plasmodium* moving junctions [@ppat.1002007-Gonzalez1]. According to this updated model ([Figure 7](#ppat-1002007-g007){ref-type="fig"}), the moving junction in wildtype parasites ([Figure 7](#ppat-1002007-g007){ref-type="fig"}, *left*) serves as a two-fold lock both inside and outside the host cell, irreversibly directing the parasite toward complete penetration of its target. In knockout parasites ([Figure 7](#ppat-1002007-g007){ref-type="fig"}, *right*), the loss of the intracellular clasp provided by RON8 severely weakens the moving junction\'s hold on its target, resulting in frequent detachment from host cells and thereby debilitating the invasive capacity of *Toxoplasma*. ::: {#ppat-1002007-g007 .fig} 10.1371/journal.ppat.1002007.g007 Figure 7 ::: {.caption} ###### Model of the *Toxoplasma* moving junction in wildtype and Δ*ron8* parasites. The model shows a diagram of a partially invaded parasite with moving junction proteins at the interface of the host and invading parasite. The transmembrane proteins RON2 and AMA1 form the bridge between the host cell and the invading parasite with the soluble proteins RON4, RON5 (processed into N and C fragments) and RON8 exposed to the host cell cytoplasm. RON8 forms a stable intracellular clamp that commits the parasite to invasion, potentially by binding host elements in the cortical cytoskeleton. In the absence of RON8, the MJ is frequently unstable, leading to frequent abortive attachment and invasion. :::  ::: In addition to disrupting contacts with host components important for parasite adherence, the loss of RON8 imparts further structural abnormalities in the moving junctions of parasites that overcome this defect, as evidenced by the appearance of MJ components in trails dragging behind newly invaded *Δron8* parasites ([Figure 4B, C](#ppat-1002007-g004){ref-type="fig"}). Disorganized trails are not ubiquitous during pulse invasion of these parasites (∼20%), suggesting that the structure is not entirely disorganized, a finding which is supported by the recovery of the remainder of the junction complex by co-immunoprecipitation ([Figure 5](#ppat-1002007-g005){ref-type="fig"}). Junctions that are compromised enough to form trails could indicate difficulties in PVM separation from the host membrane without RON8, and the low numbers of evacuoles observed in cytD-arrested *Δron8* parasites suggests RON8 function is a prerequisite for rhoptry bulb secretion. The latter phenotype is reminiscent of defects observed in parasites depleted of AMA1, which reorient after initial contact with the host cell but generally do not form a moving junction [@ppat.1002007-Alexander1] or inject evacuoles [@ppat.1002007-Mital1]. Similarly, preventing AMA1-RON complex formation using inhibitory peptides abolishes evacuole secretion in *Plasmodium* [@ppat.1002007-Richard1]. While this indicates some signal occurs after microneme and rhoptry neck secretion to allow rhoptry bulb secretion, it is unclear how such a signal is transmitted between apical organelles of the invading parasite. We have demonstrated that RON8 is processed at its C-terminus ([Figure 2D](#ppat-1002007-g002){ref-type="fig"}) at a site that cannot be readily resolved by Western analysis comparing the size of the HA-tagged RON8 precursor with its mature form. This cleavage event is distinct from RON8\'s processing at the N-terminus demonstrated in Besteiro et al [@ppat.1002007-Besteiro1], which also showed all MJ proteins are subject to proteolysis, probably at ROP1-like SφXE sites which are believed to be processed by the subtilisin protease TgSUB2 [@ppat.1002007-Miller1]. There are two candidate processing sites near the C-terminus that match this consensus site (SAME at residues 2652--2655 and SAGE at 2702--2705), but cleavage at these sites would remove ∼30--35 kDa which might be resolvable by SDS-PAGE. Additional sites are also possible as we have seen some minor variations in the residues present in these cleavage sites (Hajagos and Bradley, unpublished results) and others have highlighted a potential cleavage site at the extreme C-terminus of RON8 that would remove only nineteen amino acids from the protein (residues 2958--2961, SFLQ, [@ppat.1002007-Besteiro1]). Our HA-tagged construct and additional fusions will undoubtedly aid in determining the precise cleavage site and the role of processing in RON8 targeting, complex formation and function. RON8\'s exposure to the host cytosol during invasion makes it a prime candidate for restricting access of host transmembrane proteins to the nascent PVM, which may occur in conjunction with the stabilizing grip on host components identified in this work. Our initial attempts to examine the ability of *Δron8* parasites to exclude host Na^+^/K^+^ ATPase (known to be sieved during invasion) by immunofluorescence has not revealed any gross differences from the wildtype strain, although minor changes would be difficult to detect (data not shown). While we cannot exclude a role for this protein in some aspect of molecular sieving, it is possible that RON8\'s sole function is to firmly grip host peripheral components to completely anchor the parasite, leaving other cytosolic-exposed junction proteins (RON4 and RON5) to conduct filtration. It is currently unknown whether the individual RON proteins of the MJ complex each contain their own rhoptry neck targeting information or if one protein can escort the others as is seen in adhesive complexes secreted from the micronemes and for the *Plasmodium* rhoptry protein RAP1 [@ppat.1002007-Reiss1], [@ppat.1002007-Richard2], [@ppat.1002007-Saouros1]. While the precise cleavage site of RON8\'s N-terminal prodomain has not been determined, this region does appear to contain some but not complete rhoptry targeting information. Addition of the C-terminal region fully restores rhoptry neck targeting, which is likely due to the ability to associate with other complex members ([Figure 5E](#ppat-1002007-g005){ref-type="fig"}). The lack of a functional rescue in parasites lacking the N-terminal region of RON8 indicates that this region contains domains that are critical for invasion. These domains could play a direct role in the mechanics of invasion or provide additional interaction domains, thus strengthening the integration of RON8 into the complex. Deletion of the N-terminal region could also impact folding of the remaining C-terminal portion in the complex, which our previous exogenous expression data suggests is binding to host components at the periphery of the cell [@ppat.1002007-Straub1]. While a large number of functional domains are likely to be present in the 330 kDa RON8 protein, the data presented here demonstrate that functional complementation promises to be a powerful tool to further dissect specific regions of RON8 involved in targeting, complex formation, and function. We have shown that the RON8-deficient strain is more than five logs less virulent than wildtype *Δku80* parasites in outbred CD1 mice ([Figure 6](#ppat-1002007-g006){ref-type="fig"}). The contribution of MJ/RONs to the host response during *in vivo Toxoplasma* infection is likely minimal, due to the short period of time the moving junction is present during invasion (∼30 sec) as well as the topology of RON8 and other soluble MJ/RONs beneath the plasma membrane in this structure [@ppat.1002007-Besteiro1], [@ppat.1002007-Straub1]. Studies of RON4 antigenicity in *Plasmodium falciparum* [@ppat.1002007-Morahan1] and *P. yoelii* [@ppat.1002007-Narum1] support this, as this protein displays extensive sequence conservation with concordantly little apparent immune pressure. The dramatic impact on virulence we observe in *Δron8* parasites is therefore in all likelihood a consequence of the significant defects in host cell entry displayed *in vitro* rather than any enhanced host reaction to the parasite. In conclusion, we have utilized RON8 knockouts to make greater inroads into analyzing rhoptry neck protein function than previously achieved in *Toxoplasma*. In the process, we have expanded the model of *T. gondii* invasion by implicating at least one rhoptry protein in ensuring the parasite\'s commitment to entering its target cell. Our complementation strategy will enable further dissection of RON8 functional and interaction domains, which promises a more detailed understanding of the molecular architecture of this phenomenal structure. Future experiments will also examine RON8\'s contribution to egress, as moving junctions are observed during pathogen exit from the host cell [@ppat.1002007-Alexander1]. Although RON8 is a coccidia-specific member of the complex, identifying host contacts with this protein will likely prove useful in preventing invasion across the phylum, as host proteins regulating cytoskeletal filament assembly are recruited to the junctions of invading *Plasmodium* and *Toxoplasma* [@ppat.1002007-Gonzalez1]. Through this work and upcoming studies, a complete understanding of the moving junction\'s participation in *Toxoplasma*\'s remarkable success at a parasitic lifestyle will be at hand. Materials and Methods {#s4} ===================== Ethics statement {#s4a} ---------------- *Toxoplasma* infections in mice and antibodies raised in mice were performed under the guidelines of the Animal Welfare Act and the PHS Policy on Humane Care and Use of Laboratory Animals. Specific details of our protocol were approved by the UCLA Animal Research Committee (ARC\# 2004-055). Parasite and host cell culture {#s4b} ------------------------------ The RHΔ*hpt* and RHΔ*hptΔku80* strains of *Toxoplasma* have been previously described [@ppat.1002007-Beck1], [@ppat.1002007-Donald1], and were maintained on confluent monolayers of human foreskin fibroblasts (HFFs) grown in Dulbecco\'s modified eagle medium supplemented with 5% fetal bovine serum, 5% Cosmic Calf Serum (Hyclone), and 2 mM glutamine [@ppat.1002007-Bradley1]. Antibodies and Western blot analysis {#s4c} ------------------------------------ The following antibodies were used in immunofluorescence and Western blot assays: mouse polyclonal anti-RON8 (1∶400) [@ppat.1002007-Straub1], rabbit polyclonal anti-RON4 (1∶7000) [@ppat.1002007-Alexander1], mouse polyclonal anti-RON2 (1∶800) [@ppat.1002007-Bradley1], rabbit polyclonal anti-RON2 (1∶1000) (described below), mouse polyclonal anti-RON5N and RON5C (both 1∶600) [@ppat.1002007-Straub1], mouse polyclonal anti-RON1 (1∶300) [@ppat.1002007-Alexander1], mouse monoclonal anti-ISP1 [@ppat.1002007-Beck1], rabbit polyclonal anti-SAG1 (1∶100000), a gift from John Boothroyd, rabbit polyclonal anti-SAG2 (1∶4000, John Boothroyd), mouse monoclonal T34A7 against ROP2/3/4 (1∶300) [@ppat.1002007-Sadak1], rabbit polyclonal anti-ROP13 (1∶1000) [@ppat.1002007-Turetzky1], mouse monoclonal 1B10 anti-ROP7 (1∶1000) [@ppat.1002007-Rome1], rabbit polyclonal anti-pro-ROP4 (1∶1000) [@ppat.1002007-Carey1], rabbit polyclonal anti-HA (1∶300) (Invitrogen), and mouse monoclonal anti-HA (1∶500) (Covance). SDS-PAGE gels were used to resolve proteins by Western blot analysis as previously described [@ppat.1002007-Bradley2]. Secondary antibodies were horseradish peroxidase (HRP)-conjugated goat anti-mouse and goat anti-rabbit used at a dilution of 1∶500--2000 (Sigma) and detected using the ECL Western Blot Detection Kit (Thermo Scientific). To generate a RON2 fusion protein with a N-terminal 6xHis tag for antibody production, the *RON2* cDNA encoding amino acids 25 to 313 (residue numbers are from Genbank accession HQ110093 with residue 1 as the start methionine) was amplified from a *Toxoplasma* RH strain cDNA library using primers P21 and P22 ([Table S1](#ppat.1002007.s002){ref-type="supplementary-material"}). The amplified product was cloned into *pET28a(+)* (Novagen) using the HindIII and XhoI sites encoded in the primers. Production and purification of rRON2~25--313~ from *E. coli* strain Rosetta (Novagen) using a nickel-nitrilotriacetic acid matrix were done essentially following manufacturer\'s instructions (Qiagen). Antibodies to rRON2~25--313~ were generated in rabbits by Covance, Inc. Specific antisera against mCherry (used at 1∶4000) was generated in mice using recombinant 6xHis tagged protein purified by denaturing nickel agarose chromatography. mCherry was amplified using primers p23 and p24 and the amplified product was cloned into the pET101 (Invitrogen) bacterial expression plasmid which encodes a 6xHis tag in frame with the C-terminus of the gene. BL21-DE3 cells were transformed with the construct and induced with 1 mM IPTG for 5 hours before the cells were collected and mCherry-6xHis was purified and dialyzed against PBS. BALB/c mice (Charles River) were immunized with ∼70 µg of recombinant protein on a 21 day immunization schedule. The resulting mouse polyclonal antiserum were collected and tested by Western blot analysis. Disruption of RON8 in *T. gondii* {#s4d} --------------------------------- The initial *RON8* KO vector was previously described [@ppat.1002007-Straub1]. The construct was linearized by KpnI digestion, and 30 µg of DNA were transfected by electroporation into Δ*ku80xΔhpt* strain parasites [@ppat.1002007-Beck1]. Following selection of transformants in media containing 50 µg/ml MPA and 50 µg/ml xanthine for five days, the knockout populations were scrape-syringed and cloned by limiting dilution. GFP negative parasites were screened by immunofluorescence using anti-RON8 polyclonal antisera, and a clonal line of GFP/RON8 double negative parasites was named strain *Δron8* (*1-1716, +HPT*). For generation of the *HPT KO* vector, the HPT cassette from *RON8 KO* was removed by digestion with XbaI and re-ligated, generating *RON8KOΔHPT*. To remove the remaining RON8 coding region along with the HPT marker, a new 3′ flank downstream of the *RON8* stop codon was amplified from RH strain genomic DNA using primers P9 and P10 and subcloned into *RON8KOΔHPT* using KpnI and XbaI to make *HPT KO*. This plasmid was linearized by KpnI digestion and 50 µg of DNA transfected by electroporation into *Δron8* (*1-1716, +HPT*) parasites. Exclusion of *HPT* was selected for using 350 µg/ml 6-thioxanthine (Sigma) for 5 weeks, after which populations were cloned by limiting dilution. Clones were then screened for the absence of RON8 and susceptibility to medium containing 50 µg/ml MPA and 50 µg/ml xanthine, to identify *Δron8* parasites. Generation of full-length and selectively complemented *Δron8* parasites {#s4e} ------------------------------------------------------------------------ An *HPT* cassette was amplified from the pMini-GFP.ht plasmid [@ppat.1002007-Karasov1] using primers P11 and P12 and cloned into pCR2.1-TOPO (Invitrogen) as per manufacturer\'s instructions. Sequences upstream (using primers P13 and P14) and downstream (primers P15 and P16) of the ablated *KU80* locus were amplified from RH*ΔhptΔku80* genomic DNA and subcloned into pCR2.1-TOPO+*HPT* using KpnI and SpeI for the 5′ flank (∼1.5 kb) and NsiI and XbaI for the 3′ flank (∼1.1 kb). The 12 kb RON8-HA expression cassette was subcloned into this vector at an EcoRV site following excision from *pGRA-HA\_HPT-RON8* [@ppat.1002007-Straub1] by PciI and DraIII digestion and blunting with Klenow fragment (NEB). Sequencing established the inverse orientation of the cassette relative to the *KU80* flanks and confirmed placement of the *HPT* and RON8-HA cassettes between these flanks. This vector (*R8compHPT*) was then linearized by PmeI digestion, and 25*µ*g of DNA transfected into *Δron8* parasites, followed by MPA/xanthine selection. Resistant clones were screened by IFA and RON8 complemented parasites were confirmed by Western blot analysis, identifying R8c strain parasites. The *HPT* cassette described above was also subcloned back into the *HPT KO* vector after blunting at the XbaI site. This vector, *HPTKO+HPT*, was then digested by NotI and ApaI, blunted, and ligated with the blunted 12 kb RON8 expression cassette described above to make a complementation vector encoding *HPT* and RON8-HA adjacent to RON8 flanks lying outside the entire RON8 coding sequence. After confirming forward orientation of the RON8-HA cassette relative to the *RON8* flanks and placement of the *HPT* and RON8-HA cassettes between these flanks by sequencing, this vector (*R8HPTKO+HPT*) was linearized by AflII digestion and transfected (25 µg) into *Δron8* parasites, followed by MPA/xanthine selection for 9 days and cloning. Resistant clones were screened by IFA and RON8 complemented parasites were confirmed by Western blot analysis, with positive clones named *Δron8 + RON8~1--2980~* parasites. To make fusions of RON8 fragments with mCherry, primers P17 and P18 were used to amplify the mCherry coding sequence using pmCherry Vector (Clontech) as a template. This PCR product contains SmaI at one end and AvrII/PacI sites on the other end; after SmaI/PacI digestion, it was subcloned into *R8HPTKO+HPT* digested at SmaI and PacI endogenous sites in RON8-HA. This replaced sequences encoding residues 263--2980 and the HA tag with mCherry. This vector (*R8proMCHERHPTKO*) was linearized with AflII and 25 µg of DNA transfected into *Δron8* parasites prior to 9 days MPA/xanthine selection, cloning, and IFA screening as above, generating R8~pro~mCherry parasites. For introduction of the RON8 C-terminus into *R8proMCHERHPTKO*, the sequence encoding residues 1318--2980 and the C-terminal HA tag flanked by AvrII/PacI sites was amplified from template pGRA-HA\_HPT-RON8 using primers P19 and P20. This fragment was subcloned into *R8proMCHERHPTKO* at AvrII/PacI sites, sequenced to confirm both junctions with mCherry were in frame, linearized with AflII, transfected, drug selected, cloned, and screened by IFA as above to identify *Δron8 +* R8~pro~mCherryR8C parasites. PCR amplification for verification of constructs and knockouts {#s4f} -------------------------------------------------------------- Template DNA was extracted from harvested wildtype *Δku80*, *Δron8*, or R8c parasites using the Wizard Genomic DNA Purification Kit (Promega) as per the manufacturer\'s protocol. RON8 coding sequence removal in *Δron8* parasites was confirmed using primers P1 and P2 using either *Δku80* or *Δron8* genomic DNA. The removal of the entire *RON8* genomic locus with concomitant bridging of flanking sequences normally separated by ∼20 kb was confirmed using primers P3 and P4 using *Δku80*, *Δron8*, or R8c genomic DNA as a template. Correct targeting of the full-length RON8 complementation cassette at the *KU80* locus was confirmed using primers P5 with P6 and P7 with P8. Immunofluorescence and invasion assays {#s4g} -------------------------------------- Intracellular parasites were examined by immunofluorescence as per [@ppat.1002007-Straub1]; confluent HFFs grown on glass coverslips were infected with parasites and incubated for 24--30 hours at 37°C, washed with PBS, and then fixed with either ice-cold methanol for 3 minutes or 3.7% formaldehyde/PBS for 15 minutes prior to quenching with phosphate-buffered saline (PBS) plus 0.1 M glycine for 5 min. Coverslips were then washed in PBS and blocked with PBS/3% bovine serum albumin (BSA) or a completely permeabilizing solution of PBS/3%BSA/0.1% Triton X-100 (PBT buffer) for 30 minutes. Primary antibodies were diluted in PBS/3%BSA or PBT buffer for 1 hour. Coverslips were washed five times in PBS and incubated with secondary antibodies Alexa-488 goat anti-mouse and Alexa-594 goat anti-rabbit (or Alexa 594 goat anti-mouse and Alexa 488 goat anti-rabbit, Molecular Probes, OR) diluted 1∶2000 in PBS/3%BSA for 1 hour. Following secondary washes in PBS, coverslips were then mounted onto slides using Vectashield mounting medium for fluorescence microscopy using a Zeiss upright light microscope (Zeiss Axio Imager Z1) using either a 100x or 63x oil immersion objective. All images were rendered using Axiovision software in conjunction with a Zeiss digital CCD camera (AxioCam MRm). Early invasion experiments were performed using low temperatures as per [@ppat.1002007-Straub1]. Invasion assays were conducted to distinguish between extracellular parasites and internalized parasites via a red/green invasion assay as described [@ppat.1002007-Huynh2]. In brief, equivalent cultures of wildtype *Δku80*, *Δron8*, and R8c parasites were scraped and passed through a 27-gauge needle to collect strictly intracellular parasites for use in the invasion assay. After washing with fresh medium, ∼3×10^6^ tachyzoites were resuspended in 500 µl of fresh prewarmed DMEM and placed onto separate chambers of an 8-well chamber slide (Falcon) containing confluent HFF monolayers. These parasites were allowed to invade for one hour at 37°C, after which monolayers were washed three times in PBS and fixed with EM-grade 3.7% formaldehyde/PBS (Biosciences, Inc.). After quenching as above, the samples were blocked in PBS/3%BSA for 25 minutes and incubated with rabbit anti-SAG1 diluted in PBS/3%BSA for 1 hour, then washed five times in 1X PBS and permeabilized with PBT buffer for 30 minutes prior to incubation with mouse anti-ROP7 diluted in PBT buffer as a second primary step. Secondary staining and fluorescence microscopy then proceeded as above. Parasites staining with both anti-SAG1 and ROP7 denoted attached but uninvaded parasites, while those staining only for ROP7 were scored as internalized. Nine fields were randomly counted for each chamber, yielding total counts of 250--800 parasites for all strains. Invasion experiments were conducted in triplicate and repeated at least twice. To examine whether the Δ*ron8* parasites could be rescued by host cells exogenously expressing RON8, we used a competition growth assay in cells with and without RON8 expression. No difference in the rate at which the wildtype parasites outcompeted the knockout was observed, indicating that exogenously expressed RON8 cannot complement the invasion defect. Attachment and evacuole assays utilizing cytochalasin D were performed as described [@ppat.1002007-Mital1]; wildtype, *Δron8*, and R8c parasites were scrape-syringed as above and resuspended in Endo buffer (44.7 mM K~2~SO~4~, 10 mM Mg~2~SO~4~, 106 mM sucrose, 5 mM glucose, 20 mM Tris, 0.35% wt/vol BSA, pH 8.2) containing 1 µM cytochalasin D (Sigma-Aldrich). After incubation at room temperature for 10 min, ∼3×10^6^ parasites from each strain were used to infect HFF monolayers grown on 8-chamber slides and incubated for 20 min at 37°C. Media was replaced with prewarmed DMEM/10% FBS containing 1 µM cytochalasin D and incubation continued for another 15 min at 37°C before fixation with formaldehyde and immunofluorescence/counting as described above for red/green invasion assays (except that instead of anti-ROP7 antibody, evacuoles were labeled with monoclonal anti-ROP2/3/4 antibody). To examine whether MJ component trails are present in extracellular parasites, intracellular wildtype or *Δron8* parasites were scrape-syringed, washed once in PBS, and allowed to adhere to slides prior to coating with 3.7% formaldehyde/PBS fixative for 15 min prior to quenching with 0.1 M glycine/PBS as above. Following fixation, blocking/permeabilization in PBT buffer and staining with rabbit anti-RON4 was performed as before. Coimmunoprecipitation of the MJ complex from complemented *Δron8* strains {#s4h} ------------------------------------------------------------------------- Rabbit polyclonal anti-RON2 was cross-linked to Protein A-Sepharose beads (Amersham) using dimethyl pimelimidate as described [@ppat.1002007-Alexander1]. RON2-linked beads were then incubated with parasite lysates made in a modified RIPA buffer + Complete Protease Inhibitor described in [@ppat.1002007-Alexander1] (50 mM Tris-Cl pH 8.0, 5 mM EDTA, 75 mM NaCl, 1% NP40, 0.5% DOC, 0.005% SDS). In brief, ∼2×10^8^ extracellular parasites were centrifuged at 3000 *g* for 20 min. The parasites were washed once in 1X PBS and then lysed on ice for 20 min prior to removing insoluble material by centrifugation at 10000 *g* for 20 min. Antibody-coupled beads were incubated with lysate at 25°C for 3 hours before four washes with lysis buffer. The bound proteins were eluted using 100 mM triethylamine pH 11.5, and lyophilized to remove the triethylamine and concentrate the eluate. The resulting products were analyzed by Western blot using antibodies against RON/MJ proteins. Gliding motility trail assays {#s4i} ----------------------------- Motility experiments were performed largely as described [@ppat.1002007-Dobrowolski1]; wildtype or *Δron8* parasites were resuspended in Hank\'s buffered saline solution (HBSS) and allowed to glide on serum-coated glass coverslips for 20 min at 37°C. Coverslips were rinsed twice in PBS and fixed with EM-grade formaldehyde for 15 min prior to immunofluorescence with rabbit anti-SAG2 antibodies as described above. Mouse virulence assays {#s4j} ---------------------- Intracellular wildtype (Δ*ku80*Δ*hpt* parental), Δ*ron8*, and R8c parasites were scrape-syringed from infected HFF monolayers and resuspended in Opti-MEM prior to intraperitoneal injection of 50 wildtype (Δ*ku80*Δ*hpt* parental); 50, 500, 5×10^3^, 5×10^4^, or 5×10^5^ *Δron8*; or 50, 500, or 5×10^3^ R8c strain tachyzoites in outbred CD1 female mice, making a total of 9 groups of 4 mice each. Mice in each group were bled both prior to injection and surviving mice bled 15 days post infection, and serum used in Western blot analysis lysates from wildtype parasites to test for seroconversion. Prism GraphPad software was used to determine the LD~50~ of *Δron8* parasites from an analysis of the results by a standard sigmoidal dose-response curve. Mice were monitored for 25 days, and surviving mice "protected" by *Δron8* immunization were challenged by intraperitoneal injection with 1×10^4^ wildtype tachyzoites at day 30 and assessed for an additional 30 days. All care and handling of animals was in accordance with institutional guidelines and approved by the UCLA Animal Research Committee. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### Gross assessment of microneme function by examining MIC2 protein levels and gliding motility in Δ*ron8* and control parasites. **A**) IFA of wildtype and Δ*ron8* parasites showing no noticeable change in localization of MIC2. **B**) Western blot analysis of total parasite lysates from wild-type, *Δron8*, and R8c parasites showing approximately even levels of MIC2 protein are present. RON2 is used as a loading control. **C**) Ethanol induced secretion shows no apparent change in secreted MIC2. Parasite equivalents are shown for each sample and are used as a loading control showing that ∼33% of the MIC2 is released from both wildtype and *Δron8* parasites. **D**) Gliding motility assays staining for SAG2 deposited on serum-coated glass coverslips. Similar trails were observed for wildtype and knockout parasites showing that motility is not substantially compromised in *Δron8* parasites. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S1 ::: {.caption} ###### Oligonucleotide primers utilized in this study. (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: We acknowledge Vern Carruthers for supplying the Δ*ku80*Δ*hpt* strain prior to publication. We thank Jean-Francois Dubremetz for anti-MIC2 antibodies, Gary Ward for anti-proROP4 antibodies, and John Boothroyd for anti-SAG1 and anti-RON4 antibodies. We also thank members of the Bradley lab for helpful discussions and comments on the manuscript. The rabbit polyclonal RON2 antisera was made in John Boothroyd\'s lab (supported by NIH grant AI21423). The authors have declared that no competing interests exist. This work was supported by a Ruth L. Kirschstein Natural Research Service Award (GM07185) and a UCLA Warsaw Fellowship to K.W.S., a Microbial Pathogenesis Training Grant (T32-AI07323) to B.E.H., an American Heart Association Western States Affiliate postdoctoral fellowship to J.S.T., and an NIH RO1 (AI064616) to P.J.B. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: KWS, PJB. Performed the experiments: KWS, EDP, BEH. Analyzed the data: KWS, EDP, BEH, PJB. Contributed reagents/materials/analysis tools: JST. Wrote the paper: KWS, PJB.

PubMed Central

2024-06-05T04:04:19.706805

2011-3-10

PMC3053351

Introduction {#s1} ============ Natural killer (NK) cells are able to lyse infected or malignant cells without prior antigenic stimulation, and thus provide an important innate defense against infectious agents and tumors [@ppat.1001316-Lanier1], [@ppat.1001316-Parham1]. NK cell activation in primates is regulated in part through interactions between the highly polymorphic killer immunoglobulin-like receptors (KIRs) expressed on NK cells and their MHC class I ligands on target cells [@ppat.1001316-Lanier1], [@ppat.1001316-Parham1]. KIRs are type I integral membrane proteins with either two or three immunoglobulin (Ig)-like extracellular domains (2D or 3D) that transduce either inhibitory or activating signals via long (L) or short (S) cytoplasmic domains, respectively. Engagement of inhibitory KIRs by MHC class I molecules on healthy cells normally suppresses NK cell activation [@ppat.1001316-Lanier1], [@ppat.1001316-Kim1], [@ppat.1001316-Valiante1]. However, if these interactions are perturbed, for instance as a result of MHC class I downregulation by HIV-1 Nef [@ppat.1001316-Schwartz1], [@ppat.1001316-Cohen1], or presentation of a peptide antagonist [@ppat.1001316-Fadda1], this inhibition is lost resulting in NK cell activation and target cell lysis. In contrast to the T cell receptor, which is highly specific for a given peptide-MHC complex, KIRs typically recognize subsets of MHC class I molecules with common amino acid motifs in their α1 domains. Based on serological epitopes that correspond to defined sequences at positions 77-83, all HLA-B molecules, and some HLA-A molecules, can be classified as either Bw4 or Bw6 allotypes [@ppat.1001316-Ayers1]. Allotypes of KIR3DL1 have broad specificity for HLA-Bw4 ligands [@ppat.1001316-Gumperz1], whereas KIRs specific for HLA-Bw6 have not been identified. All inhibitory KIRs that have been examined thus far also exhibit selectivity for peptides bound by their MHC class I ligands [@ppat.1001316-Mandelboim1], [@ppat.1001316-Thananchai1], [@ppat.1001316-Hansasuta1], [@ppat.1001316-Rajagopalan1], [@ppat.1001316-Zappacosta1], [@ppat.1001316-Peruzzi1], [@ppat.1001316-Malnati1]. These observations are consistent with crystal structures of KIR2DL1 and KIR2DL2 in complex with their HLA-C ligands showing that KIR residues contact surfaces of the HLA class I α1 and α2 domains in an orthogonal orientation across C-terminal residues of the bound peptide [@ppat.1001316-Fan1], [@ppat.1001316-Boyington1]. However, the molecular basis for the selectivity of KIRs for different peptides bound by a particular MHC class I ligand has not been defined. Genetic evidence suggests that polymorphic differences in the *KIR* and *HLA class I* genes play an important role in determining the course of infection for a number of human viral pathogens, including HIV-1 [@ppat.1001316-Martin1], [@ppat.1001316-Martin2], hepatitis C virus [@ppat.1001316-Khakoo1], human papillomavirus [@ppat.1001316-Carrington1] and cytomegalovirus [@ppat.1001316-Chen1]. In the case of HIV-1, combinations of both activating and inhibitory *KIR3DL1/S1* and *HLA-Bw4* alleles have been associated with delayed progression to AIDS [@ppat.1001316-Martin1], [@ppat.1001316-Martin2]. NK cells expressing KIR3DS1 were also shown to suppress the *in vitro* replication of HIV-1 in target cells expressing HLA-Bw4 [@ppat.1001316-Alter1]. While these observations point to a role for KIR-MHC class I interactions in determining the outcome of HIV-1 infection, studies to address the functional significance of these interactions have been limited, in part, by the lack of a suitable animal model. Simian immunodeficiency virus (SIV) infection of the rhesus macaque is an important animal model for lentiviral pathogenesis and for AIDS vaccine development [@ppat.1001316-Letvin1]. Rhesus macaques express MHC class I molecules that correspond to products of the classical *HLA-A* and *-B* genes (*[Ma]{.underline}caca [mu]{.underline}latta*; *Mamu-A* and *-B*), but not the *HLA-C* gene [@ppat.1001316-Boyson1], [@ppat.1001316-Adams1]. Consistent with the co-evolution of KIR and MHC class I molecules, genes for the two-domain KIRs specific for HLA-C have not been identified in macaques [@ppat.1001316-Hershberger1], [@ppat.1001316-Bimber1]. Instead, macaques have an expanded repertoire of *KIR3DL* genes characterized by extensive polymorphism and gene duplication [@ppat.1001316-Hershberger1], [@ppat.1001316-Bimber1], [@ppat.1001316-Sambrook1], [@ppat.1001316-Blokhuis1], [@ppat.1001316-Kruse1]. Here we identify Mamu-A1\*00201, a common rhesus macaque MHC class I molecule with a Bw6 motif, as a ligand for multiple allotypes of Mamu-KIR3DL05. We show that the binding of Mamu-KIR3DL05 to Mamu-A1\*00201 is peptide-dependent, and that the relative avidity and peptide-selectivity of binding is determined by polymorphisms in the D0 and D1 domains. We also demonstrate that target cells expressing Mamu-A1\*00201 suppress the degranulation of primary Mamu-KIR3DL05^+^ NK cells. These observations reveal a previously unappreciated role for D1 polymorphisms in determining the selective recognition of MHC class I-bound peptides by KIRs, and define the first functional KIR-MHC class I interaction in the rhesus macaque. Results {#s2} ======= Peptide-dependent tetramer staining of primary rhesus macaque NK cells {#s2a} ---------------------------------------------------------------------- Samples of peripheral blood from *Mamu-A1\*00201^+^* rhesus macaques were stained with Mamu-A1\*00201 tetramers folded with SIV peptides to establish baseline CD8^+^ T cell responses prior to beginning a vaccine study. To our surprise, Mamu-A1\*00201 in complex with the Gag~71-79~ GY9 peptide stained a subset of CD8^+^CD3^--^ lymphocytes from one animal (Mm 337-07). Plasma from this animal tested negative for SIV RNA and for antibodies to viral antigens, indicating that this animal had not been previously exposed to SIV. The majority of tetramer-positive cells expressed CD8α and CD16, characteristic of NK cells that are capable of mediating cytolytic activity [@ppat.1001316-Reeves1], as well as additional NK cell markers including NKp46, NKG2A, and NKG2D ([Fig. 1A](#ppat-1001316-g001){ref-type="fig"}). A subset of these cells also cross-reacted with an antibody to human KIR2D ([Fig. 1A](#ppat-1001316-g001){ref-type="fig"}). Although most of the tetramer-positive cells were CD16^+^CD3^--^ NK cells, staining was also observed for CD8^+^CD3^+^ T cells ([Fig. 1B](#ppat-1001316-g001){ref-type="fig"}). A longitudinal comparison of the frequency of tetramer-positive CD8^+^CD3^+^ versus CD16^+^CD3^--^ lymphocytes revealed that these two populations were relatively stable in this animal over more than a year, ranging from 0.16% to 0.69% for CD8^+^ T cells and from 5.1% to 9.8% for CD16^+^ NK cells. To investigate the contribution of the peptide bound by Mamu-A1\*00201 to this unusual pattern of tetramer staining, whole blood was stained with Mamu-A1\*00201 tetramers folded with peptides corresponding to eight different CD8^+^ T cell epitopes of SIV [@ppat.1001316-Loffredo1]. In addition to Gag~71-79~ GY9, staining was also observed for Env~788-795~ RY8, but not for any of the other tetramers ([Fig. 1C](#ppat-1001316-g001){ref-type="fig"}). Thus, the tetramer staining observed for primary NK cells and CD8^+^ T cells from Mm 337-07 was dependent on the peptide bound by Mamu-A1\*00201. ::: {#ppat-1001316-g001 .fig} 10.1371/journal.ppat.1001316.g001 Figure 1 ::: {.caption} ###### Peptide-dependent tetramer staining of primary NK cells and CD8^+^ T cells from an unimmunized, uninfected rhesus macaque. \(A) To identify the tetramer-positive cells, whole blood was stained with Mamu-A1\*00201 Gag~71-79~ GY9 followed by monoclonal antibodies to the indicated cell type-specific markers. (B) The frequency of tetramer-positive CD8^+^ T cells versus CD16^+^ NK cells was determined by gating sequentially on CD8 followed by CD3 (I) or CD16 (II). (C) To determine if tetramer staining is dependent on the peptide bound by Mamu-A1\*00201, whole blood was stained with Mamu-A1\*00201 tetramers folded with eight different SIV peptides, Gag~71-79~ GY9 (GSENLKSLY), Env~788-795~ RY8 (RTLLSRVY), Env~317-325~ KM9 (KTVLPVTIM), Nef~248-256~ LM9 (LTARGLLNM), Nef~159-167~ YY9 (YTSGPGIRY), Env~296-304~ RY9 (RTIISLNKY), Vif~97-104~ WY8 (WTDVTPNY), and Vif~89-97~ IW9 (ITWYSKNFW), followed by monoclonal antibodies to CD3, CD8 and CD16. After gating on CD8^+^ lymphocytes, the percentages of tetramer-positive CD3^--^ versus CD3^+^ cells were determined. :::  ::: Identification of Mamu-KIR3DL05 as a receptor for Mamu-A1\*00201 {#s2b} ---------------------------------------------------------------- Since KIRs are known to be expressed on subsets of human NK cells and CD8^+^ T cells [@ppat.1001316-Valiante1], [@ppat.1001316-Mingari1], [@ppat.1001316-Anfossi1], we hypothesized that this pattern of tetramer staining might reflect Mamu-A1\*00201 binding to a rhesus macaque KIR. Full-length *KIR* cDNA sequences were therefore cloned from the PBMCs of Mm 337-07 and sequenced. Six *KIR3DL* alleles, three *KIR3DS* alleles and two *KIR2DL04* alleles were identified in this animal, and their predicted amino acid sequences are shown in [Fig. 2](#ppat-1001316-g002){ref-type="fig"}. ::: {#ppat-1001316-g002 .fig} 10.1371/journal.ppat.1001316.g002 Figure 2 ::: {.caption} ###### Amino acid sequence alignments for KIR alleles cloned from a rhesus macaque with a tetramer-positive NK cell population in peripheral blood. The predicted amino acid sequences are shown for six *Mamu-KIR3DL* alleles (A), three *Mamu-KIR3DS* alleles (B), and two *Mamu-KIR2DL04* alleles (C). Positions of amino acid identity with the consensus sequence shown at the top are indicated by a period and translational stop sites are indicated with an asterisk. :::  ::: To identify the receptor for Mamu-A1\*00201, Jurkat cells were transfected with constructs expressing each of the *KIR* alleles cloned from Mm 337-07 and stained with Mamu-A1\*00201 tetramers. To differentiate transfected from untransfected cells, the *KIR* alleles were expressed from a bicistronic vector that co-expresses enhanced green fluorescent protein (eGFP). Since not all KIRs are well expressed on the cell surface, and antibodies are not available to macaque KIRs, an HA tag was introduced at the N-terminus of the D0 domain of each KIR. Our rationale for introducing the HA tag at this position is based on a recent three-dimensional model showing that the N-terminus of KIR3DL1 is free and oriented away from surfaces that are predicted to contact the peptide-MHC class I complex [@ppat.1001316-Sharma1], and experiments demonstrating that the introduction of an epitope tag at the N-terminus of the D0 domain does not interfere with ligand recognition [@ppat.1001316-Carr1]. Following the electroporation of Jurkat cells with these KIR expression constructs, the cells were stained with Mamu-A1\*00201 tetramers and with a monoclonal antibody to the HA tag. Transfected cells were identified by gating on the eGFP^+^ population, the surface expression of each KIR was verified by HA staining, and binding to Mamu-A1\*00201 in complex with Gag~71-79~ GY9 versus Nef~159-167~ YY9 was assessed by tetramer staining. All of the KIRs were expressed on the cell surface under the conditions of this assay, as indicated by HA staining ([Fig. 3](#ppat-1001316-g003){ref-type="fig"}). However, only Mamu-KIR3DL05\*008 resulted in a detectable level of staining with the Gag~71-79~ GY9 tetramer ([Fig. 3A](#ppat-1001316-g003){ref-type="fig"}). At higher levels of surface expression, staining was also observed for Nef~159-167~ YY9, indicating that this tetramer can bind to Mamu-KIR3DL05\*008 under conditions of protein over expression ([Fig. 3A](#ppat-1001316-g003){ref-type="fig"}). These results identify Mamu-KIR3DL05\*008 as a receptor for Mamu-A1\*00201, and indicate that the peptide bound by Mamu-A1\*00201 can modulate this interaction. ::: {#ppat-1001316-g003 .fig} 10.1371/journal.ppat.1001316.g003 Figure 3 ::: {.caption} ###### Mamu-A1\*00201 is a ligand for Mamu-KIR3DL05. Jurkat cells were transfected with constructs expressing each of the six *Mamu-KIR3DL* (A), three *Mamu-KIR3DS* (B), and two *Mamu-KIR2DL04* (C) alleles cloned from Mm 337-07, and stained with Gag~71-79~ GY9 and Nef~159-167~ YY9 tetramers. The KIRs were expressed from a bicistronic vector designed to introduce a common leader peptide followed by an HA tag at the N-terminus of the D0 domain, and to co-express eGFP from a downstream internal ribosomal entry site. Jurkat cells were electroporated with the KIR expression constructs and stained the following day with APC-conjugated Mamu-A1\*00201 tetramers, followed by a PE-conjugated antibody to the HA tag. Tetramer versus HA staining was determined after gating on the eGFP^+^ cell population. Quadrant gates were set using empty vector-transfected controls stained with tetramer and antibody to the HA tag. :::  ::: Polymorphic differences among allotypes of Mamu-KIR3DL05 modulate the relative avidity and peptide-selectivity of binding to Mamu-A1\*00201 {#s2c} ------------------------------------------------------------------------------------------------------------------------------------------- Phylogenetic comparisons of macaque *KIR3DL* sequences revealed that *Mamu-KIR3DL05\*008* belongs to a group of similar alleles found in both rhesus and cynomolgus macaques [@ppat.1001316-Bimber1]. To determine if other allotypes of Mamu-KIR3DL05 could also bind to Mamu-A1\*00201, Jurkat cells were transfected with constructs expressing six additional *Mamu-KIR3DL05* alleles, as well as six *Mamu-KIR3DL07* alleles. The transfected cells were then stained with Mamu-A1\*00201 tetramers folded with four different SIV peptides to assess binding; Gag~71-79~ GY9, Env~788-795~ RY8, Nef~159-167~ YY9 and Vif~89-97~ IW9. One, or more, of the Mamu-A1\*00201 tetramers bound to cells expressing each of the *Mamu-KIR3DL05* alleles. Cells expressing Mamu-KIR3DL05\*004, -KIR3DL05\*003, -KIR3DL05\*010, -KIR3DL05\*008 and -KIR3DL05\*005 stained with Gag~71-79~ GY9, Env~788-795~ RY8 and Nef~159-167~ YY9, whereas cells expressing Mamu-KIR3DL05\*001 and mmKIR3DL05x stained only with Gag~71-79~ GY9 or Nef~159-167~ YY9 ([Fig. 4A](#ppat-1001316-g004){ref-type="fig"}). In contrast, none of these KIRs bound to the Vif~89-97~ IW9 tetramer ([Fig. 4A](#ppat-1001316-g004){ref-type="fig"}). Furthermore, none of the *Mamu-KIR3DL07* alleles resulted in a detectable level of staining for any of the Mamu-A1\*00201 tetramers ([Fig. S1](#ppat.1001316.s001){ref-type="supplementary-material"}). Hence, this interaction is dependent on the peptide bound by Mamu-A1\*00201 and is specific for Mamu-KIR3DL05. ::: {#ppat-1001316-g004 .fig} 10.1371/journal.ppat.1001316.g004 Figure 4 ::: {.caption} ###### Polymorphisms in the D0 and D1 domains of Mamu-KIR3DL05 modulate the avidity and peptide-selectivity of binding to Mamu-A1\*00201. \(A) Jurkat cells were transfected with HA-tagged Mamu-KIR3DL05 expression constructs, and stained the next day with APC-conjugated Mamu-A1\*00201 tetramers (Gag~71-79~ GY9, Env~788-795~ RY8, Nef~159-167~ YY9 or Vif~89-97~ IW9), followed by a PE-conjugated antibody to the HA tag. Quadrant gates were set using empty vector-transfected controls stained with tetramer and antibody to the HA tag. (B) An alignment comparing the predicted amino acid sequences of the D0, D1 and D2 domains for seven different *Mamu-KIR3DL05* alleles. Positions of amino acid identity with the consensus sequence are indicated by a period. The shaded regions correspond to loops predicted to contact surfaces of the peptide-MHC class I complex [@ppat.1001316-Sharma1]. The plus signs beneath the alignment indicate unique residues in the D1 domain of mmKIR3DL05x that coincide with, or are immediately adjacent to, predicted MHC class I-contact loops. :::  ::: Mamu-KIR3DL05\*003, -KIR3DL05\*008 and -KIR3DL05\*010, were indistinguishable in their pattern of tetramer staining ([Fig. 4A](#ppat-1001316-g004){ref-type="fig"}). This is reflected by the similarity in their values for the mean fluorescence intensity (MFI) of tetramer staining divided by the MFI of HA staining, which are provided in [Table 1](#ppat-1001316-t001){ref-type="table"} as a quantitative comparison of tetramer binding corrected for differences in surface expression for each KIR. In accordance with the rank order of tetramer staining observed for primary NK cells ([Fig. 1C](#ppat-1001316-g001){ref-type="fig"}), staining was highest for Gag~71-79~ GY9, followed by Env~788-795~ RY8, and then Nef~159-167~ YY9 ([Fig. 4A](#ppat-1001316-g004){ref-type="fig"} and [Table 1](#ppat-1001316-t001){ref-type="table"}). With the exception of a single amino acid difference in the first position of the D0 domain of Mamu-KIR3DL05\*010, each of these KIRs have identical Ig-like domains ([Fig. 4B](#ppat-1001316-g004){ref-type="fig"}). ::: {#ppat-1001316-t001 .table-wrap} 10.1371/journal.ppat.1001316.t001 Table 1 ::: {.caption} ###### Relative binding of Mamu-A1\*00201 tetramers folded with four different SIV peptides to seven allotypes of Mamu-KIR3DL05. ::: {#ppat-1001316-t001-1} Mamu-A1\*00201 tetramer -------------------- ------------------------- ------ ------ ----- Mamu-KIR3DL05\*004 1.99 0.14 0.13 --- Mamu-KIR3DL05\*003 1.57 0.06 0.03 --- Mamu-KIR3DL05\*010 1.55 0.07 0.03 --- Mamu-KIR3DL05\*008 1.49 0.07 0.03 --- Mamu-KIR3DL05\*005 0.55 0.03 0.01 --- Mamu-KIR3DL05\*001 0.02 --- --- --- mmKIR3DL05x --- --- 0.10 --- Jurkat cells were transfected with constructs expressing HA-tagged allotypes of Mamu-KIR3DL05 and stained the next day with one of the following Mamu-A1\*00201 tetramers; Gag~71-79~ GY9, Env~788-795~ RY8, Nef~159-167~ YY9 or Vif~89-97~ IW9. The cells were then stained with a monoclonal antibody to the HA tag and analyzed by flow cytometry. Values represent the MFI of tetramer staining divided by the MFI of HA staining for the transfected, eGFP^+^ cell population. Values were not calculated if the number of tetramer-positive events was less than 1000. ::: Relative to Mamu-KIR3DL05\*008, Mamu-KIR3DL05\*004 exhibited an increase in the intensity of tetramer staining for Gag~71-79~ GY9 (1.3 fold), Env~788-795~ RY8 (2.0 fold) and Nef~159-167~ YY9 (4.3 fold) ([Fig. 4A](#ppat-1001316-g004){ref-type="fig"} and [Table 1](#ppat-1001316-t001){ref-type="table"}). Since the Ig-like domains of Mamu-KIR3DL05\*004 and Mamu-KIR3DL05\*008 only differ by a single amino acid at position 138 ([Fig. 4B](#ppat-1001316-g004){ref-type="fig"}), the histidine residue at this position accounts for the increase in Mamu-KIR3DL05\*004 binding to Mamu-A1\*00201. Based on a recently proposed three-dimensional model of KIR3DL1\*015 bound to HLA-A\*2402 [@ppat.1001316-Sharma1], this residue is predicted to lie at the base of the second MHC class I-contact loop of the D1 domain, and may alter the conformation of this loop in a way that enhances binding to Mamu-A1\*00201. Compared to Mamu-KIR3DL05\*008, decreases in the intensity of tetramer staining were observed for both Mamu-KIR3DL05\*005 and -KIR3DL05\*001 ([Fig. 4A](#ppat-1001316-g004){ref-type="fig"}). The intensity of staining for Mamu-KIR3DL05\*005, which differs from Mamu-KIR3DL05\*008 by 14 amino acids ([Fig. 4B](#ppat-1001316-g004){ref-type="fig"}), was 2.7-fold lower for Gag~71-79~ GY9, 2.3-fold lower for Env~788-795~ RY8 and 3.0-fold lower for Nef~159-167~ YY9 ([Table 1](#ppat-1001316-t001){ref-type="table"}). A much greater reduction in the intensity of tetramer staining was observed for Mamu-KIR3DL05\*001. Tetramer staining for Mamu-KIR3DL05\*001 was only detectable with Gag~71-79~ GY9 at an intensity that was 75-fold lower than for Mamu-KIR3DL05\*008 ([Table 1](#ppat-1001316-t001){ref-type="table"}). Since Mamu-KIR3DL05\*001 and -KIR3DL05\*008 differ by ten amino acids in D0, but are otherwise identical in D1 and D2 ([Fig. 4B](#ppat-1001316-g004){ref-type="fig"}), this reduction in the avidity of binding to Mamu-A1\*00201 is due to polymorphic differences in the D0 domain. Thus, similar to KIR3DL1-HLA-Bw4 interactions in humans [@ppat.1001316-Sharma1], [@ppat.1001316-Khakoo2], polymorphisms in the D0 domain of Mamu-KIR3DL05 can dramatically affect binding to MHC class I ligands. In the case of mmKIR3DL05x, tetramer staining was observed for Nef~159-167~ YY9, but not for Gag~71-79~ GY9 or Env~788-795~ RY8 ([Fig. 4A](#ppat-1001316-g004){ref-type="fig"}). This shift in the pattern of Mamu-A1\*00201 tetramer staining almost certainly reflects differences in D1, since mmKIR3DL05x has a unique D1 domain, but nearly identical D0 and D2 domains to other allotypes of Mamu-KIR3DL05 ([Fig. 4B](#ppat-1001316-g004){ref-type="fig"}). Using cryopreserved PBMCs from the original source of *mmKIR3DL05x*, we verified that *mmKIR3DL05x* represents a *bona fide* allele, and not a PCR artifact, by independently cloning and confirming the cDNA sequence for this allele, and by PCR amplification of a 2.0 kb region spanning intron 4 from genomic DNA with primers to unique sequences in exons 4 and 5. Additional sequence comparisons revealed that the D1 domain of mmKIR3DL05x, as well as the leader peptide and the D0 domain, are identical to Mamu-KIR3DS02\*00402 and mmKIR3DHa ([Fig. S2](#ppat.1001316.s002){ref-type="supplementary-material"}). Thus, *mmKIR3DL05x* appears to be the product of a recombination event in which exon 4 (encoding D1) was acquired, either by the introduction of exons 1-4 of a *Mamu-KIR3DS* gene into -*KIR3DL05* or by the introduction of exons 5-9 of *Mamu-KIR3DL05* into a *-KIR3DS* gene. A closer examination of mmKIR3DL05x revealed that seven of the thirteen differences in the D1 domain coincide with, or are immediately adjacent to, loops predicted to contact surfaces of the peptide-MHC class I complex [@ppat.1001316-Sharma1]. These include a charge difference at position 144 in the second loop (L2) and a cluster of six residues at positions 164--170 in the third loop (L3) ([Fig. 4B](#ppat-1001316-g004){ref-type="fig"} and [Fig. 5A](#ppat-1001316-g005){ref-type="fig"}). To determine if these differences account for the unique binding pattern exhibited by mmKIR3DL05x, we constructed recombinants in which these sequences were exchanged with the corresponding sequences of Mamu-KIR3DL05\*008, and tested them for binding to Gag~71-79~ GY9 versus Nef~159-167~ YY9. Reciprocal L2 substitutions affected the avidity, but not the specificity, of tetramer binding ([Fig. 5B](#ppat-1001316-g005){ref-type="fig"}). In contrast, exchanging L3 residues switched the specificity, and altered the avidity, of binding to the Mamu-A1\*00201 tetramers. The 3DL05\*008/xL3 recombinant bound Nef~159-167~ YY9, but not Gag~71-79~ GY9, and the 3DL05x/\*008L3 recombinant bound both Gag~71-79~ GY9 and Nef~159-167~ YY9 ([Fig. 5B](#ppat-1001316-g005){ref-type="fig"}). Hence, these results reveal a role for polymorphisms in the third predicted contact loop of the D1 domain in determining the selective recognition of different peptides bound by the same MHC class I molecule. ::: {#ppat-1001316-g005 .fig} 10.1371/journal.ppat.1001316.g005 Figure 5 ::: {.caption} ###### Amino acid differences in the third MHC class I-contact loop of the D1 domain account for the preferential binding of mmKIR3DL05x to the Nef~159-167~ YY9 tetramer. \(A) Positions in the D1 domain of mmKIR3DL05x that differ from Mamu-KIR3DL05\*008 are highlighted in a three-dimensional model of KIR3DL\*015 bound to HLA-A\*2402 [@ppat.1001316-Sharma1]. The residues indicated in yellow are located in surface-exposed loops in close proximity to the bound peptide. The residues indicated in red represent differences in the D1 domain at sites that do not contribute directly to interactions with MHC class I ligands. The residues highlighted in magenta represent positions 77-83 of the α1 domain corresponding to the Bw6 of Mamu-A1\*00201. (B) Jurkat cells were electroporated with constructs expressing recombinants of Mamu-KIR3DL05\*008 and mmKIR3DL05x, for which residues of the second (L2) and third (L3) predicted MHC class I-contact loops in D1 were exchanged, and stained with APC-conjugated tetramer (Gag~71-79~ GY9 or Nef~159-167~ YY9) followed by a PE-conjugated antibody to the HA tag. Tetramer versus HA staining was determined after gating on the eGFP^+^ cell population and quadrant gates were set using control cells transfected with an empty vector. :::  ::: Mamu-A1\*00201 suppresses the activation of tetramer-positive NK cells from Mamu-KIR3DL05^+^ macaques {#s2d} ----------------------------------------------------------------------------------------------------- Additional *Mamu-KIR3DL05^+^* rhesus macaques were identified by sequence-specific PCR and screened for tetramer-positive NK cells and CD8^+^ T cells. The Gag~71-79~ GY9 tetramer stained subsets of CD8^+^CD3^--^ and CD8^+^CD3^+^ lymphocytes in in peripheral blood from each of the *Mamu-KIR3DL05^+^* animals, but not from *Mamu-KIR3DL05^-^* animals ([Fig. 6A and 6B](#ppat-1001316-g006){ref-type="fig"}). In accordance with the complex regulation of KIR expression, which is influenced by a number of factors including differences in gene content on different *KIR* haplotypes, differences in the repertoire of *MHC class I* genes and polymorphic differences in *KIR* genes [@ppat.1001316-Parham2], [@ppat.1001316-Gardiner1], [@ppat.1001316-Shilling1], [@ppat.1001316-Yawata1], there was considerable animal-to-animal variation in the frequency and intensity of tetramer staining ([Fig. 6B](#ppat-1001316-g006){ref-type="fig"}). Variation in the frequency of tetramer-positive CD8^+^CD3^+^ lymphocytes, particularly in *Mamu-A1\*00201^−^* animals that do not have Gag~71-79~-specific CD8^+^ T cells, may also reflect changes in KIR expression on memory CD8^+^ T cells related to age and/or prior exposure to infectious agents, since similar changes have been associated with age and encounters with viral pathogens in humans [@ppat.1001316-Anfossi1], [@ppat.1001316-Alter2], [@ppat.1001316-Bonorino1]. Overall, these results demonstrate that the presence of the *Mamu-KIR3DL05* gene is predictive of Gag~71-79~ GY9 staining in peripheral blood, that this pattern of tetramer staining is independent of *Mamu-A1\*00201* and SIV infection, and that the variability of staining is typical of the heterogeneity of KIR expression on human NK cells and CD8^+^ T cells [@ppat.1001316-Anfossi1], [@ppat.1001316-Gumperz2]. ::: {#ppat-1001316-g006 .fig} 10.1371/journal.ppat.1001316.g006 Figure 6 ::: {.caption} ###### Mamu-A1\*00201^+^ target cells suppress the degranulation of tetramer-positive NK cells. \(A) Primers specific for exon 5 of *Mamu-KIR3DL05* were used to amplify a 156 bp sequence from genomic DNA. Primers specific for a conserved 300 bp region of *Mamu-DRB* were included as an internal control. PCR products were separated on a 1% agarose gel containing ethidium bromide and visualized by UV transillumination. (B) Peripheral blood from two *Mamu-KIR3DL05^−^* and eight *Mamu-KIR3DL05^+^* rhesus macaques was stained with the Gag~71-79~ GY9 and monoclonal antibodies to CD3, CD8 and CD16. The percentages of tetramer-positive cells were determined for CD3^-^CD8^+^ versus CD3^+^CD8^+^ lymphocytes. Of the eight *Mamu-KIR3DL05^+^* animals, four were *Mamu-A1\*00201^+^* (Mm 177-05, Mm 350-04, Mm R02020 and Mm R95117) and four were *Mamu-A1\*00201^-^* (Mm 20-05, Mm 376-04, Mm RHAX18 and Mm R03035). With the exception of Mm AP78 and Mm AD73INF08, which were negative for *Mamu-KIR3DL05*, tetramer-positive NK cells were detected in peripheral blood for each of these animals. Mm AP78, Mm AD73INF08, Mm 177-05, and Mm RHAX18 were uninfected at the time of this analysis. Mm R02020 and Mm R95117 were infected with SIV~smm~E660. Mm 350-04 and 376-04 were infected with SIV~mac~239Δnef. Mm 20-05 and Mm R03035 were infected with SIV~mac~239. The SIV-infected animals are indicated with an asterisk. (C) Freshly isolated PBMC from two *Mamu-A1\*00201^+^* and two *Mamu-A1\*00201^-^* macaques were stimulated with parental 721.221 cells, or 721.221 cells expressing individual rhesus macaque MHC class I molecules. Mm 337-07 was uninfected at the time of this analysis. The cells were incubated overnight at a 5:1 PBMC to target cell ratio in the presence of a monoclonal antibody to CD107a. Following stimulation, the cells were stained with the Gag~71-79~ GY9 tetramer and antibodies to CD3, CD8, CD16 and NKG2A. After gating on CD3^-^NK2GA^+^ lymphocytes, the upregulation of CD107a on tetramer-positive versus tetramer-negative NK cells was determined. :::  ::: The role of NK cells and CD8^+^ T cells that express Mamu-KIR3DL05 in SIV-infected animals remains to be determined. However, among the eight animals represented in [Fig. 6B](#ppat-1001316-g006){ref-type="fig"}, there were no obvious differences in the percentage of tetramer-positive cells for either population that could be attributed to the status of SIV infection. Indeed, of the two uninfected animals (Mm 177-05 and Mm RHAX18), the two animals infected with attenuated SIV~mac~239 Δ*nef* (Mm 350-04 and Mm 376-04), and the two animals infected with pathogenic SIV~mac~239 (Mm R03035 and Mm 20-05), each pair had among the lowest and the highest frequencies of tetramer-positive lymphocytes ([Fig. 6B](#ppat-1001316-g006){ref-type="fig"}). Some of the tetramer-positive CD8^+^CD3^+^ lymphocytes in the *Mamu-A1\*00201^+^* animals probably represent virus-specific CD8^+^ T cells, since we cannot differentiate binding of the Gag~71-79~ GY9 tetramer to Mamu-KIR3DL05 versus the T cell receptor. Nevertheless, in two of the three SIV-infected animals (Mm 350-04 and Mm R02020), the percentage of tetramer-positive cells was actually higher for the CD8^+^CD3^−^ population than for the CD8^+^CD3^+^ population ([Fig. 6B](#ppat-1001316-g006){ref-type="fig"}). Although the explanation for this is presently unclear, it is possible that Mamu-KIR3DL05 interactions with Mamu-A1\*00201 may suppress CD8^+^ T cell responses to the Gag~71-79~ GY9 epitope in *Mamu-KIR3DL05^+^* animals, which could explain the inconsistent, and often weak, CD8^+^ T cell responses to Gag~71-79~ GY9 that we and others have observed in SIV-infected, *Mamu-A1\*00201^+^* macaques. To investigate the functional consequences of NK cell recognition of Mamu-A1\*00201, PBMC from four *Mamu-KIR3DL05^+^* animals were stimulated with the MHC class I-deficient 721.221 cell line [@ppat.1001316-Shimizu1], or with 721.221 cells expressing either Mamu-A1\*00201, -A1\*01101 (Mamu-A\*11) or -B\*010101 (Mamu-B\*01), and stained for CD107a as a degranulation marker. The cells were also stained with Gag~71-79~ GY9 to differentiate Mamu-KIR3DL05^+^ NK cells from Mamu-KIR3DL05^−^ NK cells. CD107a was upregulated on the surface of both tetramer-positive and tetramer-negative NK cells in response to parental 721.221 cells and 721.221 cells expressing Mamu-A1\*01101 or -B\*010101 ([Fig. 6C](#ppat-1001316-g006){ref-type="fig"}). In contrast, CD107a was suppressed on tetramer-positive NK cells, but not on tetramer-negative NK cells, in the presence of target cells expressing Mamu-A1\*00201 ([Fig. 6C](#ppat-1001316-g006){ref-type="fig"}). The same pattern of NK cell activation/inhibition was also observed by intracellular cytokine staining for IFNγ ([Fig. S3](#ppat.1001316.s003){ref-type="supplementary-material"}). Moreover, CD107a was suppressed on tetramer-positive NK cells from both *Mamu-A1\*00201^+^* and *-A1\*00201^−^* animals ([Fig. 6C](#ppat-1001316-g006){ref-type="fig"}), indicating that these cells were responsive to Mamu-A1\*00201 whether or not they were educated in the presence of this MHC class I molecule. These results are therefore consistent with the functional inhibition of Mamu-KIR3DL05^+^ NK cells by Mamu-A1\*00201. Discussion {#s3} ========== Polymorphic differences in the *KIR* and *HLA class I* genes play an important role in determining the course of infection for HIV-1 and for a number of other viral pathogens [@ppat.1001316-Martin1], [@ppat.1001316-Martin2], [@ppat.1001316-Khakoo1], [@ppat.1001316-Carrington1], [@ppat.1001316-Chen1], [@ppat.1001316-Alter1]. However, studies to address the functional significance of KIR-MHC class I interactions have been hampered by the lack of a suitable animal model. In the present study, we identify Mamu-A1\*00201, an MHC class I molecule present in approximately 20% of Indian origin rhesus macaques [@ppat.1001316-Kaizu1], as a ligand for multiple allotypes of Mamu-KIR3DL05. Although the frequency of specific alleles of *Mamu-KIR3DL05* remains to be determined, the *Mamu-KIR3DL05* gene was present in 42% of the rhesus macaques (43 of 103 animals) recently screened at the New England Primate Research Center by sequence-specific PCR. This suggests that animals expressing both Mamu-KIR3DL05 and -A1\*00201 are sufficiently common for use in future studies to investigate the functional implications of this interaction with respect to the pathogenesis of SIV infection. Genotyping for *Mamu-KIR3DL05* was predictive of Mamu-A1\*00201 tetramer staining for primary NK cells and CD8^+^ T cells in peripheral blood. The pattern of staining observed for subsets of CD8^+^CD3^−^ and CD8^+^CD3^+^ lymphocytes from *Mamu-KIR3DL05* ^+^ animals, but not from *Mamu-KIR3DL05* ^−^ animals, is consistent with the variegated expression of KIRs on human NK cells and CD8^+^ T cells [@ppat.1001316-Valiante1], [@ppat.1001316-Mingari1], [@ppat.1001316-Anfossi1], [@ppat.1001316-Raulet1], [@ppat.1001316-Pascal1], [@ppat.1001316-Davies1]. Tetramer staining was independent of *Mamu-A1\*00201*, reflecting the segregation of *KIR* and *MHC class I* genes on different chromosomes, and was detectable regardless of the status of SIV infection. Moreover, the variability in the frequency and intensity of tetramer staining among *Mamu-KIR3DL05* ^+^ animals was typical of the heterogeneity of KIR expression on human NK cells and CD8^+^ T cells [@ppat.1001316-Anfossi1], [@ppat.1001316-Gumperz2]. Although tetramer staining has been reported for NK cell clones and for transfected cells expressing human KIRs [@ppat.1001316-Thananchai1], [@ppat.1001316-Hansasuta1], to our knowledge this is the first report of direct *ex vivo* tetramer staining of primary NK cells. Incubation of peripheral blood lymphocytes from *Mamu-KIR3DL05^+^* macaques with target cells expressing Mamu-A1\*00201 specifically suppressed the degranulation of tetramer-positive NK cells. These results are consistent with the functional inhibition of primary NK cells expressing Mamu-KIR3DL05 by Mamu-A1\*00201. Furthermore, this inhibition was observed for tetramer-positive NK cells from *Mamu-A1\*00201^−^* as well as from *Mamu-A1\*00201^+^*animals, indicating that these cells were responsive to Mamu-A1\*00201, whether or not they were educated in animals that express this ligand. Although the mechanisms of NK cell education are not fully understood [@ppat.1001316-Hoglund1], there is evidence that the maturation of NK cells expressing inhibitory KIRs is dependent on interactions with self-MHC class I molecules, and that NK cells expressing a particular inhibitory KIR in the absence of an appropriate MHC class I ligand are rendered hyporesponsive [@ppat.1001316-Kim1], [@ppat.1001316-Kim2], [@ppat.1001316-Anfossi2]. Thus, the *in vitro* suppression of tetramer-positive NK cells from *Mamu-A1\*00201^−^* animals by target cells expressing Mamu-A1\*00201 implies that these cells were educated for recognition of another MHC class I ligand. This is perhaps not surprising given the complexity of the rhesus macaque MHC class I genes [@ppat.1001316-Wiseman1], [@ppat.1001316-Otting1], and the ability of KIRs to recognize multiple MHC class I ligands with common amino acid motifs in their α1 domains [@ppat.1001316-Gumperz1], [@ppat.1001316-Mandelboim2]. Based on haplotype modeling and phylogenetic comparisons, *Mamu-KIR3DL05* is predicted to represent a single genetic locus [@ppat.1001316-Bimber1]. Although *KIR3DL05* is not orthologous to any of the human *KIR* genes, interactions between Mamu-KIR3DL05 and Mamu-A1\*00201 resemble features of KIR3DL1 binding to HLA-Bw4. A three-dimensional model of KIR3DL1\*015 bound to HLA-A\*2402 was recently constructed based on a crystal structure of KIR2DL1 in complex with HLA-C\*04 [@ppat.1001316-Fan1], [@ppat.1001316-Sharma1]. This model predicts that surface-exposed loops in each of the three Ig-like domains of KIR3DL1 contact the HLA class I molecule over the C-terminus of the bound peptide, and that the specificity of KIR3DL1 for HLA-Bw4 is dependent on a salt bridge between glutamate 282 in the D2 domain of KIR3DL1 and arginine 83 in the α1 domain of HLA-Bw4 [@ppat.1001316-Sharma1]. Consistent with this model, polymorphisms in the Ig-like domains of Mamu-KIR3DL05 were associated with differences in binding to Mamu-A1\*00201. Amino acid differences in D0 affected the relative avidity of Mamu-KIR3DL05 binding to Mamu-A1\*00201. Compared to Mamu-KIR3DL05\*003/\*008, tetramer staining was diminished for both Mamu-KIR3DL05\*001 and -KIR3DL05\*005, which differ by eight and ten residues in D0 respectively. In the case of Mamu-KIR3DL05\*001, which is otherwise identical to Mamu-KIR3DL05\*003/\*008 in D1 and D2, binding to Mamu-A1\*00201 was all but eliminated. These results are analogous to previous observations showing that polymorphisms in the D0 domain of KIR3DL1 modulate the avidity of binding to HLA-Bw4 ligands [@ppat.1001316-Sharma1], [@ppat.1001316-Khakoo2]. Polymorphisms in D1 altered the selective binding of Mamu-KIR3DL05 to Mamu-A1\*00201 in complex with different SIV peptides. In contrast to other allotypes of Mamu-KIR3DL05, mmKIR3DL05x preferentially bound to Mamu-A1\*00201 folded with Nef~159-167~ YY9 rather than Gag~71-79~ GY9. This difference in peptide preference mapped to six amino acids in the third D1 loop predicted to contact surfaces of the peptide-MHC class I complex. These results support a recent three-dimensional model of KIR3DL1\*015 bound to HLA-A\*2402 [@ppat.1001316-Sharma1], and reveal a role for polymorphisms in the D1 domain in determining the selectivity of KIRs for MHC class I-bound peptides. Interestingly, mmKIR3DL05x appears to be the product of a recombination event in which exon 4 sequences coding for the D1 domain were derived from a *KIR3DS* gene; an observation that is consistent with domain shuffling as a mechanism of KIR evolution in primates [@ppat.1001316-Rajalingam1]. Unlike previously identified ligands for human KIRs, the α1 domain of Mamu-A1\*00201 contains a Bw6 motif. In contrast to Bw4, the Bw6 motif has a glycine rather than an arginine at position 83 (N~77~LRNLRG~83~). Yet, Mamu-KIR3DL05 retains a glutamate at position 285, which corresponds to glutamate 282 of KIR3DL1. Since the peptides recognized by Mamu-KIR3DL05 each contain a positively charged residue at position 6 or 8 (Gag~71-79~ GSENL[K]{.underline}SLY, Env~788-795~ RTLLS[R]{.underline}VY and Nef~159-167~ YTSGPGI[R]{.underline}Y), it is conceivable that glutamate 285 may form an alternative salt bridge with the peptide that accounts for the peptide-dependence of Mamu-KIR3DL05. However, a charge at this position does not appear to be sufficient for binding, since the Vif~89-97~ IW9 peptide, which also contains a lysine at position 6 (ITWYS[K]{.underline}NFW), did not result in detectable Mamu-A1\*00201 tetramer staining. While the molecular interactions underlying the binding of Mamu-KIR3DL05 to Mamu-A1\*00201 remain to be fully defined, these observations offer a potential explanation for the contribution of the peptide to this interaction, and perhaps suggest a more prominent role for certain peptides in KIR recognition of other Bw6 ligands. The extent to which KIR recognition of Bw6 ligands has been elaborated in the rhesus macaque is presently unclear. However, since this motif is retained in the MHC class I molecules of humans and macaques, the absence of human KIRs that recognize HLA-Bw6 appears to reflect the loss of receptors of this specificity during the course of human evolutionary history. While the reason for this is not understood, it may be related to the expansion of the lineage III *KIR* genes coding for KIR2DL/S receptors with a D1-D2 configuration, and a greater dependence on the regulation of NK cell activation through interactions with their HLA-C ligands. The identification of inhibitory KIRs that bind with high avidity to a common MHC class I molecule in the rhesus macaque in complex with SIV-derived peptides suggests a potential mechanism of immune evasion. The Nef proteins of HIV-1 and SIV selectively downregulate MHC class I molecules from the surface of infected cells to evade destruction by virus-specific CD8^+^ T cells [@ppat.1001316-Cohen1], [@ppat.1001316-DeGottardi1]. However, the removal of these molecules from the cell surface increases the susceptibility of infected cells to elimination by NK cells [@ppat.1001316-Cohen1]. By acquiring changes in CD8^+^ T cell epitopes that increase the binding of MHC class I ligands to inhibitory KIRs, the virus may prevent the activation of NK cells under conditions of incomplete downregulation by Nef. This possibility is supported by recent evidence that peptides can modulate NK cell activation by varying the affinity of HLA ligands for inhibitory KIRs [@ppat.1001316-Fadda1]. Whereas Fadda *et al.* show that antagonistic peptides that disrupt MHC class I interactions with inhibitory KIRs leads to NK cell activation [@ppat.1001316-Fadda1], our data suggests that viruses may acquire changes in epitopes that stabilize these interactions to suppress NK cell activation in a way that favors virus replication. KIRs are also expressed on subsets of memory CD8^+^ T cells in HIV-1 infected individuals, and have been associated with a decrease in the responsiveness to TCR-dependent stimulation [@ppat.1001316-Alter2], [@ppat.1001316-Maria1]. Thus, peptides that stabilize interactions with inhibitory KIRs may also suppress CD8^+^ T cell activation. Deleterious combinations of *KIR* and *MHC class I* alleles may therefore select for changes in epitopes of HIV-1 and SIV that inhibit certain NK cell and CD8^+^ T cell responses; a scenario that may further undermine the host\'s ability to contain virus replication. Consistent with this hypothesis, a single nucleotide polymorphism was recently identified as a marker for two *Mamu-KIR3DL05* alleles that were more prevalent among SIV-infected rhesus macaques with high viral loads in animals [@ppat.1001316-Bostik1]. The identification of Mamu-A1\*00201 as a ligand for Mamu-KIR3DL05 now affords an opportunity to investigate the functional implications of KIR-MHC class I interactions. Using *KIR-* and *MHC class I*-defined animals, experiments can now be designed to examine the phenotypic changes that occur in a specific population of NK cells during the course of virus infection in a way the was previously only possible for CD8^+^ T cells. Characterization of the molecular interactions underlying the binding of Mamu-KIR3DL05 to Mamu-A1\*00201 also promises to yield fundamental insights regarding the role of viral peptides in modulating KIR recognition of MHC class I ligands. The binding of Mamu-KIR3DL05 to Mamu-A1\*00201 in complex with SIV peptides suggests that these interactions may be particularly important in determining the course of SIV infection. Materials and Methods {#s4} ===================== Ethics statement {#s4a} ---------------- All of the animals used for these studies were Indian origin rhesus macaques *(Macaca mulatta)*. These animals were housed at the New England Primate Research Center (NEPRC) and were maintained in accordance with standards of the Association for Assessment and Accreditation of Laboratory Animal Care and the Harvard Medical School Animal Care and Use Committee. Animal experiments were approved by the Harvard Medical Area Standing Committee on Animals and conducted according to the principles described in the *Guide for the Care and Use of Laboratory Animals* [@ppat.1001316-Anonymous1]. KIR nomenclature and Genbank accession numbers {#s4b} ---------------------------------------------- Rhesus macaque *KIR* sequences were submitted to Genbank and to the Immuno-Polymorphism Database ([www.ebi.ac.uk/ipd/kir/](http://www.ebi.ac.uk/ipd/kir/)) [@ppat.1001316-Robinson1]. Sequences that have been assigned official names are indicated with the prefix *Mamu*-*KIR*. In cases where official names have not yet been assigned, sequences are referred to using a provisional nomenclature indicated by the prefix *mmKIR*. The names and Genbank accession numbers for each of the *KIR* alleles in this study are listed in [Table S1](#ppat.1001316.s004){ref-type="supplementary-material"}. Phenotypic analysis of tetramer-positive lymphocytes {#s4c} ---------------------------------------------------- Whole blood was stained with Mamu-A1\*00201 tetramers folded with the SIV peptides Gag~71-79~ GY9, Env~788-795~ RY8, Env~317-325~ KM9, Nef~248-256~ LM9, Nef~159-167~ YY9, Env~296-304~ RY9, Vif~97-104~ WY8, or Vif~89-97~ IW9 (30 min, 37°C) followed by antibodies to cell type-specific markers (30 min, 20°C). Mamu-A1\*00201 tetramers were obtained from David Watkins\' laboratory (Wisconsin National Primate Research Center), and the quality of each tetramer lot was verified by staining CD8^+^ T lymphocytes from SIV-infected rhesus macaques. For polychromatic assays, samples were stained with anti-CD3-Pacific blue (SP34-2, BD Pharmingen), anti-CD4-AmCyan (L200, BD Pharmingen), anti-CD16-FITC (3G8, BD Pharmingen), anti-HLA-DR-PE Texas Red (Immu-257, Immunotech), anti-CD20 PE-Cy5.5 (L27, BD Pharmingen), anti-CD56 PE-Cy7 (NCAM16.2, BD Pharmingen), anti-CD8α-Alexa 700 (RPA-T8, BD Pharmingen), anti-CD14-APC-Cy7 (MphiP9, BD Pharmingen), and either anti-NKG2A-PE (Z1999, Beckman Coulter), anti-NKp46-PE (BAB21, Immunotech), anti-KIR2D-PE (NKVFS1, Miltenyi Biotec Inc.), or anti-NKG2D-PE (BAT221, Miltenyi Biotec Inc.) For four-color assays, samples were stained with anti-CD3-FITC (SP34-2, BD Pharmingen), anti-CD16-PE (3G8, BD Pharmingen), and anti-CD8α-PerCP (SK1, BD Pharmingen). Samples were treated with FACS Lysing solution (BD Biosciences) to eliminate red blood cells, washed and fixed in 2% paraformaldehyde PBS. Data was acquired using a LSRII flow cytometer (BD Biosciences) and analyzed using FlowJo 8.8.6 (Tree Star Inc.). Cloning and sequencing of rhesus macaque KIR alleles {#s4d} ---------------------------------------------------- Peripheral blood lymphocytes were isolated over Ficoll (Sigma) and aliquots of 2--10 million PBMC were frozen in Trizol (Invitrogen). Total RNA was extracted using the RNeasy kit (Qiagen) according to the manufacturer\'s instructions. KIR cDNAs were amplified by reverse transcription-polymerase chain reaction (RT-PCR) using the Superscript III One-Step RT-PCR kit (Invitrogen) with modified versions of the Ig3Up and Ig3Down primers [@ppat.1001316-Pende1]. Cycling conditions included an RT step at 55°C for 30 min, a denaturation step at 94°C for 2 min, followed by 40 cycles of denaturation (94°C for 15 sec), annealing (55°C for 30 sec) and extension (68°C for 90 sec), and a final extension step at 68°C for 5 min. PCR products were cloned into the pGEM-T Easy vector (Promega) and sequenced with T7 and SP6 sequencing primers. Sequences were analyzed using Sequencher 4.8 (Gene Codes Inc.) and MacVector 9.5.2 (MacVector Inc.) software packages. At least three identical cDNA clones were identified for each *KIR* allele. KIR expression and tetramer staining of transfected Jurkat cells {#s4e} ---------------------------------------------------------------- Rhesus macaque *KIRs* were PCR amplified from cDNA clones using primers to introduce an HA tag at the N-terminus of the D0 domain. The *KIR* cDNAs were then cloned into pCGCG, a bicistronic vector that co-expresses eGFP, in frame with an upstream sequence for the leader peptide of Mamu-KIR3DL05\*008. Jurkat cells (1×10^7^ cells) were electroporated (250V, 975µF) with plasmid DNA (40 µg) in serum-free RPMI (400 µl) in a 0.4 cm cuvette (BioRad). After resting (10 min, 20°C), the cells were re-suspended in RPMI medium (9 ml) with 10% FBS and incubated overnight at 37°C, 5% CO~2~. After 22 hours, the cells were stained with APC-conjugated tetramers (30 min, 37°C), followed by PE-conjugated anti-HA PE (GG8-IF3.3, Miltenyi Biotec Inc.) (20 min, 20°C). The cells were washed and fixed in 2% paraformaldehyde PBS. At least 200,000 events were acquired using a FACSCalibur flow cytometer (BD Biosciences) and the data was analyzed using FlowJo 8.8.6. Mamu-KIR3DL05 genotyping {#s4f} ------------------------ Genomic DNA was extracted from 1--2 million PBMC using the DNAeasy kit (Qiagen, Valencia, CA), and 10 ng was used as template in a 25 µl PCR reaction with forward and reverse primers (GAGACCCATGAACTTAGGCTTC & GCAGTGGGTCACTGGGGA) for amplification of a 156 bp sequence in exon 5 specific to *Mamu-KIR3DL05*. Primers specific for a conserved 300 bp region of *Mamu-DRB* were included as an internal control [@ppat.1001316-Kaizu1]. Cycling conditions included a denaturation step at 96°C for 2 min followed by 30 cycles of denaturation (94°C for 30 sec), annealing (63°C for 45 sec) and extension (72°C for 45 sec), and a final extension step at 72°C for 10 min. PCR products were separated on a 1% agarose gel containing ethidium bromide and visualized by UV transillumination. NK cell suppression by specific MHC class I molecules {#s4g} ----------------------------------------------------- PBMC (1×10^6^ cells) were stimulated for 18 hours with 721.221 cells, or with 721.221 cells expressing rhesus macaque MHC class I molecules, at a 5:1 ratio in the presence of anti-CD107a PE-Cy5 (clone H4A3, BD Pharmingen), Golgi-Stop and Golgi-plug (BD Pharmingen). The cells were then stained with APC-conjugated tetramers (30 min, 37°C), followed by anti-CD16-FITC, anti-NKG2A-PE, anti-CD8α-Alexa 700 and CD3 APC-Cy7 (20 min, 20°C). The cells were then permeabilized and stained for 30 min with anti-IFN-γ-PE-CY7 (Clone 4S.B3, BD Pharmingen). Samples were washed and fixed in 2% paraformaldehyde PBS. At least 200,000 lymphocyte events were collected using an LSRII flow cytometer, and the data was analyzed using FlowJo 8.8.6. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### Mamu-KIR3DL07 does not bind to Mamu-A1\*00201. (A) An alignment comparing the predicted amino acid sequences of the D0, D1 and D2 domains for six *Mamu-KIR3DL07* alleles. Positions of amino acid identity with the consensus sequence are indicated by a period. The shaded regions correspond to loops predicted to contact surfaces of the peptide-MHC class I complex. (B) Jurkat cells were electroporated with constructs expressing HA-tagged allotypes of Mamu-KIR3DL07 and stained the following day with APC-conjugated Gag~71-79~ GY9 or Nef~159-167~ YY9. The cells were then stained with a PE-conjugated antibody to the HA tag and analyzed by flow cytometry. Tetramer versus HA staining is shown after gating on the eGFP^+^ cell population. (0.85 MB TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### The D1 domain of *mmKIR3DL05x* is identical to the D1 domains encoded by *Mamu-KIR3DS* alleles. The amino acid sequences of mmKIR3DL05x, Mamu-KIR3DS02\*00402 and mmKIR3DHa are shown aligned to Mamu-KIR3DL05\*008. Positions of amino acid identity are indicated with a period and translational stop sites are indicated with an asterisk. The shaded regions correspond to loops predicted to contact the peptide-MHC class I complex. (0.28 MB TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S3 ::: {.caption} ###### Target cells expressing Mamu-A1\*00201 suppress the production of IFNγ by tetramer-positive NK cells. PBMC from two *Mamu-A1\*00201^+^* and two *Mamu-A1\*00201^--^* macaques were incubated overnight at a 5:1 effector to target cell ratio with parental 721.221 cells, or with 721.221 cells expressing individual rhesus macaque MHC class I molecules. Following stimulation, the samples were stained with Gag~71-79~ GY9 tetramer, followed by antibodies to CD3, CD8, CD16 and NKG2A. The samples were then fixed, permeabilized and stained with an IFNγ-specific monoclonal antibody. After gating on CD3-NK2GA^+^ lymphocytes, the frequency of tetramer-positive versus tetramer-negative NK cells expressing IFNγ was determined. (0.34 MB TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S1 ::: {.caption} ###### Supplemental [Table 1](#ppat-1001316-t001){ref-type="table"}. (0.04 MB DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: We are grateful to Nancy Wilson and David Watkins at the University of Wisconsin-Madison for providing the MHC class I tetramers used in this study. We also thank Jackie Gillis and Michelle Connole at the New England Primate Research Center for flow cytometry services. The authors have declared that no competing interests exist. This work was supported by Public Health Service grants AI063993, AI071306 and RR000168. DTE is an Elizabeth Glaser Scientist supported by the Elizabeth Glaser Pediatric AIDS Foundation. This project was also supported in part by intramural NIH funding to the NCI under contract number HHSN261200800001E. The content of this publication does not necessarily reflect the views or policies of the DHHS, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government. The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: ADC DTE. Performed the experiments: ADC WJN. Analyzed the data: ADC. Contributed reagents/materials/analysis tools: BNB RKR GA MA RPJ MC DHO. Wrote the paper: ADC DTE.

PubMed Central

2024-06-05T04:04:19.712874

2011-3-10

PMC3053352

Introduction {#s1} ============ To complete the early steps of infection, retroviral preintegation complexes (PICs) must access the nucleus of the infected cell and integrate the viral cDNA into host chromatin. Gammaretroviruses such as MLV require nuclear envelope breakdown during mitosis to access cellular chromosomes and complete integration [@ppat.1001313-Roe1], [@ppat.1001313-Lewis1]. In contrast, lentiviruses such as HIV can enter the nucleus in non-cycling cells, presumably by traversing the nuclear pore [@ppat.1001313-Bukrinsky1]--[@ppat.1001313-Heinzinger1]. Passage through the pore is likely a preferred route of nuclear entry for HIV-1 even in dividing cells -- several components of the nuclear pore are required for efficient infection of dividing cells, even though PICs might access the nucleus during nuclear breakdown in mitosis [@ppat.1001313-Brass1]--[@ppat.1001313-Ebina1]. Moreover, in infections initiated during interphase, integration occurs before mitosis, while integration in cells infected just prior to mitosis is delayed until the following interphase [@ppat.1001313-Katz1]. These data suggest that the steps of HIV import through the nuclear pore may be coupled to subsequent integration. In support of this hypothesis, König and colleagues found that in dividing cells depleted of some nuclear pore factors or karyopherins, HIV DNA entered the nucleus but did not integrate efficiently [@ppat.1001313-Konig1]. Thus the route of nuclear entry may influence subsequent integration, and the pore may provide the preferred route even in dividing cells. Retroviral integration is known to be modulated by several host components. Integration target site selection is guided by the genomic environment of the integration acceptor site [@ppat.1001313-Schroder1]--[@ppat.1001313-Berry1]. Lentiviruses such as HIV show a preference for integration in active transcription units, which may promote efficient expression after integration [@ppat.1001313-Schroder1], [@ppat.1001313-Verdin1]--[@ppat.1001313-Weinberger1]. Gammaretroviruses such as MLV show a preference for integration near gene 5′ ends and CpG islands [@ppat.1001313-Schroder1]--[@ppat.1001313-Wu1]. Target site preferences of HIV integration are due in part to tethering by a host chromatin binding protein, Ledgf/p75 (product of the PSIP1 gene), which binds lentiviral IN [@ppat.1001313-Cherepanov1], [@ppat.1001313-Emiliani1] and mediates IN-chromatin binding [@ppat.1001313-Maertens1], [@ppat.1001313-Llano1]. In the absence of Ledgf/p75, HIV integration is severely compromised and integration in transcription units is diminished [@ppat.1001313-Shun1]--[@ppat.1001313-Ciuffi1]. Recently, the tethering model for Ledgf/p75 function was bolstered by the finding that fusion proteins containing the IN-binding domain of Ledgf/p75 fused to alternative chromatin binding domains retargeted lentiviral integration efficiently [@ppat.1001313-Silvers1]--[@ppat.1001313-Ferris1]. Here we analyze host factors identified in genome-wide siRNA screens [@ppat.1001313-Brass1]--[@ppat.1001313-Zhou1] and find links between transport into the nucleus and subsequent integration targeting. We chose factors whose depletion, like that of Ledgf/p75, led to an infection block at nuclear entry or integration. We initially surveyed effects of knocking down expression of ten genes, then focused on two of them, TNPO3 and RANBP2, which encode components of the nuclear pore and import machinery. TNPO3 encodes Transportin-3, a karyopherin [@ppat.1001313-Lai1] that has been shown to be required for import of HIV PICs into the nucleus in cycling cell lines and macrophages [@ppat.1001313-Brass1], [@ppat.1001313-Konig1], [@ppat.1001313-Christ1]. RanBP2 (originally named Nup358), is a large cyclophilin-related nuclear pore protein involved in the Ran-GTPase cycle that orchestrates much of nuclear import and export [@ppat.1001313-Wu2], and is also required for import of HIV PICs [@ppat.1001313-Konig1]. Recently, Lee and colleagues isolated a capsid mutant (N74D) [@ppat.1001313-Lee1] that bypassed the requirement for Transportin-3 and RanBP2, but acquired a requirement for other nuclear pore factors. HIV capsid had previously been suggested to be a viral determinant of nuclear entry [@ppat.1001313-Yamashita1] and these data suggest a possible direct interaction of capsid with Transportin-3 and RanBP2. Using RNA interference, we reduced the expression of candidate genes, confirmed that HIV titer was reduced as a result, and then investigated the distribution of integration sites in the human genome using DNA bar coding and 454/Roche pyrosequencing. As controls, we studied infections and targeting by MLV. We also studied integration targeting by a derivative of HIV containing the *gag* gene (encoding the capsid structural proteins) of MLV. We found that depletion of Transportin-3 and RanBP2 resulted in marked alterations in the distribution of HIV integration sites, providing a link between nuclear entry and integration targeting. MLV integration patterns were not altered in Tranportin-3 knockdowns, and substitution of MLV Gag into HIV phenocopied the effects of the knockdowns. Several additional host gene products were also identified as candidate members of the pathway. Thus we can begin to specify a \"railroad track\" through the nuclear pore to favored sites of HIV DNA integration. Results {#s2} ======= Surveying integration site distributions after siRNA knockdown {#s2a} -------------------------------------------------------------- We initially analyzed 10 genes previously implicated as HIV cofactors at or near the integration step to determine whether they had effects on integration targeting ([Table S1](#ppat.1001313.s009){ref-type="supplementary-material"}). We selected NUP98 [@ppat.1001313-Konig1], [@ppat.1001313-Ebina1], MAP4 [@ppat.1001313-Brass1], [@ppat.1001313-Konig1], IK [@ppat.1001313-Konig1], ANAPC2 [@ppat.1001313-Konig1], [@ppat.1001313-Zhou1], PRPF38A [@ppat.1001313-Konig1], RANBP2 [@ppat.1001313-Brass1], [@ppat.1001313-Konig1], SNW1 [@ppat.1001313-Konig1], and TNPO3 [@ppat.1001313-Brass1], [@ppat.1001313-Konig1] from siRNA screens, and two other genes, WDR46 and WDHD1, the products of which bind Ledgf/p75 in yeast two-hybrid screens (unpublished data). For each gene, we tested several different siRNAs in HEK-293T cells. Reduction of mRNA levels was confirmed by quantitative RT-PCR ([Figure S1](#ppat.1001313.s001){ref-type="supplementary-material"}), and we assessed inhibition of infection by a VSVG-pseudotyped GFP reporter virus, as defined as percent of cells expressing the GFP marker 48 h after infection ([Figure S2](#ppat.1001313.s002){ref-type="supplementary-material"}), as well as toxicity of the siRNAs ([Figure S3](#ppat.1001313.s003){ref-type="supplementary-material"}). Selected knockdowns were verified by Western blot ([Figure S4](#ppat.1001313.s004){ref-type="supplementary-material"} and [Figure 1A](#ppat-1001313-g001){ref-type="fig"}). ::: {#ppat-1001313-g001 .fig} 10.1371/journal.ppat.1001313.g001 Figure 1 ::: {.caption} ###### Effects of siRNA treatments on HIV integration in gene dense regions. Cells were transfected with individual siRNAs or an siRNA pool of four siRNAs targeting the same gene as indicated, and infected 48 hr later for an additional period of 48 hr prior to integration site analysis. (A) Reduction in Transportin-3 and RanBP2 protein levels after RNAi. Protein abundance was measured at the time of infection by Western blot with β-tubulin as a loading control. For comparison, protein levels are shown in cells treated with an siRNA against firefly luciferase (GL2), a gene not found in HEK-293T cells. (B) Overview of the approach for integration site analysis. The number of genomic features of interest (blue bars), such as transcription units, is tabulated within genomic intervals (black bars) surrounding integration sites (red arrowheads) or computationally-generated matched random control sites (green arrowheads). The average number of times the genomic feature occurs within that window can be compared across datasets. (C) Histogram indicating distribution of integration sites with respect to gene density. Cells were transfected with individual siRNAs and infected as above. Sample names in legend indicate the gene targeted followed by the individual siRNA number. The number of genes in 1 Mb windows surrounding each integration site was counted as in 1B. Integration sites in each dataset were binned (along the X-axis) according to the number of genes within 1 MB interval surrounding each site. Curves were computed from histogram plot using Gaussian kernal density estimates. (D) Barplot of the average number of RefSeq genes in 1 Mb windows surrounding sites of HIV integration or computationally generated matched random controls. Mock transfected cells (no RNAi) and cells treated with the siRNA targeting luciferase GL2 (siGL2) are shown as controls. Asterisks denote significant difference from control GL2 siRNA treated cells as determined by the nonparametric Mann--Whitney test (\*P\<0.05; \*\*P\<0.01; \*\*\*P\<0.001). :::  ::: This initial scan showed robust effects on infection efficiency for the nuclear import factors Transportin-3 and RanBP2, confirming observations from earlier studies [@ppat.1001313-Brass1], [@ppat.1001313-Konig1], [@ppat.1001313-Christ1], [@ppat.1001313-Lee1]; therefore, these genes were studied in detail as described in the following sections. Results for Transportin-3 and RanBP2 have been corroborated by further studies using stable knockdowns with shRNAs in HeLa cells that achieved efficient reductions in mRNA levels (Schaller *et al.*, submitted). The remaining 8 genes were also analyzed for integration targeting using our high throughput pipeline. We return to findings for this group of genes at the end of the Results. HIV integration site selection is modified by depletion of Transportin-3 and RanBP2 {#s2b} ----------------------------------------------------------------------------------- Having confirmed that knockdown of Transportin-3 and RanBP2 reduced the efficiency of HIV infection ([Figure S2](#ppat.1001313.s002){ref-type="supplementary-material"}), we examined the effect of these factors on integration site selection using ligation-mediated PCR and 454-pyrosequencing as previously described [@ppat.1001313-Wang1]. Recovered genomic sequences were mapped to the human genome draft hg18. Association of integration sites with genomic features was then assessed (e. g. [Figure 1B](#ppat-1001313-g001){ref-type="fig"}). In the human genome, many types of features are linked--for example, gene dense regions are rich in CpG islands and DNAseI sites, high in G/C content, and rich in highly expressed genes [@ppat.1001313-Lander1], [@ppat.1001313-Venter1]. As a first step in illustrating the results, we present integration site distributions as a function of gene density. In cells depleted of Transportin-3 or RanBP2, the distribution of HIV integration sites was altered towards regions of lower gene density in comparison to control cells treated with siGL2, which targets firefly luciferase GL2, a gene not found in the HEK-293T cells ([Figure 1C](#ppat-1001313-g001){ref-type="fig"}). The trend towards integration in less gene dense regions was significant for both RANBP2 and TNPO3 knockdowns (p\<0.001, see below). There was no evidence of a bimodal distribution integration sites with respect to gene density, which would have suggested knockdown of the factors in only a portion of the cells ([Figure 1C](#ppat-1001313-g001){ref-type="fig"}). The average gene density in a one megabase window surrounding integration sites in cells depleted of either Transportin-3 or RanBP2 is plotted in [Figure 1D](#ppat-1001313-g001){ref-type="fig"}. For comparison, matched random control sites within the human genome were computationally generated and are shown in black (described in [@ppat.1001313-Mitchell1], [@ppat.1001313-Berry1] and [Text S1](#ppat.1001313.s011){ref-type="supplementary-material"}). The average gene density at integration sites in the RANBP2 and TNPO3 knockdown cells was reduced compared to cells treated with siGL2, though it remained higher than would be expected for random integration. Thus integration in gene dense regions is promoted in part by RanBP2 and Transportin-3. As a control for the fact the knockdowns diminished infection, we investigated whether infections at low MOI altered the distribution of integration sites, but MOI was not found to affect integration targeting detectably (data not shown). Analysis of integration frequency relative to a large collection of genomic features (described in [Text S1](#ppat.1001313.s011){ref-type="supplementary-material"}) showed a common set of changes in both the Transportin-3 and RanBP2 depleted cells relative to the controls ([Figure 2](#ppat-1001313-g002){ref-type="fig"} and [Figure S5](#ppat.1001313.s005){ref-type="supplementary-material"}). The reduction in integration in gene dense regions was significant for both TNPO3 and RANBP2 knockdowns when analyzed over multiple genomic intervals of different lengths. Significant differences were also seen when only expressed genes (identified by Affymetrix chip transcriptional profiling) were considered in a similar analysis (labeled "Expression Intensity" in [Figure 2](#ppat-1001313-g002){ref-type="fig"}). Genomic features that correlate with gene density such as DNase I hypersensitive sites and CpG islands were similarly enriched near control HIV integration sites but less enriched near sites from TNPO3 and RANBP2 knockdown cells. GC-rich regions, normally favored by HIV [@ppat.1001313-Schroder1], were disfavored in most window sizes in the Transportin-3 and RanBP2 knockdowns. ::: {#ppat-1001313-g002 .fig} 10.1371/journal.ppat.1001313.g002 Figure 2 ::: {.caption} ###### Effects of Transportin-3 and RanBP2 depletion on integration near multiple chromosomal features. Genes targeted by siRNA in infected cells including the control, GL2, are shown above the columns. Mock cells received no siRNA. The genomic features analyzed are shown in the rows and labeled on the left. Relationships between integration frequency and feature density are summarized using ROC curve areas [@ppat.1001313-Berry1], where increasing shades of blue indicate a negative correlation with integration frequency and increasing shades of red indicate a positive correlation with integration frequency relative to matched random control distributions. The control GL2 siRNA set was used for pairwise statistical comparisons (overlay dashes). P values summarizing the significance of the departure from the GL2 control are shown with asterisks (\*P\<0.05; \*\*P\<0.01; \*\*\*P\<0.001). Note that the asterisks and the heat map summarize different comparisons (to siGL2 and matched random controls, respectively). The base pair values in the row labels indicate the size of the genomic interval used for analysis--often the most appropriate interval is not known, so several different interval sizes are compared. A more detailed guide to the data presented in this figure can be found in [Text S1](#ppat.1001313.s011){ref-type="supplementary-material"}. An interactive version of this figure is available as [Figure S5](#ppat.1001313.s005){ref-type="supplementary-material"}. :::  ::: By contrast, gene density at integration sites was not significantly affected in Ledgf/p75 knockdowns compared to the control. The GC content and the density of CpG islands within one kb of integration sites actually increased in Ledgf/p75-depleted cells [@ppat.1001313-Shun1]--[@ppat.1001313-Ciuffi1], indicating divergent effects on integration targeting. Integration within genes, which is reproducibly diminished in Ledgf/p75-depleted cells [@ppat.1001313-Shun1]--[@ppat.1001313-Ciuffi1], was not affected by TNPO3 knockdown, and showed only a slight decrease in the RANBP2 knockdown cells. Together these data suggest that Transportin-3 and RanBP2 influence HIV integration targeting relative to a collection of features associated with gene dense regions, and do so in a manner that differs from Ledgf/p75 tethering. Effect of Transportin-3 depletion on integration site selection can be partially rescued by expression from an siRNA insensitive TNPO3 allele {#s2c} --------------------------------------------------------------------------------------------------------------------------------------------- Multiple different siRNAs directed against TNPO3 and RANBP2 mRNAs yielded similar effects on integration targeting that were not observed in control knockdowns, indicating that off-target effects were unlikely to explain the observed alterations in integration targeting. As an additional control, we analyzed complementation of the Transportin-3 depletion using a plasmid-encoded siRNA-insensitive allele generated by site-directed mutagenesis of the siRNA target sequence. The RANBP2 coding region is very large (11,711 bp), and so rescue experiments were not attempted for this factor. Co-transfection of the resistant Transportin-3 expression vector with the corresponding siRNA resulted in overexpression of Transportin-3 and restored HIV infection, increasing reporter virus GFP expression above control levels ([Figure 3a and b](#ppat-1001313-g003){ref-type="fig"}). ::: {#ppat-1001313-g003 .fig} 10.1371/journal.ppat.1001313.g003 Figure 3 ::: {.caption} ###### Transfection of a Transportin-3 allele insensitive to TNPO3 si4 restores protein expression, HIV infectivity, and partially restores wild-type HIV integration site distributions. \(A) Western blot showing Transportin-3 levels in cells treated with TNPO3 si4 in the presence or absence of the Transportin-3 rescue plasmid. Cells were cotransfected with siRNA and either empty vector plasmid or rescue plasmid encoding siRNA-resistant alleles of Transportin-3 expressed from the CMV promoter and harvested at 48 hr post-transfection for analysis. Transportin-3 is reduced after co-transfection with siRNA and empty vector, and overexpressed after co-transfection with siRNA and rescue plasmid. Endogenous levels of Transportin-3 are shown in cells transfected with the control siRNA targeting GL2 and an empty vector. (B) HIV infection in cells treated with TNPO3 si4 in the presence or absence of the Transportin-3 rescue plasmid. Cells were co-transfected as above. 48 hr after transfection cells were infected with a VSVG-pseudotyped HIV-1 vector carrying a GFP reporter. At 48 hpi cells were harvested and the percent of cells expressing GFP was determined by flow cytometry. The Y-axis shows relative infection compared to infection in the control (GL2 siRNA + empty vector-transfected) cells. (C) Average gene density in 1 Mb windows surrounding HIV integration sites in cells depleted or rescued for Transportin-3 expression. Asterisks denote significant differences as determined by the Mann--Whitney test (\*P\<0.05; \*\*P\<0.01; \*\*\*P\<0.001). :::  ::: We observed an increase in gene density near integration sites in knockdown cells co-transfected with the siRNA-insensitive TNPO3 allele compared to vector-only controls ([Figure 3c](#ppat-1001313-g003){ref-type="fig"} and [Figure S6](#ppat.1001313.s006){ref-type="supplementary-material"}). The average number of genes within 1 Mb of HIV integration sites increased from 11 (in the presence of TNPO3 si4 and an empty vector) to 14 when Transportin-3 expression was rescued (p\<0.01, [Figure 3c](#ppat-1001313-g003){ref-type="fig"}). The effect of knockdown in the presence and absence of rescue on additional genomic features is described in [Text S2](#ppat.1001313.s012){ref-type="supplementary-material"}. It is unclear why restoring Transportin-3 protein levels did not fully rescue the integration defect, but this result may be due to the abnormally high levels of Transportin-3 expressed from the siRNA-resistant construct. Nevertheless, these data support the idea that off-target effects of the TNPO3 siRNA do not account for the phenotypes observed. Transportin-3 depletion has no detectable effect on gene density surrounding MLV integration sites {#s2d} -------------------------------------------------------------------------------------------------- As a control, we tested whether MLV integration, which requires cell division for infection and is not dependent on Transportin-3 [@ppat.1001313-Konig1], [@ppat.1001313-Krishnan1], showed altered integration targeting in the Transportin-3-depleted cells. We found that treatment with siRNA targeting TNPO3 mRNA, either in the presence or absence of the rescue plasmid, did not affect MLV infection efficiency ([Figure 4a](#ppat-1001313-g004){ref-type="fig"}). We sequenced MLV integration sites from knockdown and control cells ([Table S1](#ppat.1001313.s009){ref-type="supplementary-material"}), and found no significant changes in MLV integration frequency in gene dense regions ([Figure 4b](#ppat-1001313-g004){ref-type="fig"}), within transcription units, or with respect to GC content (data not shown). These data indicate that Transportin-3 depletion does not affect MLV integration targeting as it does for HIV. ::: {#ppat-1001313-g004 .fig} 10.1371/journal.ppat.1001313.g004 Figure 4 ::: {.caption} ###### Depletion of Transportin-3 does not alter MLV integration targeting. \(A) Infection levels of MLV in cells co-transfected with the control siRNA to GL2 plus an empty vector, TNPO3-si4 plus an empty vector, or TNPO3 si4 plus a vector encoding the siRNA-resistant Transportin-3 allele. (B) Average gene density in 1 Mb windows surrounding MLV integration sites in cells depleted or rescued for Transportin-3 expression. No sets showed significant differences from GL2-treated cells as determined by the Mann--Whitney test. :::  ::: Other nuclear factors may participate in directing integration to gene dense regions {#s2e} ------------------------------------------------------------------------------------ Integration site data sets were also acquired for cells treated with siRNAs for NUP98, MAP4, IK, ANAPC2, PRPF38A, SNW1, WDR46 and WDHD1 ([Table S1](#ppat.1001313.s009){ref-type="supplementary-material"}). For many of these, considerable toxicity was detected ([Figure S3](#ppat.1001313.s003){ref-type="supplementary-material"}). Thus interpretation of integration targeting results for these factors is more tentative than for Transportin-3 and RanBP2. Data sets were analyzed for their association with gene density as for Transportin-3 and RanBP2 ([Figure 5](#ppat-1001313-g005){ref-type="fig"}). Knockdown of several of the factors (ANAPC2, SNW1, PRPF38, WDH1, and IK) led to decreased integration in gene dense regions. MAP4 depletion was also seen to modestly decrease integration preference for gene dense regions in some experiments. For two of these genes, SNW1 and ANAPC2, we confirmed that although MLV infection is diminished in the knockdowns as previously noted [@ppat.1001313-Konig1], the gene density at MLV integration sites is unchanged ([Figure S7](#ppat.1001313.s007){ref-type="supplementary-material"}), suggesting that, like Transportin-3, the factors encoded by these genes are potentially involved in targeting pathways specific for HIV. By contrast, gene density at integration sites in cells stably depleted of Ledgf was not significantly decreased compared to the siGL2 control. ::: {#ppat-1001313-g005 .fig} 10.1371/journal.ppat.1001313.g005 Figure 5 ::: {.caption} ###### Depletion of additional host factors and their effects on HIV integration in gene dense regions. Integration sites were isolated from cells treated with siRNAs targeting the indicated genes. The average numbers of RefSeq genes in 1 Mb windows surrounding integration sites are shown. Data for a given gene knockdown is the average over multiple siRNA knockdowns using different siRNAs and pools of siRNAs targeted to the same gene, except for Mock, control siRNA (GL2), and LEDGF knockdown conditions for which single treatments were used. Asterisks denote significant difference from control GL2 siRNA-treated cells as determined by the Mann--Whitney test (\*P\<0.05; \*\*P\<0.01; \*\*\*P\<0.001). Error bars represent standard error for biological replicates. For LEDGF only one dataset was available. :::  ::: For those knockdowns where we could sequence at least 200 integration sites, the global integration site patterns were investigated by assessing integration frequency relative to many genomic features for each knockdown, and the patterns were clustered using a conditional logit model to conduct pairwise comparisons of the datasets (details are in [Text S3](#ppat.1001313.s013){ref-type="supplementary-material"}). The dendrogram in [Figure 6](#ppat-1001313-g006){ref-type="fig"} shows that the controls clustered in a group separate from Transportin-3 and RanBP2 knockdowns. Data sets for several additional gene knockdowns clustered in the TNPO3/RANBP2 group, including IK, ANAPC2, SNW1, WDHD1 and PRPF38A. For MAP4 and WDR46 different siRNAs fell in different groups, and so these have an indeterminate effect. Thus the IK, ANAPC2, SNW1, WDHD1 and PRPF38A genes encode candidates for additional factors acting in the same pathway with Transportin-3 and RanBP2. The Ledgf/p75 knockdown was an outlier in the control cluster. This is consistent with Ledgf/p75 knockdown leading to effects not seen in depletion of RanBP2, or Transportin-3. For this analysis we investigated both low MOI (30--60% infected wild type cells) and high MOI (90--100% infected cells) infections. In most cases the MOI made no difference on the overall position of a knockdown within the tree, suggesting that the roles of the factors are not saturable under the conditions tested. ::: {#ppat-1001313-g006 .fig} 10.1371/journal.ppat.1001313.g006 Figure 6 ::: {.caption} ###### Dendrogram showing clustering of integration site data sets from knockdowns of Transportin-3, RanBP2, and several additional factors. Only sets containing at least 200 integration sites were used for the analysis. A conditional logit model was used to cluster integration sites data sets based on annotation of in or out of annotated transcription units, gene density, expression density, CpG islands, G/C content, nearby oncogenes, and local sequence features ([Text S3](#ppat.1001313.s013){ref-type="supplementary-material"}). Sets were clustered based on their overall similarity in a pairwise analysis. The \"Control\" cluster is so named because it contains the Mock and siGL2 control data sets. Branch labels indicate the siRNA used for the analysis, and indicates the name of the targeted gene (e.g. TNPO3 si4). Infections were performed using enough HIV vector stock to infect 30--60% of untreated cells except where marked as "highMOI" where 90--100% of untreated cells were infected. :::  ::: HIV *gag* is a determinant of integration targeting to gene dense regions {#s2f} ------------------------------------------------------------------------- We previously studied integration targeting in HeLa cells using HIV chimeras containing MLV *gag*, MLV *IN*, or both, in place of their HIV counterparts [@ppat.1001313-Lewinski2]. We found that MLV IN was a dominant determinant of MLV-like integration, resulting in integration near transcription start sites by HIV derivatives containing MLV IN. Similar chimeric viruses have been used to show that HIV capsid is a dominant viral determinant of HIV nuclear entry in non-dividing cells [@ppat.1001313-Yamashita2]. Recently, Lee and colleagues [@ppat.1001313-Lee1] suggested that the HIV CA protein might determine the interactions between HIV PICs and nuclear pore components. These findings led us to reinvestigate integration targeting by the HIV chimera containing MLV *gag* in place of HIV *gag* (HIVmGag; [Fig. 7A](#ppat-1001313-g007){ref-type="fig"}) [@ppat.1001313-Lewinski2]. We found that HIVmGag showed a shift in distribution of integration sites towards less gene dense regions compared to the unmodified control ([Figure 7B](#ppat-1001313-g007){ref-type="fig"}). The average number of genes within 1 MB of HIVmGag integration sites was 11 as compared to 20 for the unmodified HIV control (A Chi square test over ranked comparisons of gene density values between the two sets attains a p value of \<2.22--16). A comparison over many genomic features ([Figure 7C](#ppat-1001313-g007){ref-type="fig"} and [Figure S8](#ppat.1001313.s008){ref-type="supplementary-material"}) showed a pattern of HIVmGag integration similar to that seen for HIV in Transportin-3 and RanBP2 depleted cells (compare [Figure 2](#ppat-1001313-g002){ref-type="fig"}), including reduced density of genes, CpG islands, DNase I hypersensitive sites and reduced GC content surrounding integration sites. Thus substitution of HIV *gag* with MLV *gag* phenocopied the TNPO3 and RANBP2 knockdowns. ::: {#ppat-1001313-g007 .fig} 10.1371/journal.ppat.1001313.g007 Figure 7 ::: {.caption} ###### A chimeric derivative of HIV containing MLV *gag* (HIVmGag) shows reduced integration frequency in gene dense regions. \(A) Genetic map of HIV proviruses containing wild type *gag* (HIVPuro) or a chimera encoding MLV Gag (MA, p12, and CA) in place of HIV MA and CA (HIVmGag). Both viruses have inactivated *vpr* and *env* and a puromycin selectable marker in place of *nef*. (B) Histogram indicating distribution of HIVPuro and HIVmGag integration sites with respect to gene density measured in 1 Mb intervals surrounding integration events. Data is plotted as in [Fig. 1C](#ppat-1001313-g001){ref-type="fig"} and curves are computed using Gaussian kernel density estimates. (C) Genomic heatmap of HIVPuro and HIVmGag datasets. Significant differences are shown by asterisks (\*p\<0.05; \*\*p\<0.01; \*\*\*p\<0.001). Annotations at the left of the heat map are as in [Figure 2](#ppat-1001313-g002){ref-type="fig"} and described in [Text S1](#ppat.1001313.s011){ref-type="supplementary-material"}. An interactive version of this figure is available as [Figure S8](#ppat.1001313.s008){ref-type="supplementary-material"}. :::  ::: Knockdowns of RANBP2 or TNPO3 do not cause HIV to favor integration near transcription start sites {#s2g} -------------------------------------------------------------------------------------------------- A model to explain the altered integration site patterns of HIV in TNPO3 or RANBP2 knockdowns is that in the absence of these pore proteins the HIV PIC accesses chromatin during nuclear breakdown during mitosis. MLV employs such a mechanism for nuclear entry, so we wondered whether the HIV integration site distributions in the knockdowns might resemble the normal pattern for MLV. We asked whether HIV integration in cells knocked down for TNPO3 and RANBP2 shows the most characteristic feature of MLV integration, favored integration near transcription start sites ([Figure 8](#ppat-1001313-g008){ref-type="fig"}). We found that HIV in the knockdowns disfavors transcription start sites, paralleling HIV integration in unmodified cells. MLV showed strongly favored integration in transcription start sites in the 293T cells studied, and in 293T cells knocked down for TNPO3. We conclude that obstructing the normal HIV pathway of integration by knocking down RANBP2 or TNPO3 does not result in an MLV-like integration targeting pattern. This is consistent with the observation that IN is the dominant determinant of MLV like integration patterns at transcription start sites for chimeric viruses where HIV IN is replaced with MLV IN [@ppat.1001313-Lewinski2], [@ppat.1001313-Yamashita2]. ::: {#ppat-1001313-g008 .fig} 10.1371/journal.ppat.1001313.g008 Figure 8 ::: {.caption} ###### HIV and MLV integration patterns at transcription start sites are unaffected by knockdown of TNPO3 or RANBP2. The percent of integration sites within the indicated genomic distances (kb) from the transcription start site (RefSeq genes) is plotted for each dataset. Sample names indicate the VSVG-pseudotyped viral vector used (HIV or MLV) followed by the cell treatment (either control siGL2 or gene-specific siRNA used). :::  ::: Discussion {#s3} ========== Here we report that depletion of Transportin-3 and RanBP2 by RNAi affects the downstream choice of targets for HIV DNA integration, providing evidence for coupling of the nuclear translocation and integration steps. As others have noted, Transportin-3 has little or no effect on infection efficiency of MLV [@ppat.1001313-Brass1], [@ppat.1001313-Konig1], [@ppat.1001313-Christ1], which is not thought to traverse the nuclear pore, and we report that Transportin-3 did not affect integration targeting by MLV. Replacing HIV *gag* with MLV *gag* phenocopied the effects of the Transportin-3 and RanBP2 knockdowns on HIV integration targeting. These findings support a model in which HIV Gag proteins interact with Transportin-3 and RanBP2 to mediate HIV integration targeting to chromosomal regions rich in genes and associated features. We found that depletion of several additional factors previously shown to be required for efficient integration also resulted in HIV integration targeting patterns similar to those seen in Transportin-3 and RanBP2 depleted cells. These factors include a component of the anaphase promoting complex (ANAPC2) splicing factors (SNW1 and PRPF38), a WD-repeat protein (WDHD1), and nuclear DNA binding proteins (IK and SNW1). The analysis of some of these was complicated by cell toxicity, and in some cases conflicting results were obtained with different siRNAs, so effects of these factors are less well supported than those of Transportin-3 and RanBP2. It is possible that each of these factors acts in a common pathway with Transportin-3 and RanBP2 to direct integration to regions dense in genes and associated features, though depletion of some of these factors could also alter the synthesis or function of other factors acting more directly. Our studies support the hypothesis that nuclear import of HIV is linked to integration, and suggest that normal interactions with the nuclear pore help to determine integration target site distributions ([Figure 9](#ppat-1001313-g009){ref-type="fig"}). We favor a two-step model, in which passage through the pore first places the PIC in regions of high gene density, and then Ledgf/p75 tethers the PIC for integration to provide the final distribution in active transcription units. Several studies suggest that chromosomes and genes are nonrandomly distributed in the nucleus, though the organization is not fully clarified [@ppat.1001313-Xing1]--[@ppat.1001313-Osborne1]. Although the nuclear periphery is thought to be rich in heterochromatic chromosomal regions that promote gene silencing, studies in yeast and Drosophila suggested that genes can relocate to the nuclear pore upon transcriptional induction [@ppat.1001313-Casolari1]--[@ppat.1001313-Mendjan1]. Thus passage through the pore may deliver HIV to locally concentrated active gene-dense chromatin. Alternatively, interaction with Transportin-3 and RanBP2 at the pore might engage a nuclear transport system leading to gene-dense chromatin. ::: {#ppat-1001313-g009 .fig} 10.1371/journal.ppat.1001313.g009 Figure 9 ::: {.caption} ###### Model for coupling of nuclear import and integration targeting. Interaction with Transportin-3 and RanBP2 shuttles the PIC through the nuclear pore and toward gene dense regions favored for HIV integration. Interactions with additional factors in the nucleus (ANAPC2, WDH1, IK, PRPF38A, and SNW1) may also play a role in site selection upstream of the known integration cofactor Ledgf/p75, which targets integration to active transcription units. RNA Pol indicates RNA polymerase II, which is known to be required for transcriptional activity, and which promotes integration [@ppat.1001313-Schroder1]--[@ppat.1001313-Wu1], [@ppat.1001313-Berry1]. Nucleosomes are shown because target DNA is known to be wrapped in nucleosomes during the integration step [@ppat.1001313-Wang1], [@ppat.1001313-Pryciak1]--[@ppat.1001313-Pruss2]. PIC, preintegration complex; FG, phenylalanine-glycine repeat sequences of nuclear pore proteins. :::  ::: Our data is consistent with the idea that correct engagement of the Transportin-3/RanBP2-dependent targeting pathway leads to efficient integration in chromosomal regions rich in genes and associated features. Failure to engage this pathway results in targeting to less gene dense regions. Two possible scenarios can be imagined for nuclear entry and integration targeting in cells depleted for pore factors Transporting-3 and RanBP2. The first model is that in the absence of Transportin-3 or RanBP2, nuclear access of HIV is restricted to times of nuclear envelope breakdown during cell division. The shift in integration away from gene-dense regions in the TNPO3 and RANBP2 knockdowns may thus reflect changes in chromatin availability during mitosis or shortly afterwards. Consistent with this idea, the HIVmGag virus requires nuclear envelope breakdown during mitosis for infection [@ppat.1001313-Yamashita2], and it phenocopied HIV integration in the knockdown cells, showing reduced integration frequency in gene dense regions. An extreme version of this model would hold that HIV integration targeting in TNPO3 and RANBP2 knockdowns might mimic MLV targeting because in both cases the virus accesses chromatin during nuclear breakdown. However, MLV strongly favors integration near transcription start sites, and this is not seen for HIV in knockdown cells ([Figure 8](#ppat-1001313-g008){ref-type="fig"}). Similarly, if passage through the nuclear pore delivers the HIV PIC to transcription units and gene dense regions, growth arrest of cells might increase favoring of these features, since all integrants must enter through the pore in arrested cells. Integration site distributions have been investigated in growth arrested IMR90 lung fibroblasts and macrophages [@ppat.1001313-Ciuffi1], [@ppat.1001313-Barr1]. In IMR90 cells, arrest did result in more integration in transcription units and gene dense regions, but in macrophages the favoring is in fact weaker than that observed in many other cell types [@ppat.1001313-Marshall1]. Thus it is possible that passage through the nuclear pore results in favored integration in gene dense regions, but additional assumptions are needed to explain the data from macrophages. The second model (not exclusive of the first) holds that in cells depleted of TNPO3 and RanBP2, HIV integration complexes may pass through the pore but on a different pathway, interacting with different pore proteins. The idea that alternative pathways through the pore exist is supported by findings of Lee and colleagues, who found that the N74D substitution in HIV CA disrupted normal interactions with Transportin-3 and RanBP2 but created dependence on other pore proteins [@ppat.1001313-Lee1]. From our data, it is not possible to determine whether in cells depleted of Transportin-3 and RanBP2 HIV integration complexes pass through the pore on alternate pathways, or whether nuclear access during mitosis fully explains the data. Thus it will be important to analyze targeting when integration complexes pass through the pore on alternative pathways, as in the presence of the N74D CA substitution (Schaller et al., submitted). Materials and Methods {#s4} ===================== Cell culture and viral infections {#s4a} --------------------------------- HEK 293T cells were grown in D10 media (DMEM supplemented with 10% FBS and 50 ug/µL Gentamicin). For gene knockdowns, cells were grown to confluency, trypsinized and reverse transfected (100,000 cells/well in 12 well plates, 50,000/well in 24 well plates, and 8,000/well in 96 well plates) using RNAiMax (Invitrogen, Carlsbad CA) with 25 pmol/mL siRNA. The siRNAs were purchased from Qiagen (Qiagen, Valencia, CA) and are listed in [Table S2](#ppat.1001313.s010){ref-type="supplementary-material"}. Toxicity of siRNAs was measured 48 hr after transfection both visually and by the CellTiter-Glo Luminescent Cell Viability Assay (Promega, Madison WI; see [Figure S3](#ppat.1001313.s003){ref-type="supplementary-material"} for details). Transfection media was replaced after 48 hr by 500 µL of D10 plus 5 ug DEAE dextran and virus in 12 well plates. Two viral inoculums were used (0.06 µL or 1 µL concentrated virus stock corresponding to 1.32 ng or 22 ng p24 per well, values determined by titration to result in infection of 30--60% or 80--100% of cells, respectively). Virus-containing media was replaced after 10--12 hours with 1 mL D10 and incubated an additional 38 hours before harvest. Infections of LEDGF stable knockdown cell lines were performed essentially as described [@ppat.1001313-Ciuffi1]. VSV-G pseudotyped HIV vector particles were produced in HEK 293T cells by Lipofectamine (Invitrogen, Carlsbad CA) transfection of p156RRLsin-PPTCMVGFPWPRE [@ppat.1001313-Follenzi1], the packaging construct pCMVdeltaR9 [@ppat.1001313-Naldini1], and the vesicular stomatitis virus G-producing plasmid pMD.G. VSV-G pseudotyped MLV particles were produced in a similar manner but using the MLV vector segment (pMX-eGFP) and packaging construct pCGP (pCGP, kindly provided by Paul Bates). Percent infection was measured using GFP fluorescence, which is not strongly affected by integration site placement in the HIV-based vectors with strong artificial promoters used here [@ppat.1001313-Gijsbers1]. HIV infection and targeting rescue experiments were performed as described for siRNA knockdowns but with the co-transfection of siRNA-resistant or empty expression vectors (333 ng plasmid/mL). The siRNA-resistant TNPO3 allele was constructed by introducing six conservative mutations in the third position of each codon and an N-terminal 3xFLAG-tag into the TNPO3 cDNA amplified HEK-293T cells. This product was then cloned into the mammalian expression vector pLNCX (kind gift of Paul Bates), engineered to contain a WPRE. Gene expression by RNA and protein levels {#s4b} ----------------------------------------- Q-PCR (see [Figure S1](#ppat.1001313.s001){ref-type="supplementary-material"} for details) and immunoblotting were used to monitor the extent of siRNA knockdowns. Protein levels were measured by immunoblotting using antibodies against Transportin-3 (ab54353, Abcam Inc., Cambridge, MA) and RanBP2 (ab2938, Abcam Inc., Cambridge, MA). HRP conjugated secondary antibodies (p0260, DAKO A/S, Denmark, and ab6721-1, Abcam, Cambridge, MA) were used for detection with SuperSignal West Pico Chemiluminescent Substrate (Thermo Scientific, Pierce Protein Research Products, Rockford, IL). Beta-tubulin was used as a loading control, detected by the HRP conjugated antibody (ab21058, Abcam, Cambridge, MA). Integration site analysis {#s4c} ------------------------- For integration site recovery, purified genomic DNA was digested overnight with MseI, ligated at 16°C to PCR adapters, and digested a second time with SacI. Nested PCR was then performed using primers and conditions described previously [@ppat.1001313-Wang1], [@ppat.1001313-Wang2]. Amplification products between 200--600 bp were then gel-excised, purified, and sequenced on a Genome Sequencer FLX Titanium Series (Roche 454 Sequencing) at either the University of Pennsylvania or the University of Florida. Only sequences that began within three base pairs of the LTR end and showed unique best alignments to the human genome by BLAT (hg18, version 36.1, \>98% match score) were considered true integration sites. Identical integration sites identified in two or more separately amplified samples were considered to be PCR contamination and were omitted. Comparisons to genomic features were carried out as described previously [@ppat.1001313-Berry1], [@ppat.1001313-Brady1] using a combination of conditional logit regression and Bayesian model averaging. Details of statistical methods are available in [@ppat.1001313-Mitchell1], [@ppat.1001313-Berry1], [@ppat.1001313-Brady1], [@ppat.1001313-Brady2]. Methods used for statistical analysis of ROC areas ([Figures 3](#ppat-1001313-g003){ref-type="fig"} and [8](#ppat-1001313-g008){ref-type="fig"}) are summarized in [@ppat.1001313-Brady2]. Gene expression analyses utilized data from 293T cells [@ppat.1001313-Ciuffi1] with expression measured using the Affymetrix HU133 plus 2.0 gene chip array. All integration site sequences will be deposited in publicly accessible databases (NCBI) upon acceptance of this manuscript for publication. Entrez Gene ID numbers for genes mentioned in the text {#s4d} ------------------------------------------------------ NUP98: 4928, MAP4: 4134, IK: 3550, ANAPC2: 29882, PRPF38A: 84950, RANBP2/NUP358: 5903, SNW1: 22938, TNPO3: 23534, WDR46: 9277, WDHD1:11169, PSIP1/LEDGF/p75: 493969. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### mRNA levels under normal and gene knockdown conditions. 293T cells were reverse transfected as described in [Materials and Methods](#s4){ref-type="sec"} (8,000/well in 96 well plates using RNAiMax (Invitrogen, Carlsbad CA) with 25 pmol/mL siRNA, then incubated 48 hr at 37°C before harvest. RNA was purified from cells using either the RNeasy Mini Kit from Qiagen (Carlsbad, CA) or the RNAspin Mini Kit (GE Healthcare, Buckinghamshire UK) per manufacturer\'s instructions. RT-PCR was carried out using the High Capacity RNA to cDNA Kit (Applied Biosystems, Foster City CA) and relative RNA levels were measured by the ddCt method using Taqman Gene Expression Assays (Applied Biosystems, Foster City CA) with GUSB as the internal reference. Assays IDs were Hs00193785\_m1, Hs00600887\_m1, Hs00173172\_m1, Hs00273527\_m1, Hs00159048\_m1, Hs00610583\_m1, Hs01108576\_m1, Hs00203499\_m1, Hs00273351\_m1, Hs00180522\_m1 for genes measured for knockdown and product number 4333767F for the GUSB endogeneous control assay. All values were normalized the control siRNA, GL2. Data presented is representative of at least three replicate experiments. (1.58 MB TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### HIV infection levels under normal and gene knockdown conditions. 48 hr following siRNA transfection, media was replaced with 500 µL of D10 (12 well plates) plus 5 ug DEAE dextran and virus as described in [Materials and Methods](#s4){ref-type="sec"} (0.06 µL concentrated virus stock corresponding to 1.32 ng p24 per well, innoculum determined by titration to result in infection of 30--60% of cells). Virus-containing media was replaced after 10--12 hours with 1 mL D10 and incubated an additional 38 hours before harvest. Infection level was measured by flow cytometry as the percentage of GFP positive cells. All values normalized to Mock controls. (0.77 MB TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S3 ::: {.caption} ###### Cell viability after siRNA transfection. Toxicity of siRNAs was measured 48 hr after transfection both visually and by the CellTiter-Glo Luminescent Cell Viability Assay (Promega, Madison WI) following the manufacturer\'s instructions. Cells were reverse transfected in 96 well plates with the indicated siRNAs at 25 pmol/ml final concentration and incubated at 37°C. All values normalized to GL2 controls. Data shown is representative of at least two independent experiments. (1.49 MB TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S4 ::: {.caption} ###### SNW1 protein levels under normal and gene knockdown conditions. Cells were reverse transfected with SNW1 si5 or with GL2 as described, incubated 48 hr, harvested, and lysed for protein analysis. Blotting was done using rabbit polyclonal antibody from Santa Cruz Biotechnology (Santa Cruz, CA; product SC-30139 Lot B1506). Following gel transfer, PVDF membranes were incubated 2.5 hr at RT (antibodies diluted 1∶2000 in PBST, 5% milk) followed by incubation for 1 hr at RT with secondary antibody was Abcam HRP conjugated Goat anti Rabbit (goat polyclonal to Rabbit IgG; ab6721-1 lot 142201, diluted 1∶2000 in PBST, 5% milk). Knockdown of protein levels for ANAPC2 could not be confirmed by western blot (Abcam, product ab18295). (1.12 MB TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S5 ::: {.caption} ###### Effects of Transportin-3 and RanBP2 depletion on integration near multiple chromosomal features: interactive heat map. Data was analyzed and is displayed as described in [Figure 2](#ppat-1001313-g002){ref-type="fig"} and [Text S1](#ppat.1001313.s011){ref-type="supplementary-material"}. To view, download and open zip file, and follow instructions in the included ReadMe.txt document. (0.21 MB ZIP) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S6 ::: {.caption} ###### Partial rescue of HIV integration site distributions by Transportin-3 allele insensitive to TNPO3 si4. Cells were cotransfected with siRNA and either empty vector plasmid or rescue plasmid encoding siRNA-resistant alleles of Transportin-3, infected with a VSVG-pseudotyped HIV-1 vector, and harvested for integration site analysis as described. Histogram shown indicates distribution of integration sites with respect to gene density. Integration sites in each dataset were binned (along the X-axis) according to the number of genes within 1 MB interval surrounding each site (counted as shown in [Figure 1B](#ppat-1001313-g001){ref-type="fig"}). Curves were computed from histogram plot using Gaussian kernal density estimates. (1.15 MB TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S7 ::: {.caption} ###### MLV infection and integration site distributions after siRNA treatment targeting SNW1 and ANAPC2. MLV infections were carried out using VSV-G pseudotyped, single round viral vectors in the same manner described for HIV infections (see [Materials and Methods](#s4){ref-type="sec"} and Supplementary [Figure 2](#ppat-1001313-g002){ref-type="fig"}). Infection level was measured by flow cytometry as the percentage of GFP positive cells. All values normalized to GL2 controls. (1.54 MB TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S8 ::: {.caption} ###### Effects of MLV-Gag swap on integration near multiple chromosomal features: interactive heat map. Data was analyzed and is displayed as described in [Figure 7c](#ppat-1001313-g007){ref-type="fig"} and [Text S1](#ppat.1001313.s011){ref-type="supplementary-material"}. To view, download and open zip file, and follow instructions in the included ReadMe.txt document. (0.29 MB ZIP) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S1 ::: {.caption} ###### Integration site data sets used in this study. (0.08 MB PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S2 ::: {.caption} ###### DNA and RNA oligonucleotides used in this study. (0.08 MB PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Text S1 ::: {.caption} ###### Guide to Interpreting Genomic Heat Maps Summarizing Integration Site Distributions. (0.12 MB PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Text S2 ::: {.caption} ###### Distributions of HIV integration sites after TNPO3 knockdown and rescue with siRNA insensitive allele. (0.58 MB PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Text S3 ::: {.caption} ###### Integration site preference under gene silencing. (1.33 MB PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: We are grateful to members of the Bushman laboratory for help and suggestions. The authors have declared that no competing interests exist. This work was supported by NIH grants AI52845 and AI082020, the University of Pennsylvania Center for AIDS Research, and the Penn Genome Frontiers Institute with a grant with the Pennsylvania Department of Health. The Department of Health specifically disclaims responsibility for any analyses, interpretations, or conclusions. GJT is supported by Wellcome Trust Senior Fellowship no. WT076608, the UCL/UCLH National Institute of Health Research Comprehensive Biomedical Research Centre and the Medical Research Council. KEO was supported by NIH training grant T32 AI007324. TLB is a Special Fellow of the Leukemia and Lymphoma Society of America. KR was supported by NIH training grant T32 AI-07324-17. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: **¤:** Current address: Fred Hutchinson Cancer Research Center, Seattle, Washington, United States of America [^2]: Conceived and designed the experiments: KEO TLB KR FDB. Performed the experiments: KEO TLB KR AH SLR TS. Analyzed the data: KEO TLB KR SLR TS LCJ GJT JATY SKC RK NM CCB FDB. Contributed reagents/materials/analysis tools: TS LCJ GJT JATY SKC RK NM CCB. Wrote the paper: KEO TLB KR FDB.

PubMed Central

2024-06-05T04:04:19.717585

2011-3-10

PMC3053353

Introduction {#s1} ============ Blood coagulation is a very complex but well-synchronized biochemical process by which blood forms a clot and the damaged blood vessel is sealed by a platelet-rich fibrin plug leading to hemostasis. Any damage to the vascular beds due to laceration of tissues exposes tissue factor from the endothelium to the circulating blood, which initiates coagulation cascades. Once coagulation is initiated, the process leads to the generation of thrombin, which converts fibrinogen to fibrin, the building block of a hemostatic plug [@ppat.1001312-Stark1]--[@ppat.1001312-Neyman1]. In contrast, fibrinolysis is an enzymatic process wherein a fibrin clot is dissolved. Normally, in the body, both coagulation and fibrinolytic processes are precisely regulated by the measured participation of zymogens, activators, inhibitors, cofactors and receptors. Plasmin is the major fibrinolytic enzyme and is derived from the limited proteolytic cleavage of plasminogen, the circulating plasma zymogen, by its physiological activators such as tissue-type plasminogen activator (t-PA) and urokinase-type plasminogen activator (u-PA) [@ppat.1001312-CesarmanMaus1]--[@ppat.1001312-Booyse1]. Homeostasis in blood fluidity is not only vital for humans but also for hematophagous animals, which have to counteract their hosts\' hemostatic mechanisms and/or facilitate the fibrinolytic process to keep the blood in a fluid state during acquisition and digestion of blood-meals. Therefore, it is believed that blood sucking-animals require an extensive spectrum of anticoagulant and/or fibrinolytic mechanisms to maximize their feeding as part of their diverse survival strategies. Thus, hematophagous animals are thought to possess anticoagulant and/or fibrinolytic proteins that have been acquired during their evolution [@ppat.1001312-Stark1], [@ppat.1001312-Gardell1], [@ppat.1001312-Yoshida1]. The ixodid ticks (Arthropoda: Ixodidae), popularly known as hard ticks, are notorious ectoparasites and live entirely on nutrients derived from host\'s blood protein, hemoglobin. Blood-feeding is an essential biologic phenomenon for the survival of hard ticks [@ppat.1001312-Hajdusek1]--[@ppat.1001312-Akov1]. These ectoparasites firmly attach to hosts using their deeply penetrating mouthparts on hosts\' skin and cause dermatitis and severe anemia. In addition to the direct severe adverse effects on health and productivity, ticks serve as a unique vector of various deadly diseases, such as lyme disease, tick-borne encephalitis, Rocky Mountain spotted fever, babesiosis, theileriosis and anaplasmosis during hematophagy. Ticks are only second to mosquitoes as vectors of diseases of humans and animals [@ppat.1001312-Murray1]--[@ppat.1001312-Fujisaki1]. Unlike rapidly feeding hematophagous insects that suck blood directly from the blood vessels within a couple of seconds, the hard ticks feed on blood-meals for a long time (5−10 days or more), making a large blood pool beneath the host\'s skin. Although size of the blood pool produced by different species of ticks varies but it is an essential feature in the feeding mechanism of ticks [@ppat.1001312-Sonenshine1], [@ppat.1001312-Kemp1]. Blood pools contain copious unclotted blood and exudates, especially in the rapid phase of feeding [@ppat.1001312-Stark1], [@ppat.1001312-Kemp1]. It is of current interest to look at the molecular scenario inside the blood pool that maintains a large volume of unclotted blood underneath the skin. Prior studies suggest that ixodid ticks, the smart pharmacologists, produce a vast array of pharmacologically active bio-molecules that are injected into the feeding lesions during persistent blood-feeding processes, and play crucial modulatory roles in their feeding success, especially to keep the blood in a fluid state in the blood pool and within the gut as well [@ppat.1001312-Waxman1]--[@ppat.1001312-MaritzOlivier1]. However, the precise molecular mechanism(s) that prevents blood coagulation and initiates fibrinolysis in the blood pool to facilitate successful acquisition of blood-meals is still unclear. Previously, we have shown that longistatin, an ixodid tick salivary gland-derived bioactive molecule, is functionally linked to the blood-feeding processes. The molecule was shown to be secreted into the blood pool created by ticks while they attach on a robust mammalian host and suck blood to engorge [@ppat.1001312-Anisuzzaman1]. Here, we demonstrate that longistatin-specific gene-knockdown ixodid ticks such as *Haemaphysalis longicornis* completely lost their unique ability to create a blood pool and consequently were unable to feed and engorge on blood-meals as verified by RNA interference (RNAi) tool. Furthermore, we describe that longistatin binds with fibrin and specifically catalyzes the activation of plasminogen into plasmin and hydrolyzes fibrinogen *in vitro*; the latter is known to be a major component of cross-linked fibrin polymer. To the best of our knowledge, longistatin is the first plasminogen activator, isolated and characterized from hematophagous arthropods. Results {#s2} ======= Injection of ds*longistatin* inhibits the transcription and translation of endogenous longistatin {#s2a} ------------------------------------------------------------------------------------------------- Total RNA isolated from salivary glands of ticks of different feeding phases of both control and treated groups was analyzed by reverse transcription-PCR (RT-PCR) and quantitative RT-PCR (qRT-PCR) to demonstrate the effect of RNAi on the expression of longistatin-specific mRNA. The RT-PCR data revealed that injection of ds*longistatin* completely abolished the detectable mRNA expression corresponding to the longistatin-specific gene in ticks ([Figure 1A](#ppat-1001312-g001){ref-type="fig"}). Our qRT-PCR data also supported this finding. However, in the RNAi-treated group, longistatin-specific transcript, although very low compared with that of control, was detected in ticks at 24, 48 and 72 h of feeding only by qRT-PCR ([Figure 1B](#ppat-1001312-g001){ref-type="fig"}). This variation in the detection of longistatin-specific mRNA by RT-PCR and qRT-PCR may be due to the sensitivity of the techniques. Furthermore, the presence of a detectable level of mRNA in the RNAi-treated group of ticks might be due to individual variations in the tick population. Here, we used pools of salivary-gland extracts, isolated from three randomly selected ticks in each feeding phase; thus, it is quite reasonable that the effects of RNAi were not exactly uniform in each and every microinjected tick. Longistatin-specific gene expression was detected in its normal pattern [@ppat.1001312-Anisuzzaman1] in ticks injected with dsRNA complementary to the gene encoding maltose-binding protein in *Escherichia coli* (ds*mal*E) ([Figures 1A and B](#ppat-1001312-g001){ref-type="fig"}), indicating that injection of ds*longistatin* caused disruption of longistatin-specific mRNA transcription. For *in situ* detection of the effect of RNAi on the translation of endogenous longistatin, salivary glands were collected from partially fed (96 h) ticks of both treated and control groups and were subjected to immunofluorescence staining using mouse anti-longistatin sera. Longistatin-specific reactions were almost absent in ds*longistatin*-injected ticks whereas binding of mouse anti-longistatin antibody was detected in the salivary glands of ds*mal*E-injected ticks ([Figure 1C](#ppat-1001312-g001){ref-type="fig"}), suggesting that injection of ds*longistatin* efficiently silenced longistatin-specific mRNA expression and subsequent translation of longistatin. To further validate our results regarding *in vivo longistatin* gene silencing, we conducted Western blot analysis using salivary gland extracts collected from both treated and control ticks. Longistatin-specific bands were detected only in the salivary gland extracts of ds*mal*E-injected ticks and longistatin was upregulated with the feeding process of ticks ([Figure 1D](#ppat-1001312-g001){ref-type="fig"}). These findings further reinforced the evidence of longistatin-specific gene silencing by ds*longistatin* injection in ticks. ::: {#ppat-1001312-g001 .fig} 10.1371/journal.ppat.1001312.g001 Figure 1 ::: {.caption} ###### Post-transcriptional silencing of longistatin-specific gene in adult ticks by injecting dsRNA. \(A) Semiquantitative RT-PCR analysis. One microgram of *longistatin* dsRNA was injected into the hemolymph of ticks of the RNAi-treated group. Ticks of the control group were treated with 1 µg of *mal*E dsRNA. Actin was used as an internal control. Eng, engorged. (B) Quantitative RT-PCR using total RNA and primers specific for longistatin as in A. Eng, engorged. (C) *In situ* detection of longistatin expression in ticks\' salivary glands. Salivary glands from the ticks of control and RNAi-treated groups. Endogenous longistatin was reacted with mouse anti-longistatin sera (1∶100). (D) Effect of gene silencing on longistatin post-translation by Western blot analysis. Salivary gland extracts were electrophoresed and transferred onto nitrocellulose membrane. Endogenous longistatin was probed with mouse anti-longistatin (1: 100). Eng, engorged. :::  ::: RNAi-treated ticks failed to develop a blood pool and were unable to feed blood-meals from hosts {#s2b} ------------------------------------------------------------------------------------------------ All ticks microinjected with dsRNA were active and healthy during the incubation period. After placement on rabbits\' ears, all ticks of both treated and control groups actively attached. However, in the ds*longistatin*-injected group, 3 (2.5%) ticks were found dead at 72 h of attachment. All ticks in the ds*mal*E-injected group reached to repletion and detached by day 6 post-attachment. Notably, ds*longistatin* injection was shown to hamper the feeding of ticks. These ticks were poorly fed and most of them failed to engorge ([Figure 2A](#ppat-1001312-g002){ref-type="fig"}). Only two ticks (1.66%) dropped off the host following engorgement in the RNAi group. The mean body weight of the ticks collected after 6 days of feeding was significantly (P\<0.01) lower in the RNAi-treated group (53.53±50.38 mg) than that of the control group (253.43±57.91 mg) ([Figure 2B](#ppat-1001312-g002){ref-type="fig"}). Visible phenotypic changes were also obvious among the ticks of the treated and control groups. Ticks of the treated group, despite 6 days of feeding, were very small with a dull cuticle and devoid of normal cuticular wrinkling. In contrast, ticks of the control group, which dropped off the host following full engorgement, were large and glossy in appearance with dorsal cuticular wrinkling ([Figure 2B](#ppat-1001312-g002){ref-type="fig"}). ::: {#ppat-1001312-g002 .fig} 10.1371/journal.ppat.1001312.g002 Figure 2 ::: {.caption} ###### Effects of post-transcriptional silencing of *longistatin* gene on blood pool formation and blood feeding. \(A) Impact of longistatin-specific mRNA silencing on blood-meal feeding from hosts. Adult ticks were injected with *longistatin* dsRNA or *mal*E dsRNA (1 µg/tick) and were allowed to feed on a tick-naïve rabbit. RNAi-treated ticks failed to replete. (B) Postengorgement weight was significantly (p\<0.01) lower in the RNAi-treated group than that of the control group. Dotted lines indicate mean±SD of body weight of ticks. (C) Effects of post-transcriptional silencing of *longistatin* gene on blood pool formation. RNAi was performed and ticks were fed in the same manner as in A. RNAi-treated ticks failed to establish a prominent blood pool. Arrows indicate site of tick attachment. (D) Blood pools were significantly (p\<0.01) smaller in the RNAi-treated group. (E) Histopathological changes were studied using EVGS. Longistatin was detected in the feeding lesions of ticks on the host\'s tissues using mouse anti-longistatin sera (1∶100). :::  ::: A marked difference between blood pools induced by the ticks of RNAi-treated and control groups was observed. Large blood pools were developed at the site of attachment of each tick of the control group. Grossly, blood pools were large enough with an estimated mean size of 19.53±7.85 mm^2^, containing a considerable amount of exudates and blood. The affected area was cyanotic or dark bluish in color. Blood vessels in the vicinity of the blood pool were congested and distended. The surrounding area of the blood pool was markedly hyperemic. On the contrary, most of the ticks of the RNAi-treated group failed to establish a feeding lesion. Importantly, significantly (p\<0.01) smaller blood pools (4.25±6.38 mm^2^) were detected in few individual ticks. Also, no prominent gross pathological changes were detected at the site of attachment of RNAi-treated ticks ([Figures 2C and D](#ppat-1001312-g002){ref-type="fig"}). To study the histological features, we stained the thin tissue sections of the blood pool areas with Elastica-van Gieson stain (EVGS) where hemorrhagic areas exhibited a bright golden yellow color, indicating that blood pools, developed at the site of attachment of the ds*mal*E-injected ticks, were flooded with RBC. Notably, hemorrhagic changes were not detected at the biting areas of ticks of the RNAi-treated group ([Figure 2E](#ppat-1001312-g002){ref-type="fig"}), suggesting that the *longistatin* gene plays vital roles in the formation and maintenance of a blood pool as preceded by marked hemorrhage into tick-feeding lesions. In addition, an attempt was made to detect endogenous longistatin in the feeding lesions produced by the ticks of both groups using mouse anti-longistatin sera. Longistatin-specific bright green fluorescence was detected at the well-developed blood pool areas produced by the ds*mal*E-injected ticks but no such reaction was detected at the feeding lesions induced by ds*longistatin*-injected ticks ([Figure 2E](#ppat-1001312-g002){ref-type="fig"}), which further suggests a possible role of longistatin in the blood pool formation. Longistatin degrades fibrinogen and delays fibrin clot formation {#s2c} ---------------------------------------------------------------- To judge the anticoagulant function, longistatin was reacted with several coagulation factors (viz., thrombin, factors VIIa and Xa) using different approaches but none of them was affected by longistatin. Interestingly, longistatin was found to delay fibrin clot formation, degrade fibrinogen and efficiently activate plasminogen to its active form, plasmin. To evaluate the anticoagulation potential of longistatin, fibrinogen (7.5 mM) was pre-incubated in the absence or presence of longistatin (0.1, 0.2, 0.4, 0.8 and 1.6 µM) or plasmin (1.6 µM) and then thrombin (0.10 NIH unit/µl) was added as described in [Materials and Methods](#s4){ref-type="sec"}. Longistatin significantly (p\<0.01) interrupted the normal fibrin clot formation in a concentration-dependent manner. In the absence of longistatin, fibrin clot was formed within 15 min, but in the presence of longistatin (1.6 µM) or plasmin (1.6 µM), no visible clot was developed within this time ([Figure 3A](#ppat-1001312-g003){ref-type="fig"}). Longistatin was shown to delay the formation of a visible fibrin clot. Fibrin clotting time was significantly extended (up to 90 min) at a concentration of 1.6 µM longistatin ([Figure 3B](#ppat-1001312-g003){ref-type="fig"}), indicating potent anticoagulant activity of longistatin. To explore the fibrinogenolytic potential, fibrinogen was incubated in the absence or presence of longistatin (0.4, 0.8 and 1.6 µM) or plasmin (1.6 µM) and was subjected to SDS--PAGE analysis. Longistatin potently degraded the α, β and γ chains of fibrinogen in a concentration-dependent manner. Longistatin completely hydrolyzed all three chains of fibrinogen at 1.6 µM concentration, as it was done by 1.6 µM plasmin ([Figure 3C](#ppat-1001312-g003){ref-type="fig"}), implying that longistatin is an efficient fibrinogenolytic protease of ixodid ticks. ::: {#ppat-1001312-g003 .fig} 10.1371/journal.ppat.1001312.g003 Figure 3 ::: {.caption} ###### Anti-coagulation and fibrinogenolytic activity of longistatin. \(A) Effects of longistatin on the formation of fibrin clot. Fibrinogen (7.5 mM) was pre-incubated in a buffer in the absence or presence of longistatin (0.1, 0.2, 0.4, 0.8 and 1.6 µM) or plasmin (1.6 µM) and then thrombin was added (0.10 NIH unit/µl) as described in [Materials and Methods](#s4){ref-type="sec"}. Clot formation was detected visually and also by determining changes in turbidity at OD~450~ using a spectrophotometer after 15 min. (B) Longistatin (1.6 µM) delayed fibrin clot formation up to 90 min. Fibrinogen was incubated in the absence or presence of longistatin (1.6 µM) following the same procedures as mentioned in A and then thrombin was added. OD~450~ was measured at 15 min intervals. (C) Fibrinogenolytic effect of longistatin. Fibrinogen (7.5 mM) was incubated in the absence or presence of longistatin (0.4, 0.8 and 1.6 µM) or plasmin (1.6 µM). Samples were collected at the indicated time period and were subjected to 12.5% SDS--PAGE analysis under reducing conditions. A gradual degradation of the α, β and γ chains of fibrinogen was detectable with the concomitant deposition of degraded products. Asterisks (^\*^) indicate that the difference compared with the negative control group (buffer only) is significant as determined by Student\'s *t*-test with unequal variance (^\*^p\<0.05, ^\*\*^p\<0.01). :::  ::: Longistatin activates plasminogen in the presence of soluble fibrin {#s2d} ------------------------------------------------------------------- Initially, we determined the plasminogen activation potential of longistatin by measuring the amidolytic activity of activated plasminogen on the plasmin-specific fluorogenic substrate. Data revealed that the initial rate of plasminogen activation was proportional to the concentration of longistatin. Hydrolysis of plasmin-specific substrate sharply increased with the increase of concentration of longistatin. However, longistatin alone, even at a higher concentration (640 nM), was not able to induce hydrolysis of plasmin-specific fluorogenic synthetic substrate ([Figure 4A](#ppat-1001312-g004){ref-type="fig"}), indicating that longistatin activated plasminogen into its active form, plasmin, which hydrolyzed the synthetic substrate releasing MCA. Most interestingly, plasminogen activation potential of longistatin was significantly increased in the presence of soluble fibrin and induced robust amidolytic activity. Addition of fibrin cyanogen bromide (CNBr) fragments (4 µg) to the assay mixture caused an increase of plasminogen activation rate up to 4 times higher than that in the absence of fibrin CNBr fragments ([Figure 4B](#ppat-1001312-g004){ref-type="fig"}). CNBr fragments of fibrin, the soluble fibrin, is usually regarded as a t-PA stimulator and increases the rate of plasminogen activation about 5 times [@ppat.1001312-Zamarron1], [@ppat.1001312-Hoylaerts1]. Plasminogen was sparingly affected by longistatin in the absence of soluble fibrin ([Figure 4B](#ppat-1001312-g004){ref-type="fig"}), indicating the extraordinary specificity of longistatin towards fibrin-bound plasminogen rather than free circulating plasminogen. ::: {#ppat-1001312-g004 .fig} 10.1371/journal.ppat.1001312.g004 Figure 4 ::: {.caption} ###### Plasminogen activation by longistatin. \(A) Longistatin (40, 80, 160, 320 and 640 nM) was incubated without or with plasminogen (0.24 units) adding fibrin CNBr fragments (4 µg) in a total volume of 200 µl of buffer (50 mM Tris--HCl, pH 7.5; 100 mM NaCl and 5 mM CaCl~2~) at 25°C for 2 h. Then, plasmin-specific fluorogenic substrate (100 µM, final concentration) was added and substrate hydrolysis was monitored by measuring excitation and emission wavelengths of 360 nm and 460 nm, respectively, at 15 min intervals. *Inset*, initial rate of plasminogen activation at different concentrations of longistatin. (B) Effects of fibrin CNBr fragments on the activation of plasminogen by longistatin. Plasminogen (0.24 units) was incubated with longistatin (640 nM) in the absence or presence of fibrin CNBr fragments (0.25, 1 and 4 µg) as described in [Materials and Methods](#s4){ref-type="sec"}. All assays were performed in triplicate. :::  ::: Longistatin is able to lyse fibrin clot by activating plasminogen {#s2e} ----------------------------------------------------------------- Although plasmin has enzymatic activity towards a broad spectrum of substrates, but the fibrin clot is considered as its main native substrate [@ppat.1001312-Wiman1]. To further evaluate the plasminogen activation efficiency of longistatin, fibrin polymer was incubated with plasminogen--longistatin/-t-PA mixture or with buffer only at 25°C for 24 h. Data revealed that, in the presence of either longistatin or t-PA, plasminogen was able to induce lysis of fibrin clot. Visual observation and spectrometric analysis at 450 nm (OD~450~) revealed that the fibrinolytic potential of plasminogen, in the presence of longistatin, was directly proportional to the concentration of the recombinant protein used. These results firmly suggest that longistatin activates plasminogen to an active plasmin in a concentration-dependent manner, which in turn lyses the visible fibrin clot. Longistatin caused complete lysis of fibrin clot in the nanomolar range ([Figures 5A and B](#ppat-1001312-g005){ref-type="fig"}). SDS--PAGE analysis showed that longistatin efficiently cleaved plasminogen into its heavy and light chains. These results are comparable to those of purified, active and commercially available t-PA ([Figure 5C](#ppat-1001312-g005){ref-type="fig"}). Here, we also observed that longistatin activated a sufficient amount of plasminogen only in the presence of fibrin clot and induced profound fibrinolysis (data not shown). ::: {#ppat-1001312-g005 .fig} 10.1371/journal.ppat.1001312.g005 Figure 5 ::: {.caption} ###### Longistatin induced fibrinolysis by activating plasminogen. \(A) Fibrin clot was formed by incubating fibrinogen (7.5 mM) and thrombin (0.10 NIH unit/µl) and was incubated in the presence of plasminogen-t-PA/-longistatin (40, 80, 160, 320 and 640 nM) mixture or buffer only at 25°C for 24 h. Clot lysis was measured at OD~450~. Plasminogen induced complete lysis of fibrin clot in the presence of 640 nM longistatin or 154 nM t-PA. (B) Time-dependent activation of plasminogen by longistatin with concomitant lysis of fibrin clot. (C) Cleavage of plasminogen into the heavy and light chains. Digested product of fibrin clot was electrophoresed by 12.5% SDS--PAGE. Asterisks (^\*^) indicate that the difference compared with the negative control group (buffer only) is significant as determined by Student\'s *t*-test with unequal variance (^\*^p\<0.05, ^\*\*^p\<0.01, ^\*\*\*^p\<0.001). :::  ::: Longistatin binds with fibrin {#s2f} ----------------------------- To explore the fibrin-binding capability of longistatin, we conducted fibrin-binding assays. Longistatin-specific green fluorescence was detected in the fibrin meshwork when purified recombinant longistatin was incorporated with fibrin and reacted with anti-longistatin sera. However, longistatin-specific fluorescent reaction was completely absent in the absence of longistatin in the reaction mixture or when longistatin-impregnated fibrin meshwork was treated with pre-immune sera (1∶100), indicating that longistatin is able to bind with fibrin clot. To compare the fibrin binding of longistatin, we used t-PA as a positive control because t-PA is known to bind with fibrin [@ppat.1001312-Li1]. Here, t-PA-specific fluorescence was only detected when t-PA added fibrin was treated with anti-t-PA mouse monoclonal antibody (1∶100), but no such reaction was visible in the presence of pre-immune sera (normal serum) at the same concentration. To evaluate the specificity of the fibrin-binding assays, we used u-PA as a negative control because u-PA does not bind with fibrin [@ppat.1001312-Murray2]. u-PA-specific reaction was not detected in the fibrin meshwork even in the presence of a relatively high concentration (1∶20) of anti-u-PA antibody ([Figure 6A](#ppat-1001312-g006){ref-type="fig"}). By using SDS−PAGE analysis, we also observed that the concentration of residual longistatin was markedly reduced after fibrin clot formation ([Figure 6B](#ppat-1001312-g006){ref-type="fig"}), which reinforced the assertion that longistatin was bound with the fibrin clot. Furthermore, we determined the concentration of longistatin/t-PA in the supernatant obtained from the fibrin-binding assay mixer and data revealed that the percentage of binding of longistatin/t-PA with fibrin meshwork was directly proportional to the amount of fibrin. Longistatin binding was 97.93%±1.68% when 60 mM fibrinogen was used to produce fibrin clot and, under the same conditions, binding of t-PA was 90.6%±1.2%. Specificity of the binding was evidenced by the absence of u-PA binding ([Figure 6C](#ppat-1001312-g006){ref-type="fig"}). To evaluate the fibrin-binding potentials of longistatin, we determined the fibrin-binding parameters of longistatin and compared them with those of t-PA. Longistatin was shown to bind with fibrin with the estimated K~d~, B~max~ and molar binding ratio (MBR) of 145.5±3.3 nmol/L, 3.1±0.6 µmol/L and 42.3±7.4, whereas those of t-PA were 159.2±7.4 nmol/L, 1.4±0.4 µmol/L and 19.3±4.7, respectively ([Table 1](#ppat-1001312-t001){ref-type="table"}), suggesting that longistatin potently binds with fibrin. Fibrin binding is an essential feature for the plasminogen activators that specifically activate fibrin clot-bound plasminogen. For example, t-PA is a very weak activator of plasminogen in the absence of fibrin. The affinity between t-PA and plasminogen is significantly increased in the presence of fibrin of blood clot [@ppat.1001312-CesarmanMaus1], [@ppat.1001312-Tate1]. Fibrin binding was also assumed as critical in the activation process of plasminogen by longistatin. ::: {#ppat-1001312-g006 .fig} 10.1371/journal.ppat.1001312.g006 Figure 6 ::: {.caption} ###### Binding of longistatin with fibrin clot. \(A) Detection of longistatin bound on fibrin meshwork. Fibrinogen at different concentrations (3.75, 7.5, 15, 30 and 60 mM; final concentration) was mixed in the absence or presence of longistatin (10 µg) or an equal amount of t-PA or u-PA in a buffer (50 mM Tris--HCl, pH 7.5; 100 mM NaCl and 5 mM CaCl~2~) and thrombin (0.10 NIH unit/µl) was added immediately and was incubated at 25°C for 1 h. The clot was treated with anti-longistatin (1∶100), anti-t-PA (1∶100), anti-u-PA (1∶20) or pre-immune sera (1∶100). Bound antibodies were detected using green fluorescent-labeled secondary antibody (Alexa Flour 488 goat anti-mouse IgG). (B) Supernatant was analyzed by 12.5% SDS--PAGE under reducing conditions. (C) The target protein was extracted from the supernatant and its concentration was determined using micro-BCA reagent as described in [Materials and Methods](#s4){ref-type="sec"}. The results are expressed as percentage of longistatin/t-PA/u-PA bound to the fibrin clot. Data represent mean ± SD, n = 3. :::  ::: ::: {#ppat-1001312-t001 .table-wrap} 10.1371/journal.ppat.1001312.t001 Table 1 ::: {.caption} ###### Comparison of fibrin-binding parameters of longistatin with those of t-PA. ::: {#ppat-1001312-t001-1} Protein K~d~ (nmol/L) B~max~ (µmol/L) MBR[a](#nt101){ref-type="table-fn"} ------------- --------------- ----------------- ------------------------------------- Longistatin 145.5±3.3 3.1±0.6 42.3±7.4 t-PA 159.2±7.4 1.4±0.4 19.3±4.7 p\<0.05 p\<0.01 p\<0.01 a Expressed as moles of longistatin/t-PA per mole of fibrin. Determined at maximum binding. ::: Longistatin activates plasminogen present in plasma milieu and induces thrombolysis {#s2g} ----------------------------------------------------------------------------------- To demonstrate the thrombolytic capability of longistatin, we treated freshly prepared platelet-rich thrombi kept in fresh plasma. Longistatin was able to cause lysis of platelet-rich thrombi in the presence of fresh plasma in a concentration-dependent manner and efficiently lysed thrombi at 640 nM concentration ([Figure 7A](#ppat-1001312-g007){ref-type="fig"}). Longistatin induced more than 50% lysis of thrombi within 2 h at 640 nM concentration (data not shown). In the same experimental setup, longistatin (640 nM) and t-PA (154 nm) induced 93.62%±2.33% and 98.78%±2.11% lysis of thrombi, respectively, by 12 h ([Figure 7B](#ppat-1001312-g007){ref-type="fig"}). Moreover, like t-PA, longistatin efficiently digested fibrin clot produced from purified, commercially available fibrinogen and thrombin in the presence of plasma (data not shown). Taken together, our results suggest that longistatin is capable of causing thrombolysis and subsequent recanalization of occluded thrombosed vascular tree by activating the physiological level of plasminogen into plasmin. ::: {#ppat-1001312-g007 .fig} 10.1371/journal.ppat.1001312.g007 Figure 7 ::: {.caption} ###### Lysis of platelet-rich thrombi by longistatin in dog plasma. \(A) Longistatin hydrolyzes platelet-rich thrombi *ex vivo* in dog plasma. Platelet-rich clot was produced by incubating 0.2 ml of dog blood and the thrombi were treated with longistatin at various concentrations in 0.5 ml of dog plasma at 37°C for 12 h and weighed at the indicated period. (B) Comparison of thrombolytic activity of longistatin with that of t-PA. Thrombi were treated with longistatin (640 nM)/t-PA (154 nM) under the same *ex vivo* experimental conditions. :::  ::: Functional implications of longistatin in blood coagulation and fibrinolysis cascades {#s2h} ------------------------------------------------------------------------------------- We hypothesized that ixodid ticks synthesize longistatin and possibly other functionally related bioactive molecules to efficiently counteract host\'s hemostatic ability and/or to activate its own fibrinolytic machinery to create blood pools for the acquisition of blood-meals and engorgement. Our *in vivo* and *in vitro* data strongly support this hypothesis and that longistatin exerts its multifunctional roles both in coagulation cascades and in fibrinolytic pathways. In this study, we proposed a longistatin-induced anti-coagulation and fibrinolytic mechanism that works persistently against host\'s hemostatic pathways until ticks ensure full blood-meals ([Figure 8](#ppat-1001312-g008){ref-type="fig"}). ::: {#ppat-1001312-g008 .fig} 10.1371/journal.ppat.1001312.g008 Figure 8 ::: {.caption} ###### A schematic diagram showing roles of longistatin in blood coagulation and fibrinolysis events. In the initial phase, the tick bites and lacerates tissues at the site of attachment and damages vascular beds, which results in hemorrhage leading to the development of a blood pool. Longistatin is synthesized in and secreted from the salivary glands and injected into the blood pool during feeding process [@ppat.1001312-Anisuzzaman1]. Longistatin degrades fibrinogen and activates plasminogen to its active form, plasmin. HMWK, high-molecular-weight kininogen; PKK, prekallikrein; TF, tissue factor. Yellow arrows, contact activation (intrinsic) pathway; olive-green arrows, tissue factor (extrinsic) pathway; green arrows, common pathway of coagulation cascade and white arrows, fibrinolytic pathway. Figure adapted from ref. [@ppat.1001312-Stark1], [@ppat.1001312-CesarmanMaus1], [@ppat.1001312-Murray2]. :::  ::: Discussion {#s3} ========== Previous literatures on the feeding behavior and physiology of hematophagous ixodid ticks suggest that acquisition of blood-meals from mammalian hosts is modulated by a vast array of pharmacologically active bio-molecules secreted from the salivary glands of these amazing tiny arthropods. Ticks\' saliva is thought to efficiently manipulate the hosts\' strong defense mechanisms, such as hemostasis, inflammatory reactions and immune responses that are induced against the tick during hematophagy [@ppat.1001312-Stark1], [@ppat.1001312-MaritzOlivier1], [@ppat.1001312-Monteiro1]--[@ppat.1001312-vandeLocht1]. In the present study, we provide evidence that longistatin, a salivary gland protein with two functional EF-hand Ca^++^-binding domains isolated from the ixodid tick, *H. longicornis*, binds with fibrin and specifically activates plasminogen to its active form, plasmin. We also demonstrate that longistatin degrades fibrinogen and is able to delay fibrin clot formation; thus, longistatin appears to be essential to keep the blood in a fluid state in the blood pool, which enables ticks to feed and replete on blood-meals from hosts. Our *in vivo* gene silencing study revealed that longistatin-specific gene-knockdown ticks were unable to produce a blood pool and consequently failed to replete. These findings prompted us to unveil the underlying complex mechanism(s) by which longistatin exerts its modulatory roles in blood pool formation and maintenance, which is an essential pre-requisite to ticks for successful feeding and engorgement. Anticoagulation of host\'s blood is a crucial step for the development and maintenance of a blood pool. To verify this, we have conducted several anticoagulation and fibrinolysis assays. Interestingly, we have shown that longistatin significantly extends the time of fibrin clot formation up to 90 min when fibrinogen is treated with longistatin compared with untreated fibrinogen (control), which takes only 15 min to complete fibrin clot formation. Data on fibrinogenolytic assays also support this observation that longistatin is capable of degrading the α, β and γ chains of fibrinogen in a concentration-dependent manner and completely hydrolyzes these three bands. Therefore, it may be assumed that the delay in the fibrin clot formation during *in vitro* anticoagulation assays in the presence of longistatin is due to the gradual degradation of coagulable proteins in the reaction mixture. Fibrinogenolytic activity has also been reported in tick metalloprotease isolated from *Ixodes scapularis* [@ppat.1001312-Francischetti1] and in several proteases identified from snake venoms [@ppat.1001312-Matsui1] and spider toxins [@ppat.1001312-daSilveira1]. Fibrinogen is the key component for the formation of cross-linked fibrin polymer. In general, after laceration of blood vessels, platelets are exposed to the subendothelial collagen and become activated and aggregated around the injuries. Then, fibrin strands, derived from cleavage of fibrinogen, intertwine about the aggregated platelets giving rigidity and stability of the initial and preliminary platelet plugs. Finally, activated factor XIII forms covalent bonds that crosslink the fibrin polymers and thus the injured blood vessels are sealed by an extra strengthened, stable fibrin clot leading to hemostasis [@ppat.1001312-Stark1], [@ppat.1001312-Furie1], [@ppat.1001312-MacFarlane1]. Therefore, it may be predicted that fibrinogenase activity of longistatin hampers hemostasis and may facilitate hemorrhage into the blood pools from which ticks persistently feed blood-meals. We have shown that longistatin specifically activated plasminogen in the presence of fibrin clot and cleaved it into plasmin, heavy chain (α) and light chain (β). Although plasminogen is unable to cleave fibrin clot but it has a strong affinity for fibrin. Plasminogen has a secondary structure known as a kringle domain, which anchors plasminogen specifically to the carboxy-terminal arginine and lysine residues of the fibrin; thus, plasminogen becomes more concentrated on the surface of the clot, whenever and wherever it develops. As soon as plasminogen is converted into plasmin by its activators, it functions like a serine protease. Fibrin acts as a cofactor in the enzymatic activation reactions when plasminogen is activated by the fibrin-selective agents, like t-PA and its derivatives or staphylokinase and its derivatives. Through a highly orchestrated biochemical process, plasmin initially creates nicks on the fibrin and further digestion leads to the complete dissolution of fibrin meshwork into soluble fibrin degradation products [@ppat.1001312-CesarmanMaus1], [@ppat.1001312-Murray2], [@ppat.1001312-Nesheim1]. Our results clearly demonstrate that longistatin contributes to this well-coordinated fibrinolytic pathway by activating plasminogen into plasmin. Plasminogen activators have been purified from the venomous snake, *Trimeresurus stejnegeri* [@ppat.1001312-Zhang1] and from the saliva of common vampire bat, *Desmodus rotundus* [@ppat.1001312-Gardell1], [@ppat.1001312-Liberatore1]. In vampire bats, plasminogen activator is thought to be a key enzyme that plays a crucial role in the maintenance of the flow of blood during the feeding process [@ppat.1001312-TellgrenRoth1] and snake-venom plasminogen activator is associated with the pathogenesis of envenomation rendering the blood incoagulable [@ppat.1001312-Zhang1]. Maintenance of hosts\' blood fluidity at the site of biting is also very critical for the development of blood pool and subsequent blood feeding, and eventually for the survival of ixodid ticks. From the available evidence, here we assume that fibrin-specific, plasminogen-activation-dependent thrombolytic potentiality of longistatin plays significant functional roles in the pathobiology of vector ticks through the formation of feeding lesion and by the maintenance of its homeostasis throughout the entire period of blood feeding. In conclusion, longistatin, a salivary-gland protein of ixodid ticks with multitarget potential, shows high specificity for fibrin clot-bound plasminogen, and degrades fibrinogen. These findings indicate that longistatin plays significant roles in the formation and maintenance of blood pools leading to the successful acquisition of blood-meals and may be critical for the survival of ixodid ticks. Furthermore, our data suggest that longistatin may serve as a novel therapeutic target against ticks and tick-borne diseases, including human diseases such as thrombosis or other occlusive cardiovascular accidents. Materials and Methods {#s4} ===================== Ticks and animals {#s4a} ----------------- We propagated parthenogenetic Okayama strains of *H. longicornis* at the Laboratory of Parasitic Diseases, National Institute of Animal Health (NIAH), Tsukuba, Japan, by feeding on the ear of tick-naïve, specific-pathogen-free Japanese White rabbits according to methods described previously [@ppat.1001312-Anisuzzaman1]. Ticks were collected after detachment following full engorgement or at the indicated period of feeding. After collection, ticks were weighed and phenotypic differences between the ticks of RNAi and control groups were recorded. This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Animal Health. The protocol was approved by the Committee of the Ethics of Animal Experiments of the NIAH (Permit Number: 09-017, 09-018, 10-008, 10-010). All surgeries were performed under sodium pentobarbital anesthesia, and all efforts were made to minimize the animals\' suffering. Reagents {#s4b} -------- Green fluorescent-labeled secondary antibody (Alexa Flour 488 goat anti-mouse IgG (H+L)) was purchased from Invitrogen and alkaline phosphatase-conjugated goat anti-mouse IgG (H+L) was from ZYMED. Nitroblue tetrazolium/5-bromo-4-chloro-3-indolyl phosphate (BCIP/NBT) and T7 RNA polymerase were from Promega. Purified human plasmin, bovine thrombin and fibrinogen were obtained from Sigma. Purified human plasminogen, t-PA and anti-t-PA (Ab-1) mouse monoclonal antibody (GMA-043) were purchased from Calbiochem. Human u-PA was from Cosmo Bio Co. LTD and polyclonal antibody to u-PA was from Acris Antibodies GmbH. Fibrin CNBr fragment was obtained from Technoclone and Boc-Glu-Lys-Lys-MCA from Peptide Institute. Total RNA extraction kit and DNA gel extraction kit were purchased from QIAGEN. RNA interference {#s4c} ---------------- We performed RNAi using dsRNA following a protocol described previously [@ppat.1001312-Islam1]. The sequence coding longistatin was cloned into pBluescript II SK^+^ vector in *Xho*I and *Eco*RI restriction sites following a protocol described previously [@ppat.1001312-Anisuzzaman1]. The dsRNA complementary to the *E. coli mal*E gene that encodes maltose-binding protein was used as a negative control. cDNA corresponding to *mal*E mRNA was synthesized and was cloned into pBluescript II SK^+^ plasmid using the primers 5′-CCGCTCGAGCGGTTATGAAAATAAAAACAGGTGCA-3′ and 5′-GAATTCGCTTGTCCTGGAACGCTTTGTC-3′. The inserted sequences of longistatin and *mal*E were amplified by PCR using primers T7 (5′-GTAATACGACTCACTATAGGGC-3′) and CMO422 (5′-GCGTAATACGACTCACTATAGGGAACAAAAGCTGGAGCT-3′) to attach to T7 promoter recognition sites at either end. The PCR products were purified using gel extraction kit (QIAGEN). dsRNA complementary to the respective DNA inserts was synthesized by *in vitro* transcription using T7 RNA polymerase (Promega). One microgram of longistatin dsRNA (ds*longistatin*, treated group) or *mal*E dsRNA (ds*mal*E, control group) was injected into each tick through the 4^th^ coxa. Ticks were observed in a humified incubator for 24 h at 25°C prior to attaching them on the host for feeding. A total of 120 ticks, each of 60 in RNAi and control groups, were attached on the ear of tick-naïve rabbits. Ticks were collected at 24, 48, 72 and 96 h of feeding or after repletion. All ticks collected after they had dropped off the host following full engorgement were weighed individually using a digital balance (Sartorius). Semiquantitative RT-PCR and qRT-PCR {#s4d} ----------------------------------- Salivary glands from adult ticks of both control and RNAi groups at different feeding periods (24, 48, 72, 96 h and engorged) were collected as described previously [@ppat.1001312-Anisuzzaman1]. Shortly after collection, salivary glands were submerged in RNA*later*, an RNA Stabilization Reagent (QIAGEN). Total RNA was isolated using an RNeasy Mini Kit (QIAGEN) according to the manufacturer\'s protocol and 500 ng of total RNA was used for reverse transcription before PCR. Single-stranded cDNA was prepared using Takara RNA PCR Kit (AMV) Ver.3.0 (Takara) following the manufacturer\'s instructions. A series of PCRs were carried out using 500 ng of cDNA from each sample and longistatin-specific oligonucleotides (5′GCTATCTCGGCTCCTGTGTC 3′ and 5′ATCTTCGCCAGGTCCTTCTT 3′) or oligonucleotides specific for a control cDNA encoding β-actin in a final volume of 20 µl. The PCR product was subjected to electrophoresis in 1% agarose gel. The qRT-PCR was performed in a LightCycler 1.5 instrument (Roche Instrument Centre AG) using LightCycler FastStart DNA Master SYBR Green I (Roche Diagnostics) following a procedure described previously [@ppat.1001312-Tsuji1]. The reaction mixture of 20 µl contained 4 mM MgCl~2~, 0.5 µM of each primer (forward and reverse as described above) and 2 µl (250 ng/µl) of the single-stranded DNA template. The data obtained were analyzed using LightCycler Software Version 3.5. Recombinant longistatin and polyclonal antibody production {#s4e} ---------------------------------------------------------- Recombinant longistatin was produced, purified, dialyzed and its concentration was determined as previously described [@ppat.1001312-Anisuzzaman1]. Purified longistatin was kept at --20°C until further use. Polyclonal antibody was produced in BALB/c mice following a procedure described elsewhere [@ppat.1001312-Anisuzzaman1]. Immunofluorescence and histopathology {#s4f} ------------------------------------- We collected salivary glands from partially fed (96 h) adult ticks of both control and RNAi groups as described above. Salivary glands were placed on slides and were fixed with 4% paraformaldehyde. Salivary glands were then permeabilized with 0.1% tritonX-100 and were treated with mouse anti-longistatin sera (1∶100). Bound antibodies were detected using green fluorescent-labeled secondary antibody (Alexa Flour 488 goat anti-mouse IgG (H+L), Invitrogen). Slides were mounted with VECTASHIELD mounting medium containing 4′,6-diamidino-2-phenylindole (DAPI, Vector Laboratories) and examined under a fluorescent microscope (Leica). In addition, thin sections were prepared from tissues collected from the feeding lesions developed on a rabbit\'s ear at the site of attachment of the ticks. Tissue sections were then subjected to immunofluorescent staining using mouse anti-longistatin sera as described previously[@ppat.1001312-Anisuzzaman1]. Rabbit\'s tissues were also stained with EVGS as described previously [@ppat.1001312-Chino1]. Immunoblot analysis {#s4g} ------------------- Equal numbers of adult ticks from both RNAi and control groups collected at different feeding intervals (24, 48, 72, 96 h and engorged) were dissected separately in PBS and salivary glands were isolated. Antigens were prepared as previously described [@ppat.1001312-Alim1]. Equal amounts of protein (4 µg) were separated by 12.5% SDS−PAGE under reducing conditions and were transferred onto nitrocellulose membrane. The membrane was treated with mouse anti-longistatin sera (1∶100) overnight at 4°C. Bound antibodies were probed with alkaline phosphatase-conjugated goat anti-mouse IgG (H+L) (ZYMED). The membranes were developed with nitroblue tetrazolium/5-bromo-4-chloro-3-indolyl phosphate (BCIP/NBT, Promega). Anticoagulation and fibrinogenolytic assays {#s4h} ------------------------------------------- Fibrinogen (7.5 mM) was pre-incubated in a total volume of 397 µl of buffer containing 25 mM Hepes (pH 7.2) and 25 mM NaCl in the absence or presence of longistatin (0.1, 0.2, 0.4, 0.8 and 1.6 µM) or plasmin (1.6 µM) at 25°C for 3 h. Fibrin clot formation was initiated by the addition of 3 µl of thrombin (0.10 NIH unit/µl). Fibrin clot formation was detected visually and also by determining changes in turbidity at OD~450~ using a spectrophotometer (Beckman Coulter) at 15 min intervals. To detect the fibrinogenolytic activity of longistatin, fibrinogen (7.5 mM) was incubated in a total volume of 100 µl of buffer (25 mM Hepes, pH 7.2, and 25 mM NaCl) in the absence or presence of longistatin (0.4, 0.8 and 1.6 µM) or plasmin (1.6 µM) at 25°C for 48 h. Aliquots were collected and separated by 12.5% SDS--PAGE. Plasminogen activation assays {#s4i} ----------------------------- Longistatin at various concentrations (40, 80, 160, 320 and 640 nM) was incubated in a 96-well cell culture plate without or with plasminogen (0.24 units) in the absence or presence of fibrin CNBr fragments (0.25, 1 and 4 µg) in a total volume of 200 µl of buffer (50 mM Tris--HCl, pH 7.5; 100 mM NaCl and 5 mM CaCl~2~) at 25°C for 2 h. Then, 2 µl (100 µM, final concentration) of plasmin-specific fluorogenic substrate (Boc-Glu-Lys-Lys-MCA) was added. Substrate hydrolysis was monitored by measuring excitation and emission wavelengths of 360 nm and 460 nm, respectively, at 15 min intervals using a Spectra Fluor fluorometer (TECAN, Männedorf, Switzerland). All assays were performed in triplicate and activity of activated plasminogen was expressed in an arbitrary fluorescent unit per min (AFU/min). Fibrinolytic assays {#s4j} ------------------- Fibrin clot was incubated in the presence or absence of longistatin. An initial fibrin clot was produced by incubating 10 µl of fibrinogen (7.5 mM) and 5 µl of thrombin (0.10 NIH unit/µl) in a total volume of 400 µl of buffer (50 mM Tris--HCl, pH 7.5; 100 mM NaCl and 5 mM CaCl~2~) at 25°C for 1 h. An equal volume of plasminogen (2.4 units), t-PA (154 nM) or longistatin at various concentrations (40, 80, 160, 320 and 640 nM) in a total volume of 100 µl of buffer, as mentioned above, was added to the fibrin clot and incubated at 25°C for 24 h. Lysis of fibrin clot was detected visually and also by measuring changes in turbidity at OD~450~ using a spectrophotometer (Beckman Coulter) at different time intervals (0, 6, 12 and 24 h). Furthermore, to verify the plasminogen activation potential of longistatin, 20 µl of digested products of clots corresponding to each concentration were collected and subjected to 12.5% SDS--PAGE analysis under reducing conditions. Fibrin clot binding assays {#s4k} -------------------------- Fibrin binding assays were conducted with the modifications of procedures described previously [@ppat.1001312-Li1]. Briefly, purified fibrinogen (3.75, 7.5, 15, 30 and 60 mM; final concentration) was mixed in the absence or presence of longistatin (10 µg) or an equal amount of t-PA or u-PA in a total volume of 200 µl of buffer (50 mM Tris--HCl, pH 7.5; 100 mM NaCl and 5 mM CaCl~2~). Fibrin clot formation was initiated by adding 3 µl of thrombin (0.10 NIH unit/µl) immediately and was incubated at 25°C for 1 h. The clot was centrifuged at 10,000 *g* for 10 min and supernatant was collected. The remaining clot was extensively washed in PBS and treated with anti-longistatin (1∶100), anti-t-PA (1∶100), anti-u-PA (1∶20) or pre-immune sera (1∶100) overnight at 4°C. Bound antibodies were detected using green fluorescent-labeled secondary antibody. Supernatant was analyzed by 12.5% SDS−PAGE under reducing conditions. The target protein band was excised and protein was extracted from the gel using negative zinc staining kit following the manufacturer\'s instructions (Bio-Rad). Double volume of cold (−20°C) acetone was mixed thoroughly with the extracted protein, incubated at −20°C for 1 h and centrifuged at 10,000 *g* for 30 min. The pellet was air-dried and dissolved with distilled water. The concentration of the extracted protein was determined using micro-BCA reagents (Pierce). To determine the binding parameters, longistatin (2−10 µg) or an equal amount of t-PA was incorporated into the fibrin clot. Residual longistatin/t-PA was extracted from the supernatant and the concentration of the relevant protein was determined following the same methods as mentioned above. K~d~, B~max~ and MBR were calculated according to the procedures as previously described [@ppat.1001312-Siddiqi1]. *Ex vivo* thrombolysis assays {#s4l} ----------------------------- Platelet-rich thrombi were produced by incubating 0.2 ml of fresh dog blood in a 96-well flat-bottom cell culture plate for 15 min. Thrombi were removed and washed gently with normal saline and weighed using a digital balance (Sartorius). The thrombi were then treated with t-PA (154 nM) or longistatin at various concentrations (40, 80, 160, 320 and 640 nM) in a total volume of 0.5 ml of fresh dog plasma at 37°C for 12 h and weighed at 3 h intervals. Furthermore, fibrin clot was produced by incubating purified, commercially available fibrinogen (7.5 mM) and thrombin (0.10 NIH unit/µl) as mentioned above and was treated with t-PA (154 nM) or longistatin (640 nM) in fresh dog plasma following the same experimental procedures and conditions. Accession numbers of the proteins and the genes used {#s4m} ---------------------------------------------------- Plasmin from human plasma (**NP\_000292**), plasminogen from human plasma (**AAN85555**), t-PA from human plasma (**NP\_000921**), u-PA from human plasma (**NP\_002649.1**), thrombin (as prothrombin) from bovine plasma (**NP\_776302**), fibrinogen from bovine plasma (α chain, **AAI02565**; β chain, **NP\_001136389** and γ chain, **AAI02630**), longistatin from *H. longicornis* (**AB519820**) and *mal*E from *E. coli* (**ZP\_02999303**). Statistical analysis {#s4n} -------------------- Data were presented as mean ± standard error, where appropriate. Statistical significance was determined using Student\'s *t* test with unequal variance. We thank H. Shimada and M. Kobayashi for their generous help in preparing histological sections. The authors have declared that no competing interests exist. This work was supported by Grant-in-aids from the Ministry of Education, Culture, Sports, Science, and Technology of Japan and also by a grant from the Program for Promotion of Basic Research Activities for Innovative Biosciences. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: A KF NT. Performed the experiments: A MKI MAA TM TH KY NT. Analyzed the data: MAA TM TH KY. Contributed reagents/materials/analysis tools: A MAA TM TH KY. Wrote the paper: A MKI YM KF NT.

PubMed Central

2024-06-05T04:04:19.721602

2011-3-10

PMC3053354

Introduction {#s1} ============ Leaf-cutting ants are conspicuous New-World herbivores comprised of ca. 45 species divided into two genera, *Atta* and *Acromyrmex*. They form the monophyletic crown group of the attine fungus-growing ants, capable of harvesting over half a ton per year of live vegetation per colony and playing a significant role in nutrient transport and (sub)tropical ecosystems [@pone.0017506-Hlldobler1], [@pone.0017506-DeFineLicht1]. The plant cell wall is composed of structurally complex polysaccharides which are compactly arranged and extremely resistant to degradation [@pone.0017506-Pauly1]. In order to deconstruct plant biomass into usable nutrients such as simpler sugars, leaf-cutting ants rely on an obligate symbiosis with fungus gardens that are assemblies of microbes dominated by the basidiomycete fungus *Leucocoprinus gongylophorus* (Agaricales: Agaricaceae) [@pone.0017506-Hlldobler1], [@pone.0017506-Mikheyev1]. While leaf-cutting ants provide their fungus gardens with plant material to sustain its growth, the fungus provides food for the ants and their brood in the form of specialized inflated hyphal tips (gongylidia) which the ants excise, consume and feed to their larvae [@pone.0017506-Mueller1], [@pone.0017506-Mueller2]. Fungus gardens are maintained in underground nest chambers where worker ants provide a clean environment for garden growth and express multiple hygienic behaviours to inhibit parasitic fungi and other unwanted microorganisms, usually assisted by a combination of aseptic glandular secretions and symbiotic bacteria producing antibiotics [@pone.0017506-Hlldobler1], [@pone.0017506-Poulsen1]. The unique characteristics of these multipartite ant-symbiont relationships have led this mutualism to become a model system for studying social evolution at multiple levels [@pone.0017506-Mueller3]--[@pone.0017506-Boomsma1]. *Atta* and *Acromyrmex* workers deposit small leaf fragments in the upper and outer-most regions of the fungus garden, which are then progressively metabolized and transformed into fungal biomass in the middle and lower sections [@pone.0017506-Hlldobler1], [@pone.0017506-Schitt1]. This implies that different stages of plant degradation are accomplished in consecutive sections of the garden, which is to some extent reflected in their visual appearance: a dark colored top layer with newly incorporated leaf material, a middle layer where the fungal biomass increases substantially and where clusters of gongylidia are most abundant [@pone.0017506-Schitt1], and a bottom layer with dense mycelial biomass and the remaining non-degraded plant substrate. Exhausted fungus garden material is continuously removed from the lowest sections by the ant workers and deposited in debris piles away from the fungus garden [@pone.0017506-Bot1], [@pone.0017506-Hart1]. The ability of fungus gardens to efficiently degrade and metabolise fresh leaf material may explain why *Atta* leaf-cutting ants in particular have become such complex and highly evolved animal societies with colonies of up to five million workers and extensive division of labour among worker castes [@pone.0017506-Mueller2], [@pone.0017506-Hlldobler2]. However, the precise mechanisms and sequence of degradation events in fungus gardens remain obscure. Relative proportions of plant substrates in consecutive garden sections are little understood and we have no knowledge about the extent to which plant cell wall properties affect the ants\' selection criteria for accepting plant substrates into the garden, and for discarding old garden material with unused substrate. Without such information it is impossible to fully understand the dynamic processes that underpin plant biomass conversion in this symbiosis, and the resulting ecological footprint of these agricultural pest ants, which cause billions of dollars worth of damage each year [@pone.0017506-Hlldobler1]. Previous studies have utilised information about enzyme activities to infer aspects of substrate degradation both in naturally maintained fungus gardens [@pone.0017506-DEttorre1]--[@pone.0017506-DeFineLicht2] and in symbiont cultures grown *in vitro* [@pone.0017506-GomesdeSiqueira1]--[@pone.0017506-Silva2]. These studies suggest that enzyme activities originate primarily from the symbiotic fungus, but that yeasts and bacteria residing in the fungus garden may also contribute [@pone.0017506-Bacci1]--[@pone.0017506-Suen1]. Taken together, they indicated that *L. gongylophorus* mainly degrades proteins, starch, and plant cell wall polysaccharide components such as pectins and cross-linking glycans (also known as 'hemicelluloses'), whereas cellulose remains largely intact. However, this indirect evidence remains controversial [@pone.0017506-Hlldobler2], [@pone.0017506-Abril1], [@pone.0017506-Abril2], [@pone.0017506-Cherrett1]--[@pone.0017506-Bucher1], because most enzyme assays used single or very few highly specific substrates at any one time, which is problematic because the degradation of individual plant cell wall polysaccharides often requires the simultaneous action of complex multi-enzyme systems [@pone.0017506-Cooke1], [@pone.0017506-Sinsabaugh1]. A recently established technique, comprehensive microarray polymer profiling (CoMPP), utilizes carbohydrate microarray-based technology to obtain detailed information about the relative abundance of numerous plant cell wall polysaccharides within a set of biological samples [@pone.0017506-Moller1]--[@pone.0017506-Domozych1]. This technology is underpinned by the availability of a large number of monoclonal antibodies (mAbs) and carbohydrate binding modules (CBMs) with specificities for defined glycan structures (epitopes) occurring on plant cell wall polysaccharides (**[Table S1](#pone.0017506.s001){ref-type="supplementary-material"}**). CoMPP does not provide information about the absolute levels of polysaccharides, but compared to conventional techniques for cell wall analysis such as monosaccharide composition analysis, CoMPP has the advantage that it provides information about glycan epitopes that can be assigned with confidence to specific polysaccharides. The technique is well suited for tracking detailed changes in cell wall components in complex biological systems and particularly so when a paucity of prior information about cell wall composition complicates the interpretation of data from conventional biochemical techniques [@pone.0017506-AlonsoSimn1], [@pone.0017506-Singh1], [@pone.0017506-Srensen2]. We therefore used CoMPP to map the distribution of polysaccharide epitopes within consecutive sections of the fungus garden of *Acromyrmex echinatior* leaf-cutting ants, as well as in the leaves provided as forage and the debris discarded by the ants ([**Figure 1**](#pone-0017506-g001){ref-type="fig"}). Our findings reveal that fungus gardens only partially degrade the plant material provided by the ants and leave the cellulose-rich components largely untouched. ::: {#pone-0017506-g001 .fig} 10.1371/journal.pone.0017506.g001 Figure 1 ::: {.caption} ###### Analysis of plant degradation in leaf-cutting ant fungus gardens. Comprehensive Microarray Polymer Profiling (CoMPP) (A--G) and enzymatic assays (H--J) were used to assess plant degradation in *Acromyrmex echinatior* fungus gardens. (A) Leaf-cutting ants collect and transport fragments of fresh leaves back to the fungus garden where they are further fragmented and deposited in the top layer. The gradual degradation of cell wall polysaccharides, as the fungus garden grows upwards into the new substrate material, results in the plant material moving downwards as it is degraded because debris consisting of old fungus and exhausted substrate material is removed from the bottom of the fungus garden and discarded by the ants. The main steps in our CoMPP technique were: (A) collection of replicate material from leaves, top, middle and bottom layers of fungus gardens, and debris; (B) sample preparation by homogenization and precipitation of cell wall polymers; (C) sequential extraction of cell wall components with CDTA, NaOH and cadoxen; (D) printing of polysaccharides as microarrays using a robot, three concentrations, and four replicates (E); (F) probing of microarrays with monoclonal antibodies (mAbs) or carbohydrate binding modules (CBMs); (G) spot quantification and analysis. The activity of enzymes in the corresponding samples was approximated with azurine dyed and cross-linked (AZCL) polysaccharides substrates: (H) Protein extraction in tris buffer, (I) Substrate incubation, and (K) quantification of the area of blue halo (see **Text S1** for detailed methods). :::  ::: Materials and Methods {#s2} ===================== Four queenright colonies of the leaf-cutting ant *A. echinatior* were excavated in Gamboa, Panama, in May 2003 (\#Ae220) and May 2007 (\#Ae332, \#Ae334, \#Ae356). The colonies were transported to the University of Copenhagen, Denmark, where they were maintained in a climate controlled room at 25°C, 70% RH, on a diet of bramble leaves (*Rubus* spec), dry rice and pieces of fruit. Three months prior to sampling, the colonies were supplied with bramble-leaves only, to make sure that fungus gardens were exclusively handling fresh leaf material by the time of sampling. Bramble leaves are highly suitable as forage as they allow *Acromyrmex* lab colonies to grow to their natural size and produce winged sexuals periodically. Five different categories of samples were collected from the four colonies in September 2008 ([**Figure 1A**](#pone-0017506-g001){ref-type="fig"}): (1) Fresh bramble leaves, (2) The newly established top section of the fungus garden, (3) The middle section of the fungus garden containing gongylidia, (4) The bottom section of the fungus garden (see also description in [@pone.0017506-Schitt1]), and (5) The debris pile consisting of waste material removed from the fungus garden by the ants. Samples in each category were collected in eight replicate (ca. 5 g fresh weight each), pooled, and stored at -20°C until further use. Comprehensive microarray polymer profiling (CoMPP) {#s2a} -------------------------------------------------- An overview of the procedure used for CoMPP analysis is shown in [**Figure** **1B--G**](#pone-0017506-g001){ref-type="fig"}. Briefly, samples were homogenised and cell wall polysaccharides sequentially extracted from 10 mg dry weight of each sample as previously described [@pone.0017506-Moller1] ([**Figure 1B,C**](#pone-0017506-g001){ref-type="fig"}). Supernatants containing extracted cell wall polymers were spotted as microarrays onto nitrocellulose membrane (Schleicher and Schuell, Dassel, Germany) ([**Figure 1D**](#pone-0017506-g001){ref-type="fig"}) and each sample was represented on arrays in three concentrations (the original extraction plus two five-fold dilutions) and as four printing replicates (a total of twelve spots per sample) ([**Figure 1E**](#pone-0017506-g001){ref-type="fig"}). Printing was carried out using a microarray robot (Microgrid II, Genomic Solutions, Cambridge, UK) equipped with split pins (Microspot 2500, Genomic solutions, Cambridge, UK). The arrays were probed with mAbs and CBMs and developed as described previously [@pone.0017506-Moller1] ([**Figure 1F,G**](#pone-0017506-g001){ref-type="fig"}). Details of the mAbs or CBMs used to probe the arrays are listed in **[Table S1](#pone.0017506.s001){ref-type="supplementary-material"}.** All probes were obtained from PlantProbes (Leeds, UK) except BS 400-2 that was obtained from BioSupplies (Melbourne, Australia). Spot signals were quantified and analysed using Imagene 6.0 microarray analysis software (Biodiscovery, <http:/www.biodiscovery.com>) as previously described [@pone.0017506-Moller1] ([**Figure 2A**](#pone-0017506-g002){ref-type="fig"}). The variation in degradation patterns across colonies was analyzed with a general linear mixed model after correcting mean spot signal data for differences in fungal biomass ([**Figure 2B**](#pone-0017506-g002){ref-type="fig"}). Heatmaps ([**Figure 2C**](#pone-0017506-g002){ref-type="fig"}) were constructed from the mean spot signal data using online heatmapper software (<http://bbc.botany.utoronto.ca/ntools/cgi-bin/ntools>) (see **Text S1**). ::: {#pone-0017506-g002 .fig} 10.1371/journal.pone.0017506.g002 Figure 2 ::: {.caption} ###### Comprehensive microarray polymer profiling (CoMPP) of *Acromyrmex echinatior* fungus gardens. \(A) An example of a CoMPP microarray populated with material from two colonies and probed with the anti-xyloglucan mAb LM15. (B) Change in average fungal biomass during the processing of leaf material in *A. echinatior* fungus gardens (±SE). Fungal biomass was determined by measuring the amount of chitin present in fungal cell walls from five different sections: leaves, top, middle and bottom of the fungus garden, and debris pile from the four replicate colonies. (C) The distribution of cell-wall polysaccharides in fungus gardens represented as a heatmap where mean CoMPP spot signals (numbered values, corrected for changes in fungal biomass) are correlated to colour intensity. Sample locations and extraction conditions are shown on the left, polysaccharide epitopes and corresponding monoclonal antibodies mAbs and carbohydrate binding modules (CBM), are shown on the top. The highest corrected mean signal value in the data set was set to 100 and all other values adjusted accordingly. All data are averages of four experiments. :::  ::: Fungal biomass and enzyme activity assays {#s2b} ----------------------------------------- Fungal biomass in each of the five sample locations (leaves, top, middle, bottom of the fungus garden, and debris) was corrected between samples by measuring N-acetylglucosamine (GlcNAC) the principal monosaccharide constituent of the polymer chitin, and a major component of fungal but not plant cell walls. The procedures used were slightly modified from previous studies [@pone.0017506-Matcham1], [@pone.0017506-Francois1] and are described in detail in the **Text S1**. AZCL-polysaccharide colorimetric gel-based assays (Megazyme, Bray, Ireland) were used to determine enzyme activity in the fungus garden as previously described [@pone.0017506-Schitt1], [@pone.0017506-DeFineLicht2]. We applied our AZCL analysis to three sample locations within the fungus garden (top, middle, bottom) that were used in the the CoMPP analysis. See **Text S1** for details. Results {#s3} ======= Detailed mapping shows that individual plant cell wall polymers are differentially degraded {#s3a} ------------------------------------------------------------------------------------------- Our results provide detailed insight into the dynamic turnover of polysaccharide epitopes across the five sampling locations. The three major classes of cell wall polysaccharides, pectins, hemicelluloses and cellulose are held together in walls with increasing degrees of firmness by different chemical bonds and supra-molecular associations. These were sequentially extracted from the fungus garden using three solvents CDTA, NaOH and cadoxen, respectively, so that our results provide information not just about the relative abundance of polymers *per se*, but also about the disassembly of higher order cell wall architectures. A heatmap showing the mean CoMPP signals obtained for all the mAbs and CBMs is shown in [**Figure 2C**](#pone-0017506-g002){ref-type="fig"}. Combining these data across the three extractions provided an overview of the changes in cell wall epitope levels as plant material is processed through the fungus garden ([**Figure 3**](#pone-0017506-g003){ref-type="fig"}) and of the overall fold changes in polysaccharides relative to levels in the starting leaf material (**[Figure S1](#pone.0017506.s004){ref-type="supplementary-material"}**). The average levels of fungal cell wall biomass increased relative to leaf weight by a factor of 1.14, 1.31, 1.5 and 1.6 in the top, middle, bottom and debris samples, respectively ([**Figure 2B**](#pone-0017506-g002){ref-type="fig"}). This regression, which reflects the progressive conversion of plant biomass into fungal biomass, was used to normalise the mean signal values from the CoMPP arrays. ANOVA revealed no significant variation in polysaccharide abundance between the four fungus gardens in our study, confirming that our experimental design produced consistently repeatable results (**[Table S2](#pone.0017506.s002){ref-type="supplementary-material"}**). ::: {#pone-0017506-g003 .fig} 10.1371/journal.pone.0017506.g003 Figure 3 ::: {.caption} ###### Overall changes in polysaccharide occurrence over five sampling locations. The relative amount of pectins, hemicelluloses and cellulose across five different sampling locations (leaves, top, middle, and bottom layers of the fungus garden, and debris) were determined by CoMPP analysis. Data are sums that combine the mean signals for all three extractions (CDTA, NaOH and Cadoxen), averaged across the four colonies and scaled relative to 100 for each polysaccharide epitope corresponding to a specific mAb or CBM. Error bars represent standard errors (±SE). :::  ::: Homogalacturonan (HG) backbone domains of pectin (recognised by mAbs JIM5 and JIM7) were substantially degraded in, or even prior to incorporation in, the top layer of the fungus garden ([**Figure 2C**](#pone-0017506-g002){ref-type="fig"} and [**Figure 3**](#pone-0017506-g003){ref-type="fig"}). Of all the polysaccharide epitopes analysed in this study, HG was the substrate compound most significantly degraded in the fungus garden system (13 and 6.3 fold, respectively, for JIM5 and JIM7, **[Figure S1](#pone.0017506.s004){ref-type="supplementary-material"}**). Material extracted from leaves provided relatively high signals for both these anti-HG mAbs, but the same HG epitopes were essentially absent from all other sampling locations ([**Figure 2C**](#pone-0017506-g002){ref-type="fig"} and [**Figure 3**](#pone-0017506-g003){ref-type="fig"}). Both mAbs bind to partially methyl-esterified HG, but JIM5 binds preferentially to HG with a relatively low degree of methyl-esterification (DE) whilst JIM7 binds preferentially to HG with a relatively high DE. The decrease in JIM7 binding may have resulted partly from a reduction in methyl-ester groups on pectin polymers. This could have occurred as a result of endogenous pectin methyl-esterase activity in the leaves (possibly triggered by ant cutting), or by exposure to exogenous pectin methyl-esterases produced by microorganisms that the leaf material came into contact with during transport or storage, prior to incorporation into the fungus garden. However, a reduction in DE would not cause a reduction in JIM5 binding so our data indicate that cleavage of the HG backbone in pectin molecules also occurred to a significant extent. The (1→4)-β-D-galactan and (1→5)-α-L-arabinan side chains of pectin (recognised by mAbs LM6 and LM5, respectively) were detected only in the leaf material of CDTA-extracted samples, but signals persisted in NaOH and cadoxen extractions from other samples ([**Figure 2C**](#pone-0017506-g002){ref-type="fig"} and [**Figure** **3**](#pone-0017506-g003){ref-type="fig"}). Some HG epitopes were also present in cadoxen extracted samples. This indicates that some pectic domains were associated with hemicelluloses or cellulose, rendering them more recalcitrant to degradation [@pone.0017506-AlonsoSimn1]. Xyloglucan (XyG) (recognised by mAb LM15) and xylans (recognised by mAbs LM10, LM11 and CBM22) showed complex profiles of occurrence throughout the fungus garden system ([**Figure 2C**](#pone-0017506-g002){ref-type="fig"} and [**Figure 3**](#pone-0017506-g003){ref-type="fig"}). The relative level of the LM15 XyG epitope decreased sharply between leaves and samples from the top layer, increased slightly between top and middle layers and then decreased between middle layer and debris samples ([**Figure 3**](#pone-0017506-g003){ref-type="fig"}). Out of all hemicellulosic polysaccharides measured in this study, the LM15 XyG epitope decreased the most overall (3.8 fold) compared to levels in the starting leaf material (**[Figure S1](#pone.0017506.s004){ref-type="supplementary-material"}**). Changes in the overall relative levels of all three xylan-binding probes (LM10, LM11 and CBM22) were similar, with probe signals decreasing between leaves and samples from the top section, increasing between top and bottom sections (sharply in the case of LM10 and LM11), and then decreasing between bottom layer and debris samples ([**Figure 3**](#pone-0017506-g003){ref-type="fig"}). However, subtly different profiles were evident when the levels of xylan epitopes were compared across the different extraction treatments ([**Figure 2C**](#pone-0017506-g002){ref-type="fig"}). For both LM10 and LM11 the highest mean signals were obtained in bottom layer samples, but in contrast to LM10, LM11 epitope levels were also high in debris samples extracted with cadoxen. Levels of CBM22 epitope were highest in leaf material for both NaOH and cadoxen-extracted samples. For NaOH-extracted samples, levels then decreased across the other layers and the epitope was not detected in the debris. In contrast, CBM22 binding after cadoxen extraction persisted in debris samples ([**Figure 2C**](#pone-0017506-g002){ref-type="fig"}). Both LM10 and LM11 recognise unsubstituted (1→4)-β-D-xylans that are associated with secondary cell walls in dicotyledons, while LM11 also binds to more substituted xylans, such as arabinoxylan [@pone.0017506-McCartney1]. CBM22 also recognizes xylans with different degrees of substitution but has a distinctly different binding profile to LM11 on plant materials [@pone.0017506-McCartney2]. Our data therefore show that the degree to which xylan polymers are decorated with other sugars significantly affects their processing with the fungal garden. The cellulosic epitopes recognised by CBM3a and CBM4-1 increased in relative abundance across the five sampling locations ([**Figure 2C**](#pone-0017506-g002){ref-type="fig"} and [**Figure 3**](#pone-0017506-g003){ref-type="fig"}). The levels of CBM3a and CBM4-1 epitopes were higher (2.1 and 7 fold, respectively) compared to the starting leaf material (**[Figure S1](#pone.0017506.s004){ref-type="supplementary-material"}**). This indicates that these epitopes were not significantly degraded and their relative abundance increased to commensurate with decreasing levels of other cell wall polymers. Whereas CBM3a is a type A CBM that binds to crystalline cellulose, CBM4-1 is a type B CBM that binds to internal amorphous regions of cellulose microfibrils [@pone.0017506-Blake1]. Interestingly, whilst the level of CBM3a binding increased steadily throughout the fungal garden system, CBM4-1 binding was restricted mostly to cadoxen-extracted debris material. Cadoxen treatment is known to affect cellulose crystallinity and is thus likely to render it more amenable to CBM4-1 binding. However, the high level of CBM4-1 binding to debris is significant because it may reflect that the gradual removal of other cell wall polymers, particularly XyG, some of which coats cellulose microfibrils, alters the higher order structure of cellulose microfibrils [@pone.0017506-McCartney2]. Enzyme activity often correlates with polysaccharide occurrence {#s3b} --------------------------------------------------------------- Seven different azurine dyed and cross-linked (AZCL) polysaccharide substrates were used to assess the activities of cell wall degrading enzymes in top, middle, and bottom sections of fungus gardens (see **[Table S3](#pone.0017506.s003){ref-type="supplementary-material"}** for details). Enzyme activities were detected for the degradation of all seven polysaccharide substrates and in most cases, enzyme activities were correlated with the levels of cell wall epitopes, as monitored by CoMPP analysis (**[Figure S2](#pone.0017506.s005){ref-type="supplementary-material"}**). Exceptions to this were galactanase activity which was similar in all three zones, whilst galactan decreased from top to middle and remained at this level in the bottom section (**[Figure S2A](#pone.0017506.s005){ref-type="supplementary-material"}**), and xyloglucanase activity, which decreased from top to the middle section, whilst xyloglucan abundance was highest in the middle section of the fungus garden (**[Figure S2C](#pone.0017506.s005){ref-type="supplementary-material"}**). This indicates that enzyme activity generally reflects substrate availability and the positive rather than negative correlation between substrate and enzyme activities suggests an excess of substrate. However, the fact that galactan levels decreased whilst galactanase activity remained unchanged may suggest that galactan was not in excess and that a substantial proportion of this polymer was degraded. Although these findings provide useful insights into the mechanism of polysaccharide degradation in fungal gardens it should be noted that enzyme and CoMPP data cannot be precisely integrated because of the occurrence of multiple isoforms of many cell wall components. For example, variant XyGs exist with many different side chain configurations that can affect both antibody binding and enzyme activity. Discussion {#s4} ========== Pectin degradation starts almost immediately {#s4a} -------------------------------------------- One of our most striking findings was that pectin, particularly homogalacturonan, was extensively degraded in the time between when leaves are collected by the ants and incorporated into the top of the fungus garden, or in the top section itself. Leaf cutting ants first chew plant material into a pulp-like mass that is deposited in the fungus garden, after which the ants place droplets of fecal fluid containing polysaccharide degrading enzymes directly on top of the new substrate [@pone.0017506-Rnhede1]. These enzymes include many pectinases (lyases and esterases), proteases and cellulases, which have been shown to originate from the fungal symbiont itself, passing unharmed through the ant gut and remaining active after deposition [@pone.0017506-Rnhede1]. This mechanism potentially achieves a very rapid degradation of cell wall polymers and likely explains the almost complete absence of certain pectic polymers in the top section of the fungus garden. Our results are thus consistent with previous studies indicating that fecal fluid manuring is a key characteristic of this symbiosis because it accelerates access of the fungal hyphae to primary resources inside the plant cells [@pone.0017506-Schitt1], [@pone.0017506-Erthal1], [@pone.0017506-DeFineLicht2], [@pone.0017506-Silva2]. Recent studies using combinations of cell wall enzymes and antibodies have revealed that hemicelluloses are extensively masked by pectins, rendering them inaccessible to antibodies and therefore most likely enzymes as well [@pone.0017506-Marcus1]. These findings are relevant to the present study because they indicate that the extensive degradation of the pectic network that occurs prior to or just after deposition of leaf material in the top zone of fungal gardens is an essential pre-treatment for the effective utilisation of hemicellulases in the main body of fungal gardens. Pectin is a major structural component of plant cell walls, forming a gel-like matrix that is particularly abundant at cell wall interfaces in the middle lamella region of leaves, where it regulates intercellular adhesion [@pone.0017506-Willats1] ([**Figure 4**](#pone-0017506-g004){ref-type="fig"}). There is also abundant evidence that dissociation of the pectin-matrix induces changes in physiological properties of plant cells that makes them more susceptible to microbial attack [@pone.0017506-EsquerreTugaye1], [@pone.0017506-Murdoch1]. Taken together this strongly suggests that the main function of pectin degradation is to give the fungal hyphae access to superior resources such as proteins and starch inside the plant cells, rather than pectin being an important nutrient source in its own right [@pone.0017506-DeFineLicht2], [@pone.0017506-GomesdeSiqueira1], [@pone.0017506-Silva1]. ::: {#pone-0017506-g004 .fig} 10.1371/journal.pone.0017506.g004 Figure 4 ::: {.caption} ###### A simplified model of polysaccharide degradation in *Acromyrmex echinatior* fungus gardens. The diagram is a synthesis of the CoMPP results obtained in our study. It illustrates three major classes of polysaccharides (pectins, hemicelluloses and cellulose) that are typically present in leaves and the fate of individual polysaccharides when they are processed by the fungus gardens. :::  ::: Fungus gardens do not prioritize degradation of recalcitrant polysaccharides {#s4b} ---------------------------------------------------------------------------- While our analysis confirmed that protein and starch are likely to be the prime targets of leaf decomposition in leaf-cutting ant fungus gardens [@pone.0017506-DeFineLicht2], it also provided compelling evidence for cellulose and xylan not being decomposition priorities. In the fungal gardens proper there was a gradual relative reduction of xyloglucan and unsubstituted xylan epitopes from the top to bottom zones and some further degradation of pectin side chains. As shown in the general overview [**Figure 4**](#pone-0017506-g004){ref-type="fig"}, the discarded debris consisted largely of cellulose and xylan, indicating that the relative levels of these polymers necessarily increased as a proportion of the total biomass. The increase in binding of CBM4-1, which preferentially targets amorphous cellulose [@pone.0017506-Blake1], probably indicates that as other cell wall polymers are progressively removed, cellulose microfibrils become increasingly exposed and susceptible to extraction by cadoxen. Our results are in broad agreement with previous studies based on the measurement of enzyme activity in fungus gardens [@pone.0017506-Schitt1], [@pone.0017506-DEttorre1]-[@pone.0017506-Erthal1], [@pone.0017506-Silva2]. In addition, our study provides a more detailed account of the degradation dynamics in leaf-cutting ant fungus gardens via direct assessment of the abundance of plant biomass components throughout the system. The use of CBMs 3a and 4-1 that recognise distinct structural forms of cellulose provided direct and compelling evidence that cellulose is not utilised as a substrate in *Acromyrmex echinatior* fungus gardens to any significant degree. In contrast, a recent study provides evidence that cellulose is significantly degraded in field-collected fungus gardens of the leaf-cutting ant *Atta colombica* possibly by γ-proteobacteria isolated from the microbiomes of fungus gardens [@pone.0017506-Suen1]. However these results may not be directly comparable because the authors measure absolute cellulose amounts without correcting for differences in fungal biomass, in contrast to relative abundance in our study. It is also puzzling that relatively small decreases in pectin or hemicellulose polymers were observed [@pone.0017506-Suen1], since the presence of these polymers would likely limit the accessibility of cellulose degrading enzymes to their substrate [@pone.0017506-Marcus1]. Therefore these results for *Atta colombica* do not provide evidence that the fungal symbiont can degrade cellulose to any significant degree, rather that the microbial community of fungus gardens has this ability. This apparent division of labour in cellulose-decomposing ability between the fungal symbiont and other garden microbes may resolve the paradox of some earlier studies, which either find evidence of some cellulose decomposition in fungus gardens [@pone.0017506-Bacci1], [@pone.0017506-Suen1], [@pone.0017506-Martin1], or find this activity is insignificant [@pone.0017506-GomesdeSiqueira1], [@pone.0017506-Abril1], [@pone.0017506-Silva1], [@pone.0017506-Silva2]. The actual outcome could well be reinforced by active interference by the farming ants that fastidiously manage the relocation of waste and exhausted fungus material from the garden. Thus being an important structuring force of the microbial communities in their gardens [@pone.0017506-Scott1]. As our data indicate, the question of ability is largely irrelevant when it is not prioritized by the ants because they discard older fungus garden material before any significant amount of cellulose or xylan is decomposed. Based on the diversity of expressed enzyme activities that we measured, it is likely that the fungus gardens retained a high degree of plasticity to cater for many different types of resources when they are made available, as demonstrated by many saprotrophic fungi that facultatively produce extracellular carbohydrate degrading enzymes [@pone.0017506-Deacon1]. As we show here, ample provisioning with fresh leaves allows the ants to prioritize resources that are of higher value and easier to process, rather than cellulose and xylan. However, we cannot exclude that cellulases may have a more significant role under sub-optimal foraging conditions, as for example in dry seasons, when fresh leaf material is less readily available. Comparative fungus garden perspectives: Ecological footprints and evolutionary opportunities {#s4c} -------------------------------------------------------------------------------------------- The approach developed in this study offers a versatile toolkit for studying evolutionary relationships between the degradation capacities of fungal symbionts of different genera of fungus-growing ants. There are more than 230 extant species of fungus-growing ants divided into 12 genera [@pone.0017506-Schultz1]. Only a single monophyletic group of fungus-growing ants, comprised of the genera *Atta* and *Acromyrmex,* has evolved actual fresh-leaf-cutting behaviour [@pone.0017506-Hlldobler1], [@pone.0017506-Schultz1], [@pone.0017506-Hlldobler2]. The majority of attine ants are less advanced and have much smaller colonies and fungus gardens the size of a tennis or table-tennis ball, which they provide with dead plant material and leaf litter [@pone.0017506-DeFineLicht1], [@pone.0017506-Schultz1]. This suggests that the fungus gardens of these ant genera may utilize different suits of extracellular enzymes to facilitate the degradation and metabolism of plant cell wall polysaccharides [@pone.0017506-DeFineLicht2]. This could imply that the profile of carbohydrate active enzymes in these gardens resembles what we here report for the lower sections of *A*. *echinatior* fungus. The clearly visible stratification of leaf-cutting ant fungus gardens is normally not observed in gardens of more basal attine ants, which may imply that this is a derived trait associated with the fairly recent [@pone.0017506-DeFineLicht1] acquisition of a herbivorous niche with richer and more abundant resources. However, the advance into this foraging niche has apparently also made the utilization of substrate relatively wasteful, as residues appear to be discarded as soon as the pectins, starches and proteins typical for fresh leaves have been degraded. This hypothesis was recently formulated based on comparative AZCL enzyme activity data across fungus gardens of eight genera of attine ants [@pone.0017506-DeFineLicht2], and appears consistent with the results provided here. The fact that not all plant biomass is utilized, implies that much more plant substrate is needed to sustain leaf-cutting ant colonies than would be the case if all plant cell wall components were used [@pone.0017506-Abril1], [@pone.0017506-Abril2]. This may explain the large detrimental effect of leaf-cutting ant herbivory in (semi)natural and agricultural ecosystems [@pone.0017506-Wirth1], the elaborate and specialized waste management behaviours seen in *Atta* leaf-cutting ants [@pone.0017506-Bot1], [@pone.0017506-Hart1], and the large waste dumps that these colonies maintain. If wastefulness was an inevitable byproduct of acquiring the novel herbivorous niche, cellulose-rich dumps must have been an integral part of leaf-cutting ant fungus-farming since the *Atta* and *Acromyrmex* clade arose about ten million years ago [@pone.0017506-Schultz1]. This relatively long time span is consistent with distinct communities of microorganisms having evolved in these dumps [@pone.0017506-Scott1], together with highly diverse communities of invertebrate commensals that directly or indirectly exploit the resources that the ants discard [@pone.0017506-Hlldobler1], [@pone.0017506-Hlldobler2], and with leaf-cutting ant colonies having major effects on local nutrient cycling [@pone.0017506-Moutinho1], [@pone.0017506-Sternberg1]. Supporting Information {#s5} ====================== Table S1 ::: {.caption} ###### **The binding specificities of monoclonal antibodies (mAb) and carbohydrate binding molecules (CBM) probes used in this study.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S2 ::: {.caption} ###### **Analysis of colony-level variation with a general linear mixed model.** The effect of colony-level variation on the distribution of cell wall polysaccharides was analysed for each mAb and CBM with a general linear mixed model (see text S1 for details). P-values are not corrected for multiple testing as this would render the test\'s too-conservative when colony variation is expected not to influence the results (Bonferroni correction increase the significance level to α  = 0.005). These results indicate that there was no significant variation in the occurrence of cell wall polysaccharides across the fungus gardens of the four colonies used in our study. (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S3 ::: {.caption} ###### **Substrates used to assess the activity of cell wall degrading enzymes in fungus gardens.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S1 ::: {.caption} ###### **Fold changes (decreases or increases) in polysaccharide occurrence.** Numerical values represent the degree of change in polysaccharide occurrence between the leaf material input to the fungus garden and the debris output from the fungus garden. Individual mAbs and CBMs and their corresponding polysaccharide epitopes are listed in the figure. \* indicate that fold-changes were significantly different from 1 (ANOVA, p \< 0.05). NA indicates that statistical analysis could not be performed on CBM4-1 as this epitope was only available for analysis from two colonies. Error bars represent standard error (±SE). (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### **Enzyme activity in fungus gardens of** ***Acromyrmex echinatior*** **.** Bar graphs (A-H, left y-axis) indicate enzyme activity as halo area (cm^2^) in the three different sampling locations of the fungus garden as determined by AZCL-polysaccharide plate assays. Line graphs (A-H, right y-axis) indicate relative abundances of the corresponding polysaccharides as determined by CoMPP analysis (the same data as in [Figure 3](#pone-0017506-g003){ref-type="fig"} but not adjusted to become a fraction of 100). Enzyme substrates and the corresponding mAbs or CBMs used in the analysis are indicated for each graph (see also [Table S1](#pone.0017506.s001){ref-type="supplementary-material"}). Enzyme data for (1-4)-β-D-xylan are shown twice (D and F) as both mAb LM10 and CBM22 detect this substrate. Error bars represent standard errors across four colonies measured. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: We would like to thank Jelle van Zweden for helping to initiate the collaborative work towards this manuscript. We also thank the Smithsonian Tropical Research Institute, Panama, for providing logistic help and facilities to work in Gamboa, and the Autoridad Nacional del Ambiente y el Mar (ANAM) for permission to sample ants in Panama and export them to Denmark. We thank Professor Paul Knox, University of Leeds, UK, for supplying monoclonal antibodies and Professor Harry Gilbert at the Complex Carbohydrate Resource Centre, Georgia, US, for supplying CBMs. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**The authors IEM, HHDFL, and JJB were supported by the Danish National Research Foundation. JH was supported by the Villum Kann Rasmussen Foundation. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: IEM HHDFL. Performed the experiments: IEM HHDFL JH. Analyzed the data: IEM HHDFL JH JJB. Contributed reagents/materials/analysis tools: IEM HHDFL JH WGTW JJB. Wrote the paper: IEM HHDFL WGTW JJB. [^2]: ¤a Current address: Plant Cell Biology Research Centre, School of Botany, University of Melbourne, Victoria, Australia [^3]: ¤b Current address: Section for Plant Glycobiology, Department of Plant Biology and Biochemistry, Faculty of Life Sciences, University of Copenhagen, Frederiksberg, Denmark [^4]: ¶ These authors also contributed equally to this work.

PubMed Central

2024-06-05T04:04:19.726043

2011-3-10

PMC3053355

Introduction {#s1} ============ The history of Venice, Italy is tightly linked to the ancient plague and particularly to the Second Pandemic, which originated in Europe with the Black Death in the mid-14th century. The commercial activity of the Venetian Republic facilitated trade and interactions with the Southern and Oriental regions of the Mediterranean Sea, where the plague was endemic. Starting in 1348, Venice suffered several plague epidemics, most notably the Black Death [@pone.0016735-Biraben1]. Historical records indicate that a massive epidemic swept through the city during the 14th century [@pone.0016735-Signoli1], which is thought to have killed thousands of people and profoundly affected the history of this prosperous city. Following the initial wave, additional and more detrimental epidemics occurred in 1462, 1485, 1506, 1575--1577 and 1630--1632. In Venice, the number of deaths was first recorded during the 1575--1577 epidemic, with a mortality rate of 27.8%; the 1630--1632 epidemic had a mortality rate of 32.5% of the Venetian population [@pone.0016735-Biraben1], [@pone.0016735-Ell1]. The cause of these disasters is a matter of debate, and it has not been universally agreed upon that these epidemics were due to *Yersinia pestis* [@pone.0016735-Wood1]. Alternative hypotheses including influenza [@pone.0016735-Teh1], anthrax [@pone.0016735-Twigg1] and hemorrhagic fever [@pone.0016735-Duncan1] have been proposed. Using suicide PCR and a recently developed multiplex molecular approach to identify pathogens in ancient human remains [@pone.0016735-Raoult1], we demonstrate here that the Venetian epidemics were indeed plague outbreaks caused by the bacterial species *Y. pestis*. Methods {#s2} ======= Archaeological sites {#s2a} -------------------- During 2004 and 2005, the renovation of the buildings of Lazzaretto Vecchio in Venice revealed several burial sites containing victims of the plague epidemics ([Figure 1](#pone-0016735-g001){ref-type="fig"}). Skeletons from this site were collected by Michel Signoli and Luigi Fozzati. A total of 92 burial locations including graves and trenches were discovered at this site, each containing 5--184 individuals. Pottery fragments found in the sediment were used to determine the age of each site [@pone.0016735-Signoli1], [@pone.0016735-Gambaro1]. Sites 21, 24, 26, 34, 90, 91 and 92 dated to the second half of the 14th century and were organized in regular, narrow, parallel graves approximately 50 cm apart. The graves had an east--west or a west--east orientation and were mainly located in the western part of the Prato al Morti. The corpses were deposited in a supine position on the same level. In sites 26 and 34, the bodies were deposited on ceramic (graffita arcaica) dating to the mid-14th century. Burial sites dating to the 15th century could be divided into two major groups. The first consisted of regular, parallel trenches that intersected and often partially or totally destroyed earlier trenches. This suggested that the locations of the previous burial sites were not recorded. The second group consisted of several levels of large graves. Burials dating to the 16th century were in equally large and long trenches. The burials from the early 17th century epidemic were more dispersed and characterized by regular trenches in an east--west orientation or by rectangular graves with varying numbers of corpses. ::: {#pone-0016735-g001 .fig} 10.1371/journal.pone.0016735.g001 Figure 1 ::: {.caption} ###### Three views of the medieval plague burial sites in Venice, Italy. a: grave 2; b: grave 35; c: grave 44. :::  ::: Prevention of contamination {#s2b} --------------------------- Ancient teeth were collected separately from different skeletons in burial sites by archaeologists and transported to the laboratory in individual bags. The dental pulp, which is protected from external contamination in the central cavity and the root canal of the tooth, was used for molecular experiments [@pone.0016735-Drancourt1]. The teeth used in this study had closed apexes and were free of caries and trauma. All instruments used to collect dental pulp were sterilized for each tooth to prevent cross contamination, and all reagents were from new kits. The laboratory followed general procedures for decontamination including the use of decontamination solutions and sterilization by ultraviolet light before experiments. PCR experiments were performed according to the suicide PCR protocol previously used for *glp*D by our research team [@pone.0016735-Drancourt2]. The experiments were done in a laboratory where *Y. pestis* and *Y. pestis* DNA have not been previously handled. Ancient teeth collected from corpses devoid of any anthropological evidence of infection were collected from a cemetery in Moirans, France (16th--18th) in agreement with French regulations and with appropriate permission of French authorities; they were used as negative controls in the PCR analyses. High throughput detection of pathogens {#s2c} -------------------------------------- Dental pulp was recovered as previously described [@pone.0016735-Drancourt3] and incubated overnight at 56°C with 600 µL of ATL buffer and 50 µL of proteinase K. The total DNA was extracted using the QIAamp Media MDx Kit and pulverized on the BioRobot® MDx workstation in a final volume of 100 µL (Qiagen GmbH, Hilden, Germany). The high throughput detection of seven pathogens was performed as previously described [@pone.0016735-NguyenHieu1]. Briefly, DNA of *Y. pestis*, *Bacillus anthracis* (anthrax agent), *Borrelia recurrentis* (louse-borne relapsing fever agent), *Bartonella quintana* (trench fever agent), *Rickettsia prowazekii* (epidemic typhus agent), *Salmonella enterica* Typhi (typhoid fever agent) and poxvirus (smallpox agent) ([Table](#pone-0016735-t001){ref-type="table"}) was detected with high throughput multiplexed real-time PCR. Two wells containing sterile water and two containing DNA extracted from dental pulp collected from negative control corpses served as standards. ::: {#pone-0016735-t001 .table-wrap} 10.1371/journal.pone.0016735.t001 Table 1 ::: {.caption} ###### Primers and probes for the molecular detection of pathogens in ancient teeth. ::: {#pone-0016735-t001-1} Pathogen Gene Probe and primers PCR product lenght ----------------------------- -------- --------------------------------------------------- -------------------- *Bacillus anthracis* *pag* 6 FAM- TAC CGC AAA TTC AAG AAA CAA CTG C -TAMRA 94 bp 5′- AGG CTC GAA CTG GAG TGA A -3′ 5′- CCG CCT TTC TAC CAG ATT T -3′ *Borrelia recurrentis* 6 FAM- CTG CTG CTC CTT TAA CCA CAG GAG CA -TAMRA 111 bp 5′- TCA ACT GTT TTT CTT ATT GCC ACA -3′ 5′- TCC TTA TGT TGG TTA TGG GAT TGA -3′ *Bartonella quintana* ITS 6 FAM- GCG CGC GCT TGA TAA GCG TG -TAMRA 102 bp 5′- GAT GCC GGG GAA GGT TTT C -3′ 5′- GCC TGG GAG GAC TTG AAC CT -3′ *Rickettsia prowazekii* *ompB* 6 FAM- CGG TGG TGT TAA TGC TGC GTT ACA ACA -TAMRA 134 bp 5′- AAT GCT CTT GCA GCT GGT TCT -3′ 5′- TCG AGT GCT AAT ATT TTT GAA GCA -3′ *Salmonella enterica* Typhi 6 FAM- GCT TTT TGT GAA GCA ACG CTG GCA -TAMRA 138 bp 5′- CTC CAT GCT GCG ACC TCA AA -3′ 5′- TTC ATC CTG GTC CGG TGT CT -3′ Poxvirus HA 6 FAM- AAG ATC ATA CAG TCA CAG ACA CTG T -TAMRA 100 bp 5′- GAC KTC SGG ACC AAT TAC TA -3′ 5′- TTG ATT TAG TAG TGA CAA TTT CA -3′ *Yersinia pestis* *pla* 6 FAM- TCC CGA AAG GAG TGC GGG TAA TAG G -TAMRA 98 bp 5′- ATG GAG CTT ATA CCG GAA AC -3′ 5′- GCG ATA CTG GCC TGC AAG -3′ ::: *Y. pestis* DNA genotyping {#s2d} -------------------------- Further genotyping of *Y. pestis* was based on suicide PCR of the *glp*D gene [@pone.0016735-Drancourt2]. A previously reported *glp*D primer pair [@pone.0016735-Drancourt2] was used and the PCR was conducted in a laboratory in which *Y. pestis* and *Y. pestis* DNA were not previously handled. The PCR products were separated by electrophoresis at 100 V in a 2% agarose gel and sequenced using the Big Dye Terminator Kit. Sequencing products were resolved with the ABI PRISM 3130 Genetic Analyzer (Applied BioSystems, Courtaboeuf, France) and analyzed with the ABI PRISM DNA Sequencing Analysis Software version 3.0 (Applied BioSystems). Sequences were compared with those available in the GenBank database by BLAST (<http://www.ncbi.nlm.nih.gov/blast/Blast.cgi>). Results {#s3} ======= High throughput detection of pathogens {#s3a} -------------------------------------- A total of 173 dental pulp specimens from Venice were analyzed including 37 specimens dating to the 14th century, 45 from the 15th century, 48 from the 16th century and 43 from the 17th century. Negative controls were negative in all experiments. High throughput real-time PCR detected *B. quintana* DNA in five (2.9%) dental pulp specimens, including three from the 16th century and two from the 15th century, and *Y. pestis* DNA was detected in three (1.7%) specimens, including two from the 14th century and one from the 16th century. The other five tested pathogens were not detected in this study. *Y. pestis* DNA genotyping {#s3b} -------------------------- The presence of *Y. pestis* DNA was confirmed by amplifying 165 bp of the *glp*D gene in two specimens, including one specimen positive by real-time PCR (from grave 35) for *Y. pestis* and another specimen negative by real-time PCR. The sequence of the PCR product derived from the specimen of grave 35 was most closely related to that of the *Y. pestis* biotype Orientalis *glp*D gene (GenBank accession number AL59082) with 98% sequence similarity. This sequence is characterized by a 93-bp deletion compared with the *glp*D gene sequence of *Y. pestis* Antiqua (GenBank accession number NC008150). Discussion {#s4} ========== The results reported here are authentic; the negative controls remained negative in the two rounds of PCR-based experiments, and *Y. pestis* was specifically detected using two independent PCR-based experiments including suicide PCR. The specificity of the PCR products was further confirmed by sequencing [@pone.0016735-Drancourt1]. The innovative approach used in this study was based on high throughput, multiplexed detection of seven pathogens that have been implicated in several epidemics with high mortality rates [@pone.0016735-Anderson1]. Previous studies reported the detection of bacteria in the dental pulp of buried individuals [@pone.0016735-Drancourt3], [@pone.0016735-Aboudharam1]. This multiplexed approach allowed the detection of two organisms in individuals recovered from the same grave. *B. quintana* is a blood-borne organism and the etiological agent of trench fever resulting from bacteremia [@pone.0016735-Stein1]. However, asymptomatic bacteremia has also been reported [@pone.0016735-Brouqui1] indicating that only the detection of *B. quintana* DNA in the dental pulp does not definitively identify the cause of death in ancient, buried individuals. However, the same is not true for *Y. pestis*; untreated septicemia always results in death [@pone.0016735-Gage1], [@pone.0016735-Perry1]. Therefore, we interpreted the detection of *Y. pestis* DNA as indicative that these individuals died of septicemic plague. This approach eliminated five pathogens previously implicated without any experimental evidence as being responsible for the Black Death [@pone.0016735-Raoult1]. Only *B. quintana* and *Y. pestis* were detected in these Venetian individuals. *B. quintana* has previously been detected in human remains including a Neolithic individual [@pone.0016735-Drancourt4] and in Napoleon Great Army soldiers from 1815 who also had typhus [@pone.0016735-Raoult2]. We recently detected a *B. quintana* and *Y. pestis* co-infection in individuals excavated from a burial site near Paris dating to the 11th--15th centuries (Drancourt and Le Forestier, unpublished data). *B. quintana* is transmitted by the human body louse *Pediculus humanus* [@pone.0016735-Raoult3], which has been experimentally demonstrated to carry *Y. pestis* [@pone.0016735-Ayyadurai1], [@pone.0016735-Houhamdi1] and was observed during familial plague outbreaks [@pone.0016735-Blanc1]--[@pone.0016735-Blanc3]. Medieval populations are known to have been largely infested by body lice and the observation here of a co-infection with *B. quintana* and *Y. pestis* is compatible with the hypothesis that the body louse was a vector driving the Black Death epidemics in Europe [@pone.0016735-Drancourt5], [@pone.0016735-Drancourt6]. Our results detail the start of the Black Death in Europe in the mid-14th century. Several works previously documented *Y. pestis* in human remains from the Black Death ([Figure 2](#pone-0016735-g002){ref-type="fig"}) including *Y. pestis* DNA in one individual in Vilarnau, France from the 13th--15th centuries [@pone.0016735-Donat1], one individual from the second half of the 14th century in the Saint Come and Saint Damien sites in Montpellier, France [@pone.0016735-Raoult1], three individuals in Dreux, France from the 12th--14th centuries [@pone.0016735-Drancourt7], one individual in Saint-Laurent-de-la-Cabreisse, France from the AD 1348 or 1374 [@pone.0016735-Haensch1], two individuals in Bondy, France from the 11th--15th centuries (Drancourt and Le Forestier, unpublished data), two individuals in Stuttgart, Germany from the 14th--17th centuries [@pone.0016735-Pusch1], five late medieval individuals in Manching-Pichl, Germany [@pone.0016735-Wiechmann1], seven individuals in Bergen op Zoom, the Netherlands from the mid-14th century (AD 1349-50) and two individuals in Hereford, England from the AD 1335±54 [@pone.0016735-Haensch1]. In addition, immunological detection of the F1 antigen has been reported in seven individuals in Saint-Laurent-de-la-Cabreisse, France [@pone.0016735-Haensch1], one individual in Genoa, Italy from the 14th century [@pone.0016735-Cerutti1], ten individuals of Stuttgart, Germany from the 14th--17th centuries [@pone.0016735-Pusch1], three individuals in Bergen op Zoom, the Netherlands and four individuals in Hereford, England [@pone.0016735-Haensch1]. *Y. pestis* has been documented in ten Black Death burial sites scattered over five countries by using different methodological approaches, and therefore the Black Death undoubtedly was due to the plague agent *Y. pestis* [@pone.0016735-Haensch1]. In the present study, ancient *Y. pestis* DNA has been detected in only a small proportion of buried individuals in agreement with previous studies, indicating that detection of aDNA lacked sensitivity, in contrast to the immunological detection of the *Y. pestis* F1 antigen [@pone.0016735-Haensch1], [@pone.0016735-Pusch1], [@pone.0016735-Bianucci1]--[@pone.0016735-Bianucci3]. One Black Death site yielded 10/12 (83.3%) positives in the F1 dipstick assay and only 2/12 (16.7%) positives with PCR techniques [@pone.0016735-Pusch1]. Another recent study yielded only 10/72 (14%) positives with PCR and 24/47 (51%) positives by the F1 dipstick assay [@pone.0016735-Haensch1]. Molecular techniques allowed for genotyping ancient plague and yielded *Y. pestis* Orientalis on the basis of multiple spacer sequencing [@pone.0016735-Drancourt7] and a characteristic deletion in the *glp*D gene as in Venice [@pone.0016735-Drancourt2]. A recent analysis of single nucleotide polymorphisms yielded two previously unknown, non-Orientalis clades of *Y. pestis* in South France, in the Netherlands and in England [@pone.0016735-Haensch1]. In latter study, plague in 17th century Parma, another North Italy city was ascertained by immunological detection of the F1 antigen but aDNA detection failed and genotyping was not done. ::: {#pone-0016735-g002 .fig} 10.1371/journal.pone.0016735.g002 Figure 2 ::: {.caption} ###### Molecular (squares) and immunological (triangles) detection of the plague agent *Yersinia pestis* in ancient burial sites in Europe made by six teams (Marseille team, blue). References are indicated in brackets. **France:** 1. Marseille (18th) [@pone.0016735-Drancourt2], [@pone.0016735-Drancourt3], [@pone.0016735-Bianucci1]; 2. Martigues (18th) [@pone.0016735-Drancourt2], [@pone.0016735-Bianucci1]; 3. Berre l\'Etang (18th) [@pone.0016735-Bianucci2]; 4. La Chaize-le-Vicomte (17th--18th) [@pone.0016735-Bianucci3]; 5. Poitiers (16th--18th) [@pone.0016735-Bianucci3] (Drancourt, unpublished data); 6. Draguignan (17th) [@pone.0016735-Bianucci1] (Drancourt, unpublished data); 7. Saint-Maurice (17th) [@pone.0016735-Hadjouis1]; 8. Briançon (17th) [@pone.0016735-Cerutti1]; 9. Lariey (17th) [@pone.0016735-Bianucci2] (Drancourt, unpublished data); 10. Lambesc (16th) [@pone.0016735-Drancourt3], [@pone.0016735-Bianucci1]; 11. Vilarnau (13th--15th) [@pone.0016735-Donat1]; 12. Bondy (11th--15th) \[Drancourt, unpublished data\]; 13. Montpellier (13th--14th) [@pone.0016735-Raoult1], [@pone.0016735-Drancourt7]; 14. Dreux (12th--14th) [@pone.0016735-Drancourt7]; 15. Vienne (7th--9th) [@pone.0016735-Drancourt2]; 16. Sens (5th--6th) [@pone.0016735-Drancourt7]; 17. Saint-Laurent-de-la-Cabrerisse (AD 1348 or 1374) [@pone.0016735-Haensch1]. **Italy:** 18. Venice (14th--17th) \[present study\]; 19. Genoa (Bastione dell\'Acquasola) (14th) [@pone.0016735-Cerutti1]; 20. Parma (16th/17th) [@pone.0016735-Haensch1]. **Germany:** 21. Stuttgart (14th--17th) [@pone.0016735-Pusch1]; 22. Aschheim (6th) [@pone.0016735-Wiechmann2]; 23. Manching-Pichl (Late medieval) [@pone.0016735-Wiechmann1]; 24. Augsburg (16th/17th) [@pone.0016735-Haensch1]; **The Netherlands:** 25. Bergen op Zoom (Mid-14th) [@pone.0016735-Haensch1]. **England:** 26. Hereford (AD 1335±54) [@pone.0016735-Haensch1]. :::  ::: The originality in the organization of the Lazzaretto Vecchio site is owed to the fact that, unlike other plague burial sites investigated to date; this location was utilized during the Venetian plague waves for four centuries rather than only a single epidemic. This site contains multiple, simultaneous burial sites from different periods of major demographic crises that reflect the unique management of an epidemic. In Venice, the island of Santa Maria di Nazareth appears to have been used since the beginning of the Second Pandemic, if not for the care, at least for the burial of victims. While the Black Death significantly affected Venice, this medieval city imposed the most efficient prevention measures of the time by increasing the 30-day isolation decreed in Ragusa (currently Dubrovnik) to a 40-day isolation known as quarantine [@pone.0016735-Gensini1]. Shortly, all of the port cities in medieval Europe set up quarantine areas that persisted until the 20th century [@pone.0016735-Brachet1]. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**These authors have no support or funding to report. [^1]: Conceived and designed the experiments: DR MD GA. Performed the experiments: TT LF MS. Analyzed the data: MS LF GA DR MD. Contributed reagents/materials/analysis tools: LF DR. Wrote the paper: TT MS MD GA.

PubMed Central

2024-06-05T04:04:19.729056

2011-3-10

PMC3053356

Introduction {#s1} ============ The prevalence of obesity and type 2 diabetes has increased dramatically in the past decades accompanied by an array of secondary health issues [@pone.0016729-Campbell1]. Both obesity and diabetes are features of the metabolic syndrome, which increases ones risk for cardiovascular diseases. Obesity is a key factor contributing to the development of type 2 diabetes; and, weight loss is associated with improvements in glucose homeostasis. Bariatric surgery represents one of the most efficacious methods for the treatment of obesity and type 2 diabetes; and, it often leads to normalization of hyperglycemia and insulin resistance prior to significant weight loss [@pone.0016729-Bult1], [@pone.0016729-Gagliardi1]. Furthermore, bariatric surgery results in sustained weight loss and resolution of type 2 diabetes in a majority of patients with a higher success rate compared to that obtained through diet/lifestyle changes and pharmacotherapy [@pone.0016729-Buchwald1]. The main objective of the present pilot study was to examine changes in whole blood gene expression in obese subjects with type 2 diabetes before and after bariatric surgery, which resulted in weight loss and improved hyperglycemia for all subjects. We chose to observe gene expression changes in whole blood, as this tissue may reflect a systemic response to altered metabolism, and it is the easiest tissue to obtain for serial sampling in a clinical setting. In 11 subjects studied before and after surgery, we found 200 unique genes whose expression was significantly altered, many of which have been previously implicated in obesity and/or type 2 diabetes. Methods {#s2} ======= Study design {#s2a} ------------ The complete clinical characteristics of subjects that participated in this study are given in [Table S1](#pone.0016729.s001){ref-type="supplementary-material"}. The study was conducted in accordance with the appropriate clinical and experimental ethical guidelines and was approved by the Cleveland Clinic Institutional Review Board. Written informed consent was obtained by the subjects before their participation. Eleven obese subjects with type 2 diabetes, (5 females and 6 males), with an average age at enrollment 50.5±11.9 (mean ± SD) were studied. Roux-en-Y gastric bypass surgery (RYGB) was performed on seven subjects, (3 females and 4 males), while 4 subjects underwent Sleeve Gastrectomy (SG). None of the women were taking hormonal therapy, no subjects used steroid medications, and none were smokers. Of the 11 subjects, 9 were taking metformin, 2 were taking pioglitazone, and 4 were taking sulfonylureas as oral diabetes medications prior to surgery. Diabetes medications were discontinued by 6 weeks after surgery. Following a 10--12 hour overnight fast, peripheral blood was drawn for clinical laboratory tests and transcriptome analysis 2 weeks prior to and 6--12 months after intervention. All medications were withheld at least 24 hours prior to blood draw on both occasions. Isolation of total RNA {#s2b} ---------------------- Whole blood samples (2.5 ml) from each subject were collected in PAXgene Blood RNA tubes (BD, Franklin Lakes, NJ) which lyses the cells and stabilizes the RNA. Samples were incubated for two hours at room temperature to ensure the complete lysis of blood cells. Total RNA was isolated using PAXgene Blood RNA Kit (Qiagen, Valencia, CA) following manufacturer\'s guidelines and stored at −80°C. RNA quality was evaluated by incubating 200 ng RNA at 37°C overnight and analyzed by gel electrophoresis. Hybridization and microarray data analysis {#s2c} ------------------------------------------ An aliquot of total RNA for each sample was amplified and biotin labeled using Ovation RNA amplification System V2 (NuGEN, San Carlos, CA). Purified biotin-labeled cDNA was hybridized to HumanWG--6 v2 Expression BeadChip microarrays (Illumina, San Diego, CA) according to the manufacturer\'s instructions. The hybridization temperature was reduced to 48°C to adjust the conditions for hybridization of cDNA rather than cRNA. The arrays were scanned on the Illumina BeadArray Reader using Illumina BeadScan software. Data were quantile normalized without background subtraction using Illumina BeadStudio software. To reduce the chip-to-chip variability the control (pre-surgery) and post-surgery samples for each patient were put on the same microarray chip. A two-tailed paired t-test analysis was performed to identify the genes whose expression levels changed significantly after the surgery. Only the genes with an unadjusted paired t-test p-value\<0.01 were further analyzed and reported. To correct for multiple testing the False Discovery Rate (FDR) Q-value estimates were calculated using QVALUE software (<http://www.genomics.princeton.edu/storeylab/qvalue/index.html>) [@pone.0016729-Storey1], [@pone.0016729-Storey2]. Fold changes in gene expression were calculated as the average of value post/value pre from each of the 11 subjects, and converted to % changes (i.e. 1.40 fold = 40% increase and 0.60 fold = 40% decrease). In order to correlate changes in expression level of each transcript with changes in the clinical values, we calculated, expression levels and clinical values as (value post -- value pre)/value pre, and the mean for all subjects was expressed as percent change. Gene ontology, biological networks, and canonical pathways were evaluated with Ingenuity Pathway Analysis software ([www.ingenuity.com](http://www.ingenuity.com)) and the DAVID Annotation Tool website (<http://david.abcc.ncifcrf.gov>) [@pone.0016729-Huang1], [@pone.0016729-Dennis1]. The microarray dataset from this study is available through the Gene Expression Omnibus server, accession number GSE19790. All microarray data are MIAME compliant. Correlation analysis were performed to identify the significantly regulated genes whose changes in expression levels were best associated with changes in weight loss, BMI, fasting plasma glucose, and WBC count. Real-time quantitative PCR (qPCR) {#s2d} --------------------------------- To validate the microarray findings, the expression levels of cathelicidin antimicrobial peptide (*CAMP*), alpha defensin 1 (*DEFA1*), and carcinoembrionic antigen-related cell adhesion molecule 8 (*CEACAM8*) were quantified relative to the endogenous control gene, beta actin (*ACTB*) using pre-designed TaqMan gene expression assays (Applied Biosystems, Foster City, CA). Mean fold changes for each sample were calculated by using the 2^−ΔΔCt^ method as previously described [@pone.0016729-Livak1]. Protein expression and clinical assays {#s2e} -------------------------------------- Human plasma levels of lipocalin 2 (also known as NGAL) were determined using a rapid ELISA kit (BioPorto Diagnostics, Gentofte, Denmark) following the manufacturer\'s protocol. Serum gamma-glutamyltransferase 1 activity and all other clinical values were determined by standard clinical assays performed by the Cleveland Clinic Pathology and Laboratory Medicine Institute. Gamma-glutamyltransferase activity must be measured in serum, which was only available for six pairs of samples. Results {#s3} ======= Clinical and metabolic effects of bariatric surgery {#s3a} --------------------------------------------------- The diabetes related clinical characteristics of the subjects before and after surgery are presented in [Table 1](#pone-0016729-t001){ref-type="table"} and the complete clinical data for each subject are presented in [Table S1](#pone.0016729.s001){ref-type="supplementary-material"}. Overall, bariatric surgery had a positive impact on weight loss and resolution of hyperglycemia for every subject included in this study, leading to significant decreases of 22% in mean weight (p = 9.1×10^−7^), 21.0% in mean BMI (p = 9.9×10^−7^), 42% in fasting plasma glucose (p = 8.6×10^−4^), 26% in mean glycosylated hemoglobin (HbA~1C~) content (p = 5.9×10^−4^), and 80% in fasting plasma insulin (p = 0.042). Other significant changes include a 12% decrease in white blood cell count (p = 0.016), a 31% decrease in mean VLDL-cholesterol (p = 0.036), a 54% decrease in plasma alanine aminotransferase activity (p = 7.5×10^−4^), a 7% decrease in % neutrophils (p = 0.01), and a 14% increase in percent lymphocytes (p = 0.007). Although this pilot study was not designed or powered to compare the two types of surgery, we observed that the seven subjects who underwent RYGB surgery were more responsive in weight loss and resolution of type 2 diabetes than the four SG surgery subjects, with larger percent decreases in BMI (24% and 18% decreases for RYGB and SG, respectively, p = 0.016), fasting plasma glucose (45% and 28%, respectively, p = 0.086), and HbA~1C~ (30% and16%, respectively, p = 0.064). Post surgery, all seven of the RYGB subjects and only one of four SG subjects met the criterion for normal fasting plasma glucose of \<100 mg/dL (p = 0.007 by chi square test). ::: {#pone-0016729-t001 .table-wrap} 10.1371/journal.pone.0016729.t001 Table 1 ::: {.caption} ###### Diabetes related clinical characteristics of subjects involved in the study. ::: {#pone-0016729-t001-1} Clinical Characteristic Pre-surgery[a](#nt101){ref-type="table-fn"} Post-surgery[a](#nt101){ref-type="table-fn"} \% Change P- value[b](#nt102){ref-type="table-fn"} \% Change RYGB \%Change SG P-value[c](#nt103){ref-type="table-fn"} ----------------------------------------------------------------- --------------------------------------------- ---------------------------------------------- ----------- ------------------------------------------ ---------------- ------------- ----------------------------------------- Weight (kg) 140.7±35.1 110.4±29.9 −21.5 9.08E-07 −24 −18 0.033 BMI (kg/m^2^) 47.1±11.6 37.2±10.0 −21.0 9.90E-07 −24 −17 0.016 Fasting plasma glucose (mg/dL) 170±54.2 97.8±26.5 −42.5 8.57E-04 −45 −28 0.086 Fasting plasma insulin (µIU/ml)[d](#nt104){ref-type="table-fn"} 32.2±35.4 6.6±4.8 −79.5 4.25E-02 −79 −53 0.135 HbA~1c~ (%) 7.9±1.5 5.9±0.6 −26.3 5.87E-04 −30 −16 0.064 a , Data are means ± SD for 11 subjects pre- and post surgery. b , Unadjusted paired T-test comparing pre- vs. post surgery. c , T-test comparing % change by RYBG vs. SG. d , Data only available for 9 subjects before and after surgery. BMI, Body mass index. HbA~1C~, glycosylated hemoglobin A~1C~. ::: Gene expression profiling {#s3b} ------------------------- Of the 48,804 probes on the microarray, 17,115 probe IDs were called present (Illumina detection p-value≤0.05) in 10 or more of the 22 total samples, which were used in subsequent analyses (10 was chosen to reduce the noise while maintaining any potential gender specific expressed transcripts in the five paired female and the six paired male samples). Due to the small number of subjects, we were not powered to compare the effects of the two types of surgery on gene expression, thus all subjects were pooled for our analyses. 204 (∼1.7%) of the transcripts were identified as differentially expressed due to the bariatric surgery at a p-value\<0.01 ([Table S2](#pone.0016729.s002){ref-type="supplementary-material"}). Because some transcripts had more than one probe on the array, these 204 transcripts represent 200 unique genes. To correct for multiple testing, we estimated the false discovery rate (FDR) at p\<0.01 to be 0.52, thus, at the considered cutoff we would expect 107 transcripts to appear differentially expressed solely by chance (compared to the 204 differentially expressed transcripts observed). Of the 204 differentially expressed transcripts, 115 (56%) were down regulated and 89 (44%) were up regulated. The top 25 significantly differentially expressed transcripts ranked by p-value are displayed in [Table 2](#pone-0016729-t002){ref-type="table"}. None of the 204 transcripts had a 2- or greater fold change, and only 17 (8.3%) were altered by more than 25%, with 8 being down regulated and 9 up regulated ([Table 3](#pone-0016729-t003){ref-type="table"}). ::: {#pone-0016729-t002 .table-wrap} 10.1371/journal.pone.0016729.t002 Table 2 ::: {.caption} ###### Top 25 differentially expressed transcripts and their respective p-values and % expression change after bariatric surgery. ::: {#pone-0016729-t002-2} RANK PROBE\_ID GENE Symbol Paired T-Test \% Change DEFINITION ------ --------------- ------------- --------------- ----------- ---------------------------------------------------------------------- 1 ILMN\_1652604 GGT1 6.35E-05 −10% Gamma-glutamyltransferase 1, transcript variant 4. 2 ILMN\_1736238 GNMT 9.49E-05 −7% Glycine N-methyltransferase. 3 ILMN\_1688580 CAMP 1.76E-04 −41% Cathelicidin antimicrobial peptide. 4 ILMN\_1682312 CYBB 2.24E-04 −13% Cytochrome b-245, beta polypeptide (chronic granulomatous disease). 5 ILMN\_1768399 ARFIP1 3.64E-04 −9% ADP-ribosylation factor interacting protein 1, transcript variant 1. 6 ILMN\_1781028 ZMYM5 4.18E-04 20% Zinc finger, MYM-type 5, transcript variant 1. 7 ILMN\_1727740 SYNCRIP 4.59E-04 20% Synaptotagmin binding, cytoplasmic RNA interacting protein. 8 ILMN\_1797682 INSL3 4.81E-04 −13% Insulin-like 3 (Leydig cell). 9 ILMN\_1714987 TRIM54 5.04E-04 −18% Tripartite motif-containing 54, transcript variant 1. 10 ILMN\_1675413 ENPP7 5.06E-04 −11% Ectonucleotide pyrophosphatase/phosphodiesterase 7. 11 ILMN\_1813091 ARL1 5.84E-04 9% ADP-ribosylation factor-like 1. 12 ILMN\_1748384 BOC 6.02E-04 15% Boc homolog (mouse). 13 ILMN\_1759017 ZNF333 6.44E-04 15% Zinc finger protein 333. 14 ILMN\_1912287 HS.133181 7.69E-04 −11% BX093329 Soares\_parathyroid\_tumor\_NbHPA cDNA clone. 15 ILMN\_1775570 FLJ44635 8.33E-04 −9% TPT1-like protein. 16 ILMN\_1656371 TRPA1 9.35E-04 −10% Transient receptor potential cation channel, subfamily A, member 1. 17 ILMN\_1689294 LOC85390 9.46E-04 16% RNA, small nucleolar on chromosome 11. 18 ILMN\_1911981 HS.197709 9.90E-04 22% AGENCOURT\_8291102 Lupski\_sympathetic\_trunk cDNA clone 19 ILMN\_1679357 DEFA1/DEFA3 1.08E-03 −46% Defensin, alpha 1/Defensin alpha 3. 20 ILMN\_1804631 CNBP 1.11E-03 13% CCHC-type zinc finger, nucleic acid binding protein. 21 ILMN\_1725661 DEFA1/DEFA3 1.16E-03 −45% Defensin, alpha 1/Defensin alpha 3. 22 ILMN\_1815777 NBPF12 1.16E-03 −11% PREDICTED: neuroblastoma breakpoint family, member 12. 23 ILMN\_1687567 CUTL1 1.25E-03 −10% Cut-like 1, CCAAT displacement protein, transcript variant 1. 24 ILMN\_1672246 OR4C15 1.30E-03 −11% Olfactory receptor, family 4, subfamily C, member 15. 25 ILMN\_1893704 HS.541159 1.31E-03 14% 7n45b12.x1 NCI\_CGAP\_Lu24 cDNA clone IMAGE:3567335 3. ::: ::: {#pone-0016729-t003 .table-wrap} 10.1371/journal.pone.0016729.t003 Table 3 ::: {.caption} ###### Differentially expressed transcripts with \>25% change in expression following bariatric surgery. ::: {#pone-0016729-t003-3} PROBE\_ID GENE Symbol P-value[a](#nt107){ref-type="table-fn"} \% Change DEFINITION --------------- ------------- ----------------------------------------- ----------- ------------------------------------------------------------ ILMN\_1679357 DEFA1/DEFA3 1.08E-03 −46% Defensin, alpha 1/definsin alpha 3. ILMN\_1692223 LCN2 4.37E-03 −46% Lipocalin 2 (oncogene 24p3). ILMN\_1725661 DEFA1/DEFA3 1.16E-03 −45% Defensin, alpha 1/defensin alpha 3 ILMN\_1693262 DEFA1/DEFA3 1.65E-03 −45% Defensin, alpha 1/defensin alpha 3 ILMN\_1806056 CEACAM8 3.73E-03 −45% Carcinoembryonic antigen-related cell adhesion molecule 8. ILMN\_1688580 CAMP 1.76E-04 −41% Cathelicidin antimicrobial peptide. ILMN\_1723035 OLR1 8.09E-03 −37% Oxidized low density lipoprotein (lectin-like) receptor 1. ILMN\_1762713 C19ORF59 6.53E-03 −29% Chromosome 19 open reading frame 59. ILMN\_1690546 PPP3CC 6.21E-03 26% Protein phosphatase 3 catalytic subunit, gamma isoform. ILMN\_1805271 ZNF721 6.81E-03 26% Zinc finger protein 721. ILMN\_1813400 CBR4 8.96E-03 26% Carbonyl reductase 4. ILMN\_1702858 ADHFE1 3.78E-03 27% Alcohol dehydrogenase, iron containing, 1. ILMN\_1748476 NOP5/NOP58 2.49E-03 28% Nucleolar protein NOP5/NOP58. ILMN\_1661940 CAMTA1 4.40E-03 28% Calmodulin binding transcription activator 1. ILMN\_1656111 MYLIP 3.83E-03 29% Myosin regulatory light chain interacting protein. ILMN\_1797893 PFAAP5 5.93E-03 31% Phosphonoformate immuno-associated protein 5. ILMN\_1679045 SBDS 6.03E-03 39% Shwachman-Bodian-Diamond syndrome. a , Paired T-test pre- and post-surgery. ::: The transcript with the most significant p-value overall encodes for gamma-glutamyltransferase 1 (*GGT1*); and, although its expression was only reduced by 10% post-surgery, this change was consistent in all subjects. The transcript with the third most significant p-value overall encodes for cathelicidin antimicrobial peptide (*CAMP*), which was reduced by 41% post surgery. The microarray contained three probes that align to both alpha defensin 1 (*DEFA1*) and alpha defensin 3 (*DEFA3*), which ranked first, third and fourth of those genes with the largest percent change in gene expression (45 to 46% down regulated, [Table 3](#pone-0016729-t003){ref-type="table"}), demonstrating internal consistency in this finding. Thus, two separate antimicrobial peptides (*CAMP* and *DEFA1/A3*) were relatively highly down regulated. Other transcripts that were relatively highly down regulated include lipocalin 2 (*LCN2*) and carcinoembryonic antigen-related cell adhesion molecule 8 (*CEACAM8*), with 46% and 45% reductions post-surgery, respectively ([Table 3](#pone-0016729-t003){ref-type="table"}). Quantitative real-time PCR (qPCR) {#s3c} --------------------------------- To confirm the microarray data we performed qPCR for three relatively highly regulated and robustly expressed transcripts: *DEFA1*, *CAMP*, and *CEACAM8* ([Table 4](#pone-0016729-t004){ref-type="table"}). Paired t-test p-values for the effect of bariatric surgery on transcript expression levels were found to be consistent with the microarray data; and, the percent changes after surgery were similar or greater than the percent changes detected by microarray ([Table 4](#pone-0016729-t004){ref-type="table"}). The expression levels of these three transcripts were compared between the microarray and qPCR data for the 22 (pre and post-surgery) samples using linear regression analysis; and, each transcript had a very strong positive correlation (R values ranged from 0.75 to 0.88). ::: {#pone-0016729-t004 .table-wrap} 10.1371/journal.pone.0016729.t004 Table 4 ::: {.caption} ###### qPCR validation of microarray results for select transcripts. ::: {#pone-0016729-t004-4} Gene/Transcript Symbol Array vs. qPCR Correlation (R)[a](#nt108){ref-type="table-fn"} qPCR Paired T-TestP-value[b](#nt109){ref-type="table-fn"} Array Paired T-TestP-value[b](#nt109){ref-type="table-fn"} qPCR% Change Array% Change ------------------------ ---------------------------------------------------------------- ----------------------------------------------------------- ------------------------------------------------------------ -------------- --------------- DEFA1 0.88 2.99E-03 1.08E-03 −64% −46% CAMP 0.75 1.01E-03 1.76E-04 −43% −41% CEACAM8 0.86 1.08E-02 3.73E-03 −57% −45% a , P-values of linear regression analysis are all significant (p\<0.05). b , Two-tailed paired t-test based on comparing transcript levels pre- vs. post- surgery. ::: Plasma protein assays {#s3d} --------------------- To determine whether the protein levels follow the same expression pattern as their transcripts we measured the activity or level of two serum proteins, gamma-glutamyltransferase and lipocalin 2. Serum gamma-glutamyltransferase activity was reduced in the post-surgery samples by an average of 38% (p = 0.05, paired t-test), consistent with the direction of change observed for *GGT1* mRNA levels in whole blood. Plasma levels of lipocalin 2 were determined in 8 subjects pre and post surgery, and its levels were significantly increased after the surgery by an average of 34% (p = 0.008), which is opposite to the whole blood *LCN2* mRNA levels that were significantly decreased. This discrepancy may be due to the majority of plasma lipocalin 2 being synthesized in tissues other than blood such as adipocytes and liver [@pone.0016729-Yan1], [@pone.0016729-Esteve1]. Pathway and correlation analyses {#s3e} -------------------------------- To identify the biological mechanisms, pathways and functions most relevant to the genes of interest, the differentially expressed transcripts (p\<0.01) were subjected to Ingenuity Pathway Analysis. The top scoring network identified was "lipid metabolism, small molecule biochemistry and free radical scavenging" ([Figure 1](#pone-0016729-g001){ref-type="fig"}). Of the 23 genes involved in the network, 16 were down regulated and seven were up regulated. Also, the differentially expressed transcripts were subjected to canonical pathways analysis via Ingenuity Pathways Analysis software. The following pathways (together with their respective genes) were the most significantly represented canonical pathways (Fisher\'s exact test p-value\<0.05): myc-mediated apoptosis signaling (FAS↑, SHC1↓ and TP53↓); thyroid cancer signaling (SHC1↓ and TP53↓); fatty acid metabolism (ADHFE1↑, CPT1A↓, CYP51A1↓ and SLC27A6↓); iCOS-iCOSL signaling in T-helper cells (PLEKHA3↑, PPP3CC↑ and SHC1↓); and glycine, serine and threonine metabolism (GNMT↓, SARDH↓ and SMOX↓). ::: {#pone-0016729-g001 .fig} 10.1371/journal.pone.0016729.g001 Figure 1 ::: {.caption} ###### Lipid metabolism, small molecule biochemistry, free radical scavenging network. The straight lines represent direct relationships and the dotted lines represent indirect relationships. Up regulated and down regulated genes that meet the P-value cutoff of \<0.01 are shown in red and green shading respectively, with color intensity related to the fold change in expression. The molecules that do not meet the abovementioned P-value cutoff are shown in gray, while the molecules that are incorporated in the network through relationships with other molecules and are not user specified are shown in white. Symbols used in the figure represent: trapezoid, transporter; circle, other; concentric circles, complex/group; dotted square, growth factor; vertical rhombus, enzyme; horizontal rhombus, peptidase; vertical oval, transmembrane receptor; horizontal oval, transcription regulation. :::  ::: Linear regression analysis was performed in order to identify changes in transcript expression that were best correlated with changes in weight, FPG, HbA~1C~, and WBC count, using an arbitrary R^2^ cutoff of ≥0.25. We found that expression changes in 20 regulated transcripts (9.8% of the 204 identified) were highly correlated with changes in weight ([Table S3](#pone.0016729.s003){ref-type="supplementary-material"}). Of these 20 transcripts, 12 were found to be inversely correlated and 8 were positively correlated with weight change. The transcript best and inversely correlated with changes in weight encodes for prenyl (decaprenyl) diphosphate synthase subunit 1 (*PDSS1*, R = −0.80). Gene ontology (GO) analysis of the 20 transcripts best correlated with weight change revealed two related GO terms that were significantly represented, cholesterol biosynthetic process (p = 1.23E-04) and sterol biosynthetic process (p = 2.21E-04, [Table S4](#pone.0016729.s004){ref-type="supplementary-material"}). Changes in 35 regulated transcripts (17.2% of the 204 identified) were highly correlated with changes in FPG. More transcripts were inversely correlated than positively correlated (23 versus 12). The transcript best correlated with changes in FPG was *HHAT* (R = −0.79), which encodes the hedgehog acyltransferase enzyme. Changes in 32 regulated transcripts (15.7% of the 204 identified) were strongly correlated with changes in HbA~1C~ levels, of which 23 were inversely and 9 transcripts were positively correlated ([Table S3](#pone.0016729.s003){ref-type="supplementary-material"}) The distributions for positive and/or inverse correlations were not significantly different for this or any of the other correlations via chi-square tests. The transcript best and inversely correlated with changes in HbA~1C~ content (R = −0.75) encodes for the WD repeat 35 protein (*WDR35*). We looked for transcripts that were correlated to more than one of these clinical traits, and we found many that were highly correlated with two or more traits, and seven that were highly correlated with all three traits ([Figure 2](#pone-0016729-g002){ref-type="fig"}). These seven transcripts were *WDR35*, *FLJ45244*, *DHCR24*, *TIGD7*, *TOPBP1*, *TSHZ1* and *FAM8A1* ([Table S3](#pone.0016729.s003){ref-type="supplementary-material"}). ::: {#pone-0016729-g002 .fig} 10.1371/journal.pone.0016729.g002 Figure 2 ::: {.caption} ###### Venn diagram representing the number of transcripts whose change in expression after bariatric surgery were highly correlated with percent changes in weight loss, fasting plasma glucose, and HbA~1C~ content. Seven transcripts were highly correlated with changes in all three of these clinical characteristics. :::  ::: Changes in the expression of 27 transcripts (13.24% of the 204 identified) were strongly correlated with changes in WBC count, with 16 inversely and 11 positively correlated. Two of our top three differentially expressed transcripts, *GGT1* and *CAMP*, were highly and positively correlated with changes in WBC count (R = 0.63 and 0.56, respectively, [Table S3](#pone.0016729.s003){ref-type="supplementary-material"}). Discussion {#s4} ========== This pilot study was designed to examine the clinical and biochemical changes induced by bariatric surgery and their association with changes in whole blood gene expression in obese subjects with type 2 diabetes. This is the first study that we are aware of comparing the whole blood transcriptome in subjects before and after bariatric surgery. Our pilot study identified ∼200 transcripts whose expression levels in whole blood were significantly changed after bariatric surgery, some of which were previously reported to be implicated in obesity and/or type 2 diabetes. Although we observed consistent reductions in body weight, BMI, fasting plasma glucose, and HbA~1C~ levels following bariatric surgery, we did not specifically measure changes in body fat composition. Thus, we were not able to determine the degree of correlation between changes in gene expression and changes in body fat composition among the 11 subjects. Overall, we observed more changes in gene expression correlated with changes in the measures of diabetes related traits (fasting plasma glucose and HbA~1C~ levels) than we observed with changes in body weight, thus implying that the resolution of diabetes had a stronger influence on whole blood gene expression than the loss of body weight. We searched the literature and the Gene Expression Omnibus database (<http://www.ncbi.nlm.nih.gov/geo>) for prior transcriptome profiling studies in humans following bariatric surgery or weight loss. One study examined the skeletal muscle transcriptome in three human subjects before and after bariatric surgery. Twenty genes were differentially expressed using a paired t-test, and all of them were reduced 12 months after surgery with decreases ranging from 25 to 66% [@pone.0016729-Park1]. In another study, Ghosh *et al.*, compared whole blood gene expression profiles in lean and obese subjects [@pone.0016729-Ghosh1]; although, the Ghosh study used the same source of RNA as the current study, it did not compare subjects prior to and after weight loss using a paired t-test analysis. However, Ghosh deposited additional whole blood gene expression profiles from 10 obese subjects before and after a 6-week diet period which led to weight loss (GEO accession \# GSE18897). We analyzed this dataset using the same paired t-test as algorithm as in the current study, and found 388 probes that were significantly different using the unadjusted p-value threshold of \<0.01. The expression levels of 194 genes were reduced after weight loss with decreases ranging from 7 to 63%. In the present study, we found that bariatric surgery led to significantly decreased levels of serum alanine aminotransferase (ALT) (54% decrease), plasma VLDL-cholesterol, and serum gamma-glutamyltransferase (GGT) activity. Given previously reported association of elevated serum VLDL, ALT and GGT levels with hepatic steatosis in obese and type 2 diabetes subjects, and the positive effects of bariatric surgery to reverse hepatic steatosis and the levels of these markers [@pone.0016729-Mathurin1], their decreased levels in our study likely reflect the reduction of hepatic fat and improvement in liver function following bariatric surgery. Other studies report that serum levels of GGT are positively and strongly associated with the increased risk of type 2 diabetes over a period of three years [@pone.0016729-Andre1], [@pone.0016729-Andre2], and associated with increased risk for fatal cardiovascular disease (CVD) [@pone.0016729-Strasak1]. We found a decrease in whole blood *GGT1* mRNA levels, the transcript with the most significant p-value in our study. Since the liver is thought to be the major source of serum GGT activity [@pone.0016729-Emdin1], we speculate that the liver and blood levels of *GGT1* mRNA may be coordinately regulated. The human *GGT1* gene has multiple promoters and is alternatively spliced leading to several mRNA isoforms [@pone.0016729-Visvikis1]. Several studies have reported a complex regulation of *GGT1* transcription upon exposure to oxidative stress [@pone.0016729-Zhang1]; however, it not known whether liver and blood express the same isoforms of *GGT1* mRNA and the pathways that regulate GGT1 expression in these tissues are not well defined. In conclusion, our observed decrease in whole blood GGT1 mRNA levels along with decreased FPG levels after bariatric surgery in the current study are consistent with these prior studies [@pone.0016729-Andre1]--[@pone.0016729-Strasak1] that showed increased serum GGT activity in those at most risk for type 2 diabetes. A variety of acute-phase proteins are increased in subjects with obesity and type 2 diabetes [@pone.0016729-FernandezReal1]; while, several acute phase reactant proteins have been shown to decline following bariatric surgery [@pone.0016729-Compher1]. We found that the mRNA levels of two antimicrobial peptides, cathelicidin antimicrobial peptide (*CAMP*) and alpha defensin 1/3 (*DEFA1/DEFA3*), were significantly decreased after bariatric surgery. In two prior studies, plasma levels of alpha defensins were found to be decreased by 55 to 60% in women after bariatric surgery, and their levels were correlated with changes in plasma triglyceride levels and pathology based measures of liver steatosis and inflammation [@pone.0016729-Manco1], [@pone.0016729-Manco2]. Thus, our finding of decreased *DEFA1/DEFA3* mRNA levels after bariatric surgery agrees with prior findings from plasma protein level measurement assays. Lipocalin 2 (*LCN2*) is another gene whose expression was significantly (46%) reduced after surgery. *LCN2* has been classified as an adipokine that increases expression of *PPARγ* and adiponectin in 3T3-L1 adipocytes, and suppresses LPS-induced macrophage cytokine production [@pone.0016729-Zhang2]. *Lcn2* mRNA is increased in the liver and adipose tissue of obese *Lepr^db/db^* vs. wild type mice [@pone.0016729-Wang1]. Also, lipocalin 2 protein levels are higher in obese vs. lean humans [@pone.0016729-Wang1]. Furthermore, plasma lipocalin 2 levels are positively correlated with BMI, adiposity, hyperglycemia, and insulin resistance [@pone.0016729-Wang1]. Lipocalin 2 has also been implicated in innate immunity, as *Lcn2*-deficient mice are more susceptible to bacterial infection [@pone.0016729-Wang1]. Although blood *LCN2* mRNA was decreased after surgery in our study, we found that plasma lipocalin 2 levels increased after surgery, which may be attributed to increased secretion from another tissue source. Although our observed effects of surgery on plasma lipocalin 2 levels are not concordant with the prior correlations, we are not aware of any other studies measuring plasma lipocalin 2 before and after bariatric surgery. Tumor protein p53 (*TP53*) is another transcript that we identified that was down regulated after bariatric surgery. A recent study has reported that adipose tissue p53 plays a crucial role in the regulation of insulin resistance [@pone.0016729-Minamino1]. Increased expression of *p53* was found to be associated with an increased production of proinflammatory cytokines that led to insulin resistance, while decreased expression had the opposite effect on proinflamatory cytokines and was associated with improved insulin resistance in mice with type 2 diabetes-like disease [@pone.0016729-Minamino1]. Thus, our finding of decreased whole blood *TP53* mRNA after bariatric surgery, weight loss, and remission of type 2 diabetes is consistent with this prior study. The lectin-like low density lipoprotein receptor 1 (*OLR1*) gene, also known as *LOX1*, encodes for a protein that can be proteolytically cleaved and released as soluble form in serum. Whole blood *OLR1* mRNA levels were significantly decreased by 37% after bariatric surgery in our study, consistent with a previous study reporting that the serum levels of the sLOX1 protein is higher in type 2 diabetes patients than in controls [@pone.0016729-Tan1]. Human genome wide association studies have identified many SNPs associated with type 2 diabetes. Four genes with common SNPs that were previously identified to be associated with type 2 diabetes [@pone.0016729-Wellcome1]--[@pone.0016729-Rampersaud1] are found in our list of 204 transcripts whose expression was altered after bariatric surgery. Surgery induced changes in expression of all four of these genes were found to be highly correlated with surgery induced changes in HbA~1C~ content. These four genes are: contactin associated protein-like 5 (*CNTNAP5*, decreased after surgery), 24-dehydrocholesterol reductase (*DHCR24*, decreased), hedgehog acyltransferase (*HHAT*, increased), and sarcosine dehydrogenase (*SARDH*, decreased). The finding that two distinct methods, genetic association with type 2 diabetes and transcriptome profiling after bariatric surgery, yield four identical gene hits highlights a potential for gene environmental interactions playing an important role for these specific genes and their influence on glucose homeostasis. *DHCR24* was also one of seven genes whose change in expression level after bariatric surgery was positively and highly correlated with changes in the three examined clinical phenotypes: HbA~1C~ content, weight, and FPG. *DHCR24* encodes an enzyme participating in sterol biosynthesis; however, it also binds to p53 and protects it from degradation leading to cellular defense against oxidative stress [@pone.0016729-Wu1], [@pone.0016729-Otis1]. *DHCR24* is up regulated in endothelial cells by exposure to HDL, and the HDL mediated lowering of *NF-kB* and *VCAM* expression is dependent upon *DHCR24* expression [@pone.0016729-McGrath1]. The decrease in expression of both *DHCR24* and *TP53* in whole blood after surgery reinforces the concept that surgically induced weight loss and remission of type 2 diabetes may alter WBC p53 pathways; however, the role of *DHCR24* in WBC metabolism and inflammation is not known. One of the advantages of our study is the use of whole blood rather than isolated peripheral mononuclear cells to avoid gene expression changes due to variable handling times during processing [@pone.0016729-Whitney1], in addition to capturing changes in more cell types. Adipose tissue would be an attractive tissue to analyze in regard to changes in gene expression following bariatric surgery, since a higher proportion of adipose transcripts than whole blood transcripts were found to be correlated with BMI in a population study [@pone.0016729-Emilsson1]. However, blood is simpler to obtain, particularly for serial sampling; and, genetic cis effects on gene expression are mostly conserved between adipose tissue and whole blood [@pone.0016729-Emilsson1]. Furthermore, white blood cell expression profiling has been shown to be useful for screening of diseases of non-blood tissues, such as colorectal cancer [@pone.0016729-Han1]. In prior studies, whole blood transcriptome profiling was used to compare obese and lean subjects, yielding sets of transcripts that were differentially expressed [@pone.0016729-Ghosh1], [@pone.0016729-Hindle1]. One potential complication of the use of whole blood for transcriptome analysis is that the abundant globin mRNA can interfere with subsequent analyses. There are several ways to counter this issue, and we chose NuGen Ovation RNA amplification kit to eliminate this problem, as it has been shown to be the most efficient method and yields increased sensitivity [@pone.0016729-Burian1]. The present study had several limitations. The small sample size led to a high FDR and prevented the analysis of differential transcriptome responses to the RYGB and SG surgeries. The paired t-test analysis with each subjects\' pre and post surgery whole blood RNA placed on the same Illumina BeadChip controlled for chip to chip variation; but, this design also precluded the use of multiple logistic regression to correct for covariates such as gender and surgery type. Overall, we did not observe large fold effects on transcript levels, but variations of 10% to 60% can have meaningful physiological consequences, particularly in our experimental design that determines changes in whole blood gene expression levels before and after bariatric surgery within each individual subject, thus removing all other genetic and environmental differences among the subjects. The expression of many genes decreased after bariatric surgery, and we presented these reductions as % changes derived from fold changes of post/pre values, which appears to under represent the fold effect. For example, a 50% reduction in gene expression actually represents a −2-fold change. We directly compared the % changes of two transcripts found significantly altered by weight loss in both our study and our analysis of the data from Ghosh *et al.*, (GEO Accession \# GSE18897). We observed that expression of *CEACAM8* decreased by 45% (−1.8-fold effect) after bariatric surgery in our study and by 19% (−1.23-fold effect) after diet induced weight loss in the data of Ghosh *et al*. We observed that the expression of *MYLIP* (myosin regulatory light chain interacting protein) increased by 29% (1.29-fold) in our study and by 75% (1.75-fold) in the Ghosh study. Overall, the range of gene expression changes that were decreased after weight loss was fairly comparable in the present study (4 to 46% decreased) to our analysis of the Ghosh data (7 to 63% decreased). Several of the genes we described whose expression in whole blood changed after bariatric surgery have been previously found to be associated with obesity and/or type 2 diabetes. Thus, one might be able to compare how different diets, drug therapies, and surgical interventions alter whole blood gene expression profiles, and determine the correlations of gene expression changes with clinical improvements in order to examine which pathways are affected by these different therapeutic regimens. In addition baseline whole blood transcriptomes can be examined among subjects who respond differently to different therapeutic treatments for obesity and type 2 diabetes. In the future, whole blood gene expression profiling might be used for personalized medicine in order to make informed clinical decisions among the different therapeutic options. In conclusion, our pilot study identified many genes whose expression was markedly altered in whole blood following bariatric surgery, some of which are related to inflammation and lipid metabolism. We identified seven genes whose expression changes were correlated with changes in HbA~1C~ content, weight, and FPG. Thus, whole blood expression levels of these and potentially other transcripts identified in this study may be useful as biomarkers indicative of type 2 diabetes susceptibility, as well as response to various therapeutic regimens. In addition, this type of study may illuminate new targets and pathways for intervention. Supporting Information {#s5} ====================== Table S1 ::: {.caption} ###### **Clinical data measured in 11 subjects that participated in the study before undergoing bariatric surgery (RYGB, Roux-en-Y Gastric Bypass; SG, Sleeve Gastrectomy).** Clinical characteristics that were found to change significantly after bariatric surgery are shown in bold (paired t-test p-value\<0.05). (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S2 ::: {.caption} ###### **List of significantly differentially expressed (paired t-test p-value\<0.01) transcripts IDs with their respective gene symbol, definition and percent change after bariatric surgery.** The genes discussed in the manuscript are shown in bold. (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S3 ::: {.caption} ###### **List represents significantly differentially expressed (paired t-test p-value\<0.01) transcripts IDs whose changes in gene expression are correlated to changes in at least one of the following four clinical characteristics: weight, FPG, HbA~1C~ or WBC, with an R^2^≥0.25.** Values with R^2^≥0.25 are shown in bold. Gene symbols in blue are correlated to changes in weight, FPG, and HbA~1c~. (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S4 ::: {.caption} ###### **Table shows the gene ontology (GO) terms and pathways identified by gene ontology and functional analyses (DAVID) that were overrepresented (P-value\<0.01) in our list of 204 significantly differentially expressed transcripts after bariatric surgery.** GO terms discussed in the manuscripts are shown in bold. (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: We thank the personnel of the Cleveland Clinic Clinical Research Unit for their assistance. A portion of this work was published as an abstract at Diabetes Journal, Volume 58, Supplement 1; A295 (June 2009). **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was supported by National Institute of Health Grants P50-HL077107 and R01-HL098193 (<http://www.nih.gov/>) to J.D.S, by National Center for Research Resources Multidisciplinary Clinical Research Career Development Programs Grant 5K12RR023264 (<http://www.ncrr.nih.gov/index.asp>) and Clinical and Translational Science Award Grant 1UL1RR024989 (<http://www.ctsaweb.org/>) to S.R.K.. S.Z.B. was supported by Cellular and Molecular Medicine Coordinating Committee Research Assistantship from Cleveland State University (<http://www.csuohio.edu/sciences/dept/biology/cmms/index.htm>). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: SZB DS PS SRK JDS. Performed the experiments: SZB DS PS SRK JDS. Analyzed the data: SZB DS SRK JDS. Contributed reagents/materials/analysis tools: DS PS SRK JDS. Wrote the paper: SZB DS SRK JDS.

PubMed Central

2024-06-05T04:04:19.731353

2011-3-10

PMC3053357

Introduction {#s1} ============ The epidermis of a plant acts as a barrier to the outside world, providing a waterproof layer that prevents dehydration of internal tissues. The majority of plant epidermal surfaces are composed of essentially flat cells. The occurrence of protruding cells, particularly trichomes (hairs) and papillae (single cells in the shape of cones), is associated with specific functions. For example, trichomes may be involved in deterring predators as well as moderating leaf boundary layer, and have also been found to influence the degree to which water is retained on the plant epidermis -- its wettability [@pone.0017576-Brewer1]. Previous studies have indicated that approximately 80% of plant species analysed have petal epidermal surfaces composed exclusively, or almost exclusively, of conical cells [@pone.0017576-Kay1]. The restriction of conical cells to the petal epidermis, and the frequency with which they are found on petals, has led several authors to conclude that they must function to enhance the attractiveness of the corolla to pollinating animals. There has been considerable debate as to how conical cells might function to increase floral attractiveness [@pone.0017576-Kay1]--[@pone.0017576-Whitney2]. It is also possible that petal cell shape affects floral surface wettability. Structures on the plant surface and surface chemistry can both have a significant effect on hydrophobicity or hydrophilicity [@pone.0017576-Brewer1], [@pone.0017576-Koch1], [@pone.0017576-Koch2]. The behaviour of surface water on a rough surface such as the plant epidermis was established by Wenzel [@pone.0017576-Wenzel1] and Cassie and Baxter [@pone.0017576-Cassie1]. In 'Wenzel wetting' the water is in close contact with the surface ([Figure 1A](#pone-0017576-g001){ref-type="fig"}) while in 'Cassie-Baxter wetting' air is trapped between parts of the surface and the drop ([Figure 1B](#pone-0017576-g001){ref-type="fig"}). Due to the presence of air-pockets under the droplet in Cassie-Baxter wetting, the water has less physical contact with the surface. In this case the drop has a very small contact angle hysteresis and rolls easily off the surface. In this case, the surface is superhydrophobic [@pone.0017576-Koch1], [@pone.0017576-Koch3]. ::: {#pone-0017576-g001 .fig} 10.1371/journal.pone.0017576.g001 Figure 1 ::: {.caption} ###### Diagram illustrating wettability behaviour of water on a rough surface. **A**. Wenzel wetting, where the water is in close contact with the surface. **B.** Cassie-Baxter wetting where air is trapped between parts of the surface and the drop. :::  ::: Such superhydrophobicity was previously observed on the leaves of the Sacred Lotus, *Nelumbo nucifera*, where extracellular conical wax extrusions present on these leaves reduce the wettability of the surface, allowing water droplets to bead and roll off [@pone.0017576-Barthlott1], [@pone.0017576-Neinhuis1]. This adaptation is clearly important for an aquatic plant, but it may also be of significance in maintaining leaves and petals of other species free from waterlogging. A consequence of the lack of wettability of the Lotus leaf is its "self-cleaning" properties. Water droplets rolling off the leaf remove particles of dirt, generating a clean surface, an effect known as "the Lotus effect", which has provided the inspiration for biomimetic applications [@pone.0017576-Barthlott1]. A lack of wettability on the leaf surface is thought to be a distinct advantage for several other reasons including that water reduces photosynthetic gas exchange, may promote pathogen infection and may enhance pollutant deposition [@pone.0017576-Brewer2], [@pone.0017576-Evans1]. In the case of flowers, wettability could influence pollinator preference. The presence of conical cells has already been shown to impact on pollinator choices [@pone.0017576-Glover1]--[@pone.0017576-Whitney1]. If the conical cells of petals generate a lack of wettability similar to that found in the Lotus effect, this may have a range of adaptive consequences, including potentially impacting on the ability of an insect to successfully grip a flower, a factor that has been shown to influence pollinator preferences [@pone.0017576-Whitney1]. The potential self-cleaning properties could also have an impact on the presence of scent marks, colour-obscuring dust, and bacterial cells and fungal spores, all of which could affect pollinator choice and contribute to general plant health [@pone.0017576-Schmitt1]--[@pone.0017576-Saleh1]. However, the wettability properties of the plant surface have so far only been studied either through biomimetics or through comparing the structural and chemical differences between the epidermal surfaces of different plant species from different habitats [@pone.0017576-Neinhuis1], [@pone.0017576-Brewer2], [@pone.0017576-Feng1]--[@pone.0017576-Aryal1]. Biomimetic methods can precisely determine how individual surface features influence wettability, and comparisons of different species determine how overall surface may influence plant ecology. This creates a difficulty in directly determining the ecological impact of specific surface wettabilites due to the range of interspecies difference in surface properties. The ability to manipulate individual components of the plant surface, and so directly impact on surface wettability, would be an invaluable tool to investigate the ecological relevance of this physical property. This can be achieved by the use of genetically modified or mutant lines. One mutant line that has been invaluable in the study of the function of the plant epidermis in plant-insect interactions is the *mixta* mutant of Antirrhinum. The *mixta* mutant lacks conical papillate petal cells, and instead produces flat petal epidermal cells, more similar to leaf epidermal cells ([Figure 2](#pone-0017576-g002){ref-type="fig"}). The mutant is a lesion in a single gene and has no other consequences for plant phenotype, including in the composition of cuticular waxes [@pone.0017576-Glover1], [@pone.0017576-Noda1]. The use of isogenic lines differing only in the *MIXTA* locus and therefore only in the shape of the petal epidermal cells allows accurate and sensitive dissection of the function of these specialised cells. Using these lines we have previously shown that conical cells increase fruit set and reduce pre-landing and post-landing rejection of flowers by bees [@pone.0017576-Glover1], [@pone.0017576-Comba1]. ::: {#pone-0017576-g002 .fig} 10.1371/journal.pone.0017576.g002 Figure 2 ::: {.caption} ###### Conical-celled and flat-celled petal surfaces. **A.** Scanning Electron Microscope image (SEM) of wild-type Antirrhinum petal. **B.** SEM of *mixta* mutant Antirrhinum petal. :::  ::: In this paper we consider the effects of conical cells on petal wettability. This is the first study to test the impact on plant surface wettability of changes in the plant epidermis due to a single gene, and to provide a model system in which the ecological importance of this property can be tested. Since bees can distinguish between the different colours of wild type and *mixta* flowers, and can learn to associate those colour differences with different rewards [@pone.0017576-Dyer1], [@pone.0017576-Dyer2], they could be using them as a cue associated with some other physical property of the flower. If cell shape significantly affects the wettability, and pollinators exhibit discrimination between flowers of different dryness, then this physical effect could also explain the preference of bees for conical-celled flowers. Here we show that petal cell shape has a significant influence on floral wettability. Conical cells have been shown to be multifunctional, and thus we conclude that the ability of conical cells to influence floral wettability could be one of the factors by which conical petal cells enhance plant fitness. We conclude that the use of isogenic lines is a powerful tool for the study of the ecological importance of plant surface wettability. Materials and Methods {#s2} ===================== Plant lines and growth conditions {#s2a} --------------------------------- Two lines of Antirrhinum plants were used, varying in petal cell shape. The isolation of the isogenic wild-type (*Mx^+^*), and *mixta* (*mx^−^*), lines is described in [@pone.0017576-Glover1]. These lines have been maintained by self-pollination since 1995 so genetic variation between individuals is almost absent. The *mixta* mutant has flat petal epidermal cells [@pone.0017576-Noda1]. Six plants of each of the Antirrhinum lines were grown under greenhouse conditions at 23°C in 4-inch pots in Levington\'s (UK) M3 compost. During the growth period plants received supplemental lighting from 400 Osram (Osram, München, Germany) lamps on a 16 hr light/8 hr dark photoperiod. Inflorescences with fully opened flowers were selected from each plant as they became available, and flowers within the inflorescence further selected to avoid flowers with petal surface damage or irregularities. The effect of petal cell shape on wettability {#s2b} --------------------------------------------- Surface wettability can be mostly characterised by two measurable quantities, the static contact angle of the water droplet and the contact angle hysteresis. The wettability of the fresh petals of the conical-celled wild type and flat-celled *mixta* mutant was determined by measuring the surface contact angle of single water drops as they were extruded onto the surface via a syringe needle (the advancing contact angle) or were removed (receding contact angle). The difference between the advancing and receding contact angles defines the contact angle hysteresis. If part of a droplet of water remained after attempted removal, a measurement of the static contact angle was taken. Two samples were cut from each of the lobes of 10 individual flowers from each line. All samples were laid flat and attached to a glass microscope slide by means of double-sided tape. A visual inspection of the sample confirmed that sample preparation had not damaged the petal surface, and any samples showing damage or irregularities were discarded. Using the methods detailed in [@pone.0017576-vanderWal1], measurements from were each sample were obtained using a contact goniometer (KSV CAM 200) equipped with a digital image acquisition system and an automatic liquid dispenser. Individual water drops were dispensed and then removed from the petal surface. If the sample was large enough, a second measurement was subsequently taken on a separate region of the sample. Images of the water drop as it advanced and receded across the petal surface were taken ([Figure 3](#pone-0017576-g003){ref-type="fig"} shows images of advancing and receding water drops on the petals of both wild type and *mixta* lines). The contact angles were determined from these side-on images of the drops using a numerical fitting algorithm [@pone.0017576-vanderWal1]. ::: {#pone-0017576-g003 .fig} 10.1371/journal.pone.0017576.g003 Figure 3 ::: {.caption} ###### Measurement of floral surface wettability. **A1**. Advancing angle of drop on surface showing Cassie-Baxter wetting. **A2**. Receding angle of drop showing Cassie-Baxter wetting. **A3**. Ease of drop removal on a surface showing Cassie-Baxter wetting. **B1**. Advancing angle of drop on surface showing Partial Cassie-Baxter wetting. **B2**. Receding angle of drop showing Partial Cassie-Baxter wetting. **B3**. Drop removal on a surface showing Partial Cassie-Baxter wetting, showing that while initial removal is similar to perfect Cassie-Baxter wetting, a localized point remains. **C1**. Advancing angle of drop on surface showing Wenzel wetting. **C2**. Receding angle of drop showing Wenzel wetting. **C3**. Attempted drop removal on a surface showing Wenzel wetting. :::  ::: Results {#s3} ======= Petal cell shape has a significant effect on petal wettability {#s3a} -------------------------------------------------------------- The wetting behaviours showed by the Antirrhinum petals could be categorised into three types of wetting behaviour; (**A**) Wenzel wetting (where the water is in close contact with the surface), (**B**) Cassie-Baxter wetting (where air is trapped between parts of the surface and the drop) and (**C**) Partial Cassie-Baxter wetting [@pone.0017576-Wenzel1], [@pone.0017576-Cassie1]. This third, intermediate type of wetting behaviour was found on the wild-type petals, where the drop was found to have very small contact angle hysteresis, but part of a droplet of water remained after attempted removal. Examples of all three types of wetting on Antirrhinum petal surfaces are shown in [Figure 3](#pone-0017576-g003){ref-type="fig"}. Of the 31 measurements made of the conical celled wild-type petals, drop behaviour corresponded 4 times to complete Cassie-Baxter wetting, 7 times to partial Cassie-Baxter wetting and 20 times to Wenzel wetting (see [Table 1](#pone-0017576-t001){ref-type="table"} for details and properties of the different wettability categories). Of 14 measurements made of the flat-celled *mixta* mutant, all showed Wenzel wetting. If the mutants showed the same wetting as the wild types, the probability of an individual mutant sample showing Wenzel wetting would be 0.645 and the probability of 14 mutant samples (out of 14) showing Wenzel wetting can therefore be calculated as 0.002 (exact value from binomial test). Mutants therefore have significantly different wetting properties to wild type flowers. ::: {#pone-0017576-t001 .table-wrap} 10.1371/journal.pone.0017576.t001 Table 1 ::: {.caption} ###### Designated wettability criteria and occurrence in Antirrhinum wild type (*Mx^+^*) and *mixta* (*mx^−^*) lines. ::: {#pone-0017576-t001-1} Perfect Cassie-Baxter wetting Partial Cassie-Baxter wetting Wenzel wetting ----------------------------------------------------- ------------------------------- --------------------------------------------- ------------------------------ Ease of drop removal Very easy Easy but with one localised point remaining Not easy Number of examples in wild-type Antirrhinum flowers 4 7 20 Average angle when drop was: Average angle when drop was: Average angle when drop was: advancing (148±6) advancing (140±18) advancing (135±17) receding (109±22) receding (91±16) receding (73±19) Number of examples in *mixta* Antirrhinum flowers 0 0 14 Average angle when drop was: advancing (120±14) receding (74±14) Picture sequences refer to [Figure 3](#pone-0017576-g003){ref-type="fig"}. ::: Discussion {#s4} ========== We find that, if studied independently of surface chemistry by using an isogenic mutant line, cell shape does play an important role in determining petal wettability. The presence of conical cells renders the surface weakly superhydrophobic. This weak superhydrophobicity is not present for the *mixta* mutant, which lacks the conical-celled surface corrugation. Generally, if they had an identical cuticular composition, flat-celled petals would therefore be more wettable than conical-celled flowers, whether of the same or different species. This finding is in line with studies by Barthlott and Neinhuis [@pone.0017576-Barthlott1], [@pone.0017576-Neinhuis1], which showed that the papillate leaf cells of the Sacred Lotus *N. nucifera* were significantly less wettable than ordinary flat leaf cells. They concluded that this lack of wettability helped the plant to prevent waterlogging in its aquatic habitat, and showed that the formation of beads of water caused dirt to be more easily removed from the leaf. The conical cells found on Antirrhinum flowers produce a much less robust hydrophobicity than that shown on *N. nucifera* as the exact contact angle was strongly influenced by very local petal surface properties. As well as Wenzel and Cassie-Baxter wetting, a third type of wetting was also found on the wild type flowers. While the division between Wenzel and Cassie-Baxter wetting is usually very distinct, surfaces can display mixtures of these two behaviours. This is the case for surfaces that are close to the crossover between these two regimes, for which both wetting types have similar free energies (weakly superhydrophobic surfaces). On natural surfaces, even small amounts of structural damage can induce Wenzel wetting on surfaces that otherwise have Cassie-Baxter properties. This suggests that even small regions of damage caused by pathogens or insects or small amounts of contamination, such as grains of pollen, could be enough to seriously perturb the degree of hydrophobicity. The observed weak superhydrophobicity might therefore have a self-cleaning function for dewdrops that form directly on the petal surfaces, but probably not for impacting drops (such as rain). The self-cleaning effect observed for Lotus leaves could therefore explain the prevalence of conical cells on Angiosperm petals. Petals are exposed to dust, dirt and pollen, which might interfere with their display and their attraction of pollinators. They are also exposed to pathogenic bacteria and fungal spores. Ability to remove all of these contaminants by self-cleaning could significantly enhance pollination success, although each contaminant will be differently influenced by the surface, depending on viscosity. The self-cleaning properties observed in petals were much less robust than those found on Lotus leaves. This possibly arises from the difficult handling of the petals in the contact angle goniometer. However, lotus-like surface features (micrometre-sized wax crystals on surfaces) could potentially interfere with other floral roles, such as pollinator handling and colour display. Conical cells, on the other hand, not only do not interfere with other floral roles, but actually enhance both the visual display and ease of pollinator handling while providing a degree of hydrophobicity [@pone.0017576-Whitney1], [@pone.0017576-Whitney2], [@pone.0017576-Noda1] and therefore, though less robust than the lotus-like surface features, may be optimal for both display and protection. Analysis of the distribution of conical petal cells with relation to habitat, dew and rainfall and pollinator type would allow discrimination between some of these possibilities. As mentioned before, the petal epidermis is highly multifunctional. Individual floral features, for example conical cells, have been implicated in a wide range of floral properties. Due to the *mixta* Antirrhinum line, the various impacts of conical cells have begun to be untangled, and they have been found to impact directly on floral colour, floral temperature, flower shape, pollinator grip and now flower wettability (this study). How much each of these factors contributes to the maintenance of the presence of conical cells in such a high proportion of flowering plants, how much the production of conical cells costs, and the trade-off between these factors will be the subject of on going ecological studies both to establish the contributions of these cells in different situations and to determine why a feature that appears to be useful in such a variety of ways is absent from a proportion of Angiosperm species.. A similar approach could be used to determine the strengths of the hypotheses regarding wettability of leaves. Hydrophobicity in leaves is thought to be maintained due to a number of advantages, including that a layer of water reduces photosynthetic gas exchange, promotes pathogen infection and enhances pollutant deposition [@pone.0017576-Brewer2], [@pone.0017576-Evans1]. Using isogenic lines that vary in wettability traits within ecologically relevant contexts could test each of these ideas. In conclusion, we have shown that petal cell shape does significantly influence the physical properties of petals. The commonly found conical petal epidermal cells make a great deal of difference to wettability. Thus, besides providing visual and tactile cues to pollinating animals, conical petal cells may also act to prevent waterlogging and to optimise shedding of dirt and pathogens. Conical cells have been found to be multifunctional, and these many roles may explain the prevalence of the conical petal cell form in the flowering plants. We thank Matthew Dorling and the staff of the Cambridge University Botanic Garden for excellent plant care and Sean Rands for help and advice with statistics. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**HMW was supported by a NERC grant NE/C000552/1 (to BJG and LC) and by the Cambridge University Research Exchange. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: HMW RP US LC BJG. Performed the experiments: HMW RP. Analyzed the data: HMW. Contributed reagents/materials/analysis tools: HMW RP US BJG. Wrote the manuscript: HMW RP US LC BJG.

PubMed Central

2024-06-05T04:04:19.736087

2011-3-10

PMC3053358

Introduction {#s1} ============ Small airway disease frequently occurs in chronic obstructive pulmonary disease, asthma, and cystic fibrosis (CF) [@pone.0017588-Contoli1]. Despite lack of respiratory symptoms in *e.g.* adult smokers [@pone.0017588-Stanescu1], children with asthma [@pone.0017588-Gustafsson1] or CF [@pone.0017588-Sly1]; [@pone.0017588-Tiddens1], small airway malfunction may be present and is regarded as important sign of early lung disease. Conventional lung function techniques such as spirometry are considered to be not sensitive enough to detect small airway malfunction [@pone.0017588-Gustafsson1]; [@pone.0017588-Aurora1]--[@pone.0017588-Estenne1]. Inert tracer gas washout tests over single or multiple breaths (SBW or MBW) provide a more sensitive alternative to spirometry in tracking small airway malfunction, *e.g.* altered ventilation inhomogeneity (VI), and they are more accurate in reflecting structural changes in lung periphery [@pone.0017588-Gustafsson2]; [@pone.0017588-VanMuylem1]--[@pone.0017588-Aurora3]. However, these tests are not commonly used in clinical routine as they rely on custom made, expensive, and bulky setups, *e.g.* mass spectrometer (MS), and require coordination of vital capacity manoeuvres for SBW or cooperation during 20 minutes of tidal breathing for MBW [@pone.0017588-Coates1]. Recently, the ultrasonic flowmeter (USFM) technique was applied in MBW studies [@pone.0017588-Buess1]--[@pone.0017588-Pillow1]. The USFM measures total molar mass (MM), a sum signal derived from measured gas density [@pone.0017588-Latzin1]--[@pone.0017588-Fuchs2]. Tracer gas mixtures had different MM compared to air and contained a single tracer gas, either sulfur hexafluoride (SF~6~) or helium (He) [@pone.0017588-Buess1]--[@pone.0017588-Pillow1]. MM of SF~6~ (146 g/mol) is much higher than MM of He (4 g/mol), thus SF~6~ and He distribute unequally in lung periphery where diffusion predominates. The diffusion front for He is thought to arise in the zone of the entrance to the *acinus*. In contrast, the diffusion front for SF~6~ is predicted to occur more distally within the *acinus* [@pone.0017588-Lacquet1]. Different structural asymmetries within the lung zones where the diffusion fronts of He and SF~6~ arise thus lead to unequal washout patterns of He and SF~6~. Using both gases for a SBW may provide more specific information about ventilation in these peripheral lung zones [@pone.0017588-Paiva1]--[@pone.0017588-Gustafsson3]. A tracer gas mixture containing SF~6~ and He and exhibiting similar MM as air could be measured by an USFM to assess washout patterns of SF~6~ and He. Using the USFM for a modified SBW procedure would eliminate some shortcomings of current tracer gas washout tests in clinical routine. The aim of this study was to develop and validate a simple technique for a new tidal SBW of SF~6~ and He using an USFM. Materials and Methods {#s2} ===================== Ethics statement {#s2a} ---------------- The study was approved by both the Ethics Committee of the Canton of Bern, Switzerland (Kantonale Ethikkommission Bern) and the Research Ethics Committee of University Hospital Bern (Inselspital). All participants provided written informed consent for this study. Study design {#s2b} ------------ In this feasibility study, a SBW of a double tracer gas mixture (DTG-SBW) using an USFM was applied in 13 healthy adults during tidal breathing. The subjects\' mean (SD) age was 35.2 (9.4) years. Accuracy of the USFM compared to mass spectrometry (MS), and repeatability and reproducibility of this DTG-SBW test were assessed. Double tracer gas {#s2c} ----------------- SF~6~ and He, two inert tracer gases, were employed in a SF~6~/He ratio of 1/5.26 to establish similar MM of DTG compared to dry medical-grade air (28.9 g/mol). DTG contained 5% SF~6~, 26.3% He, 21% oxygen (O~2~), and balance nitrogen (N~2~) (Carbagas, Domdidier, Switzerland) and was applied in all DTG-SBW tests (n = 60). Tidal single breath washout {#s2d} --------------------------- Subjects were measured in an upright sitting position, wearing a nose clip, and breathing through a disposable bacterial filter (air™ Vickers Industrial Estate, Lancashire, UK) attached to the flow-head ([figure 1](#pone-0017588-g001){ref-type="fig"}). Prior to the DTG-SBW, subjects tidally breathed air for 20 seconds until steady shapes of flow-volume-loops were established. At the beginning of an expiration, DTG was switched on manually to flush the system. A tidal volume of DTG was inhaled from functional residual capacity (FRC) prior to exhaling back to FRC. DTG-SBW was technically accepted if the test breath had a similar flow-volume-loop as pre-test breaths. A minimum of ten subsequent breaths of air were required prior to the next DTG-SBW, and three DTG-SBW tests were done per test occasion. ::: {#pone-0017588-g001 .fig} 10.1371/journal.pone.0017588.g001 Figure 1 ::: {.caption} ###### Ultrasonic flowmeter setup. Molar mass was measured in a sidestream ultrasonic flowmeter (USFM) and flow was measured in a mainstream USFM. Oxygen (O~2~), and carbon dioxide (CO~2~) were measured in sidestream sensors. :::  ::: Washout analysis {#s2e} ---------------- The wave form of naturally exhaled MM is attributed to the increasing carbon dioxide (CO~2~) fraction [@pone.0017588-Thamrin1]. During DTG-SBW, CO~2~, SF~6~, and He fractions were expected to give rise to the USFM derived MM (MM~USFM~) signal. Therefore, we transformed the CO~2~ signal into MM and subtracted this signal from MM~USFM~. We then obtained a single MM signal potentially reflecting SF~6~ and He washout (SF~6~-He~USFM~). The main outcome of the DTG-SBW analysis was the shape of the SF~6~-He~USFM~ signal plotted as expirogram, *i.e.* MM against expired volume. Despite some limitations, MS is still regarded as current gold standard for quantification of respiratory gases [@pone.0017588-Pillow1]; [@pone.0017588-Fuchs1]. During DTG-SBW we compared MM~USFM~ and SF~6~-He~USFM~ signals with the two corresponding MS signals (MM~MS~, SF~6~-He~MS~) which were derived as follows. The MM signal (MM~MS~) was calculated by summing fractional MM of respective gas concentrations. The SF~6~ and He signals were normalized by dividing them by their starting concentrations [@pone.0017588-Robinson1]. The normalized He signal was then subtracted from the normalized SF~6~ signal to obtain a single tracer gas signal (SF~6~-He~MS~). Accuracy and reproducibility of the single breath washout method {#s2f} ---------------------------------------------------------------- Three DTG-SBW were applied in six healthy adults on a single test occasion to compare USFM signals with MS signals. MM~USFM~ and SF~6~-He~USFM~ were compared graphically with MM~MS~ and SF~6~-He~MS~. Signal-to-noise ratio of MM~USFM~ (MM~USFM~ mean/MM~USFM~ standard deviation (SD)) was assessed for air and DTG at 1 L/s flow during ten seconds. To assess repeatability and reproducibility of the DTG-SBW test, three DTG-SBW tests were applied in seven healthy adults on two test occasions 24 hours apart. We calculated area under the washout curve (AUC) by integrating a best-fit double exponential curve (Matlab; The Mathworks Inc., Natick, MA, USA) to mathematically describe the shape of the SF~6~-He~USFM~ expirogram. We constrained AUC calculation to phases II and III of the CO~2~ expirogram ([figure 2](#pone-0017588-g002){ref-type="fig"}) to assess washout patterns from bronchial and alveolar gases. While during the first phase CO~2~-free gas is exhaled from airway dead space, the rapidly increasing CO~2~ fraction forms a sigmoidal curve of the bronchial phase II ending up in the plateau of the alveolar phase III [@pone.0017588-Tang1]. ::: {#pone-0017588-g002 .fig} 10.1371/journal.pone.0017588.g002 Figure 2 ::: {.caption} ###### Area under the single breath washout curve. The ultrasonic flowmeter (USFM) derived sulfur hexafluoride (SF~6~) and helium (He) molar mass signal (SF~6~-He~USFM~) was plotted as expirogram. The carbon dioxide (CO~2~) expirogram (dashed line) was plotted to determine area under the washout curve (AUC, grey area) during washout of bronchial and alveolar gas fronts [@pone.0017588-Tang1]. As SF~6~-He~USFM~ had been calculated by subtracting CO~2~ from molar mass (MM~USFM~), a "negative" SF~6~-He~USFM~ signal resulted from a low MM~USFM~ signal reflecting relatively more He than SF~6~ contribution to MM~USFM~. In phase I, relatively more He than SF~6~, but during phases II and III, relatively more SF~6~ than He sequentially arrived at the mouth. :::  ::: Hardware {#s2g} -------- We used a commercially available USFM setup (Exhalyzer D®, Eco Medics AG, Duernten, Switzerland) described previously ([figure 1](#pone-0017588-g001){ref-type="fig"}) [@pone.0017588-Schibler1]; [@pone.0017588-Pillow1]. The USFM measured MM in sidestream sampling mode with a sample flow of 200 mL/min via a Nafion® tube to allow equilibration of ambient temperature and humidity [@pone.0017588-Fuchs1]; [@pone.0017588-Fuchs2]. Measurement precision was 0.01 g/mol at a sampling frequency of 200 Hz [@pone.0017588-Schibler1]; [@pone.0017588-Fuchs2]. Tidal flows and derived volumes were measured in mainstream gas using the flow-head USFM. Into this flow-head, a dead-space reducer (DSR size 3) and a disposable hygienic insert (Spirette), both provided by the manufacturer (Eco Medics AG), were inserted. Total dead space of the flow-head plus the bacterial filter attached was 40 mL. A three-way valve system operated manually administered air or DTG via by-pass flow at 1 L/s effecting a resistance of 0.01 kPa s L-1. Prior to measurements, the USFM was calibrated for inspiratory and expiratory volumes using a precision syringe. Gas concentrations (SF~6~, He, N~2~, O~2~, and CO~2~) using a respiratory mass spectrometer (AMIS 2000, Innovision A/S, Odense, Denmark) and respiratory flows using a heated pneumotachograph were measured near airway opening as previously described [@pone.0017588-Gustafsson4]. An additional sidestream sampling Nafion® tube was introduced between the mouthpiece and the flow-head. The sample flow of the MS was 20 mL/min and the gas signals were updated at a rate of 33.3 Hz. The PC based data acquisition setup recorded flow and dry gas concentrations at 100 Hz. Software {#s2h} -------- A software package (WBreath® 3.28; ndd Medical Technologies, Switzerland) was used for collection of USFM and MS signals. Signals were aligned in time as previously described [@pone.0017588-Pillow1]; [@pone.0017588-Gustafsson4]. Tidal flows and derived volumes were converted to body temperature and ambient pressure, and saturated with water vapour (BTPS) conditions. Statistical analysis {#s2i} -------------------- The association of MM~USFM~ and SF~6~-He~USFM~ with MM~MS~ and SF~6~-He~MS~ signals were assessed graphically and using a linear regression model accounting for clustered data within individuals. Means of MM~USFM~ and MM~MS~ were compared using two-tailed paired *t*-tests. Accuracy of MM~USFM~ compared to MM~MS~ was determined using the *Bland and Altman* method [@pone.0017588-Bland1] by plotting differences of paired measurements against means of paired measurements. Intra-test repeatability of DTG-SBW was calculated as intra-subject mean coefficient of variation (CV%  =  SD/mean\*100) of AUC. DTG-SBW curves and AUC from test occasions 24 hours apart were compared graphically, and using the two-tailed paired *t*-test. Between-test reproducibility of DTG-SBW was assessed graphically, and by the coefficient of repeatability of AUC representing the 95% range of differences between two repeated measurements and was calculated as twice the square root of the mean of squared differences of paired measurements [@pone.0017588-Bland1]; [@pone.0017588-Chinn1]. Means, SD, and 95% confidence intervals (95% CI) were reported, p-values \<0.05 were considered significant, and all analysis were done using Stata™ (StataCorp. 2009. Stata Statistical Software: Release 11. College Station, TX: StataCorp LP). Results {#s3} ======= DTG-SBW was feasible and technically acceptable in all subjects (n = 13; 7 males) during all tests (n = 60). Mean (SD) duration of one DTG-SBW test was 76 [@pone.0017588-Bland1] seconds. The MM~USFM~ signal of the DTG-SBW was significantly different to the naturally exhaled MM~USFM~ signal due to CO~2~ as shown in one male subject ([figure 3](#pone-0017588-g003){ref-type="fig"}). Different patterns of SF~6~ and He washout was observed throughout all DTG-SBW tests reflecting a non-linear washout relationship of these tracer gases. An increase in MM~USFM~ indicated an increase in SF~6~ washout relative to He washout. This pattern was mainly observed in phase II of the CO~2~ expirogram ([figures 2](#pone-0017588-g002){ref-type="fig"} and [3](#pone-0017588-g003){ref-type="fig"}). ::: {#pone-0017588-g003 .fig} 10.1371/journal.pone.0017588.g003 Figure 3 ::: {.caption} ###### Comparison of single breath washout signals. Typical ultrasonic flowmeter (USFM) and mass spectrometer (MS) signals from a single breath washout (SBW) using sulfur hexafluoride (SF~6~) and helium (He) in a healthy male adult. USFM derived molar mass (MM~USFM~ black solid line) and MS derived molar mass (MM~MS~ grey dashed line) reflected changes in SF~6~ washout relative to He washout measured using MS: SF6~MS~ (black dashed line), and He~MS~ (black dotted line). The MM signal derived from CO~2~ (MM~CO2~ grey solid line) reflected naturally exhaled MM similar to MM signals from pre-test breaths. :::  ::: Accuracy of the single breath washout method {#s3a} -------------------------------------------- MM~USFM~ and MM~MS~ signals were compared using paired data from 18 tests. The association of MM~USFM~ and MM~MS~ signals was high, the adjusted linear regression coefficient r^2^ was 0.98 ([figure 4a](#pone-0017588-g004){ref-type="fig"}). Strong agreement of MM~USFM~ and MM~MS~ was found in the *Bland and Altman* plot ([figure 4b](#pone-0017588-g004){ref-type="fig"}) without graphical evidence of systematic bias or significant outliers. Mean (95% CI) difference in MM was −0.0004 (−0.0021 to 0.0013) g/mol. The range of 95% of all differences was 0.13 g/mol (0.45% of mean MM of both methods). The MM~USFM~ signal-to-noise ratio for air and DTG was 3490 and 3140, respectively. ::: {#pone-0017588-g004 .fig} 10.1371/journal.pone.0017588.g004 Figure 4 ::: {.caption} ###### Accuracy of the ultrasonic flowmeter signal. Six healthy adults performed three tidal single breath washout tests of sulfur hexafluoride (SF~6~) and helium (He) during tidal breathing. Molar mass was measured using an ultrasonic flowmeter (MM~USFM~) and mass spectrometer (MM~MS~). [Figure 4a](#pone-0017588-g004){ref-type="fig"}: Paired MM~USFM~ and MM~MS~ data from 18 tidal single breath washout tests in six subjects were plotted against each other. Accounting for clustered data, adjusted linear regression coefficient r^2^ = 0.98 (p\<0.001). [Figure 4b](#pone-0017588-g004){ref-type="fig"}: *Bland and Altman* plot of MM~USFM~ - MM~MS~ differences against mean MM of both methods [@pone.0017588-Bland1]. Dashed lines indicated the mean difference of MM (−0.0004 g/mol), and upper and lower limits of agreement (mean of difference ± 2 SD of differences): 0.131 to −0.132 g/mol). :::  ::: Calculation of SF~6~-He~USFM~ and SF~6~-He~MS~ signals was feasible in all DTG-SBW tests. SF~6~-He~USFM~ washout curves were strongly associated with SF~6~-He~MS~ washout curves ([figure 5a](#pone-0017588-g005){ref-type="fig"}). Paired data of SF~6~-He~USFM~ and SF~6~-He~MS~ from 18 tests were plotted against each other, the adjusted linear regression coefficient r^2^ was 0.96 ([figure 5b](#pone-0017588-g005){ref-type="fig"}). A relative increase in either tracer gas was reliably reflected in the SF~6~-He~USFM~ signal. ::: {#pone-0017588-g005 .fig} 10.1371/journal.pone.0017588.g005 Figure 5 ::: {.caption} ###### Comparison of sulfur hexafluoride and helium washout curves. Sulfur hexafluoride (SF~6~) and helium (He) washout signals measured using an ultrasonic flowmeter (SF~6~-He~USFM~) and mass spectrometer (SF~6~-He~MS~) derived from a tidal single breath washout (SBW). [Figure 5a](#pone-0017588-g005){ref-type="fig"}: Typical SBW signals from one healthy male adult. SF~6~-He~USFM~ (circles) and SF~6~-He~MS~ (triangles) were plotted as expirogram against expired volume. [Figure 5b](#pone-0017588-g005){ref-type="fig"}: SF~6~-He~USFM~ was plotted against SF~6~-He~MS~ derived from 18 tidal SBW tests in six subjects. Accounting for clustered data, adjusted linear regression coefficient r^2^ was 0.96 (p\<0.001). :::  ::: To investigate if the shape of the SF~6~-He~USFM~ expirogram was robust against technical factors, we assessed the impact of possible VI caused by the measurement setup and variable dead space, respectively, on the shape of SF~6~-He~USFM~. First, three DTG-SBW were performed using a 500 mL precision syringe at 40 tidal strokes per minute, and a 1000 mL precision syringe at 20 tidal strokes per minute, respectively. DTG-SBW tests (n = 6) in these precision syringes resulted in flat SF~6~-He~USFM~ signals similar to signals of pre-test strokes using air. Second, pre- and post-capillary dead spaces were increased step-wise using 3.5 mL tubes resulting in 17.5 mL additional dead space on either side of the sidestream MM~USFM~ sampling inlet. During each step, one DTG-SBW test was applied in one healthy adult. For each step (n = 10), SF~6~-He~USFM~ expirograms were similar to those recorded using the original setup. Reproducibility of the single breath washout method {#s3b} --------------------------------------------------- The shape of SF~6~-He~USFM~ expirograms was repeatable and reproducible ([figure 6a](#pone-0017588-g006){ref-type="fig"}). Calculation of AUC was feasible in all tests (n = 42). On day one, mean (SD) AUC was 24.5 (6.7) g/mol\*%volume, and on day two, mean (SD) AUC was 24.6 (6.7) g/mol\*%volume ([figure 6b](#pone-0017588-g006){ref-type="fig"}). Within-test repeatability given as mean (SD) intra-subject CV was 6.8% (3.2%). Between-test reproducibility assessed graphically was good without evidence of systematic bias or significant outliers. The coefficient of repeatability was 2.9 g/mol\*%volume corresponding to 11.8% of mean AUC of both visits. Mean (95% CI) difference of AUC between the two test occasions was −0.15 (−0.82 to 0.50) g/mol\*%volume. ::: {#pone-0017588-g006 .fig} 10.1371/journal.pone.0017588.g006 Figure 6 ::: {.caption} ###### Reproducibility of the single breath washout. Seven healthy adults performed three double tracer gas single breath washout tests (DTG-SBW) 24 hours apart. Ultrasonic flowmeter (USFM) derived sulfur hexafluoride (SF~6~) and helium (He) washout (SF~6~-He~USFM~) signals were plotted as expirogram against percentage of total expired volume. [Figure 6a](#pone-0017588-g006){ref-type="fig"}: All SF~6~-He~USFM~ signals of DTG-SBW tests (n = 42) of visit one (black lines) and two (dashed lines) were plotted per subject. [Figure 6b](#pone-0017588-g006){ref-type="fig"}: AUC of the SF~6~-He~USFM~ from three DTG-SBW tests were plotted per test occasion and subject with a single symbol for each subject. Intra-individual changes of AUC between visit one and two were tracked via connecting lines. :::  ::: Discussion {#s4} ========== The USFM accurately measures relative changes in SF~6~ and He washout. DTG-SBW tests are repeatable and reproducible in healthy adults. A tracer gas mixture of similar MM as air is suitable to explore SF~6~ and He washout using an USFM. The shape of the MM expirogram reflects the sequential arrival of SF~6~ and He at airway opening. During a single tidal breath, SF~6~ and He washout patterns are different as hypothesized due to their different physical characteristics and their interaction with normal airway structure. The AUC of DTG-SBW showed high intra-test repeatability and between-test reproducibility in healthy adults. Strengths of the single breath washout method {#s4a} --------------------------------------------- This DTG-SBW method has several strengths with regards to practicability, informative value, and hardware. DTG-SBW tests are easy and quick to perform, requiring only 30 to 40 seconds for a few tidal breaths, and minimal cooperation. Compared to tracer gas SBW tests based on vital capacity maneuvers, classical outcomes, *e.g.* closing volume, are not assessed. However, indices derived from SBW tests near FRC may be even more sensitive for small airway disease [@pone.0017588-VanMuylem1]. We used low tracer gas concentrations minimizing respective physical interaction and consumption per test [@pone.0017588-Worth1]. Compared to MBW tests, consumption of SF~6~, a known greenhouse gas, is considerably smaller in SBW tests. USFM setups are economic and handy, thus probably more suited for clinical routine compared to MS [@pone.0017588-Coates1]; [@pone.0017588-Pillow1]; [@pone.0017588-Fuchs1]; [@pone.0017588-Fuchs3]. While the USFM does not allow measurements of single gas concentrations, it accurately measures relative changes of SF~6~ and He at a single spot. Using MS, a delay correction for each gas signal is needed [@pone.0017588-Estenne1]; [@pone.0017588-Gustafsson3]. With regards to higher signal-to-noise ratio, signal resolution, and lower technical dead space, the USFM technique is probably better qualified than MS to assess gas signals with high fluctuation in *e.g.* infants [@pone.0017588-Latzin1]; [@pone.0017588-Frey1]. Accuracy of the single breath washout method {#s4b} -------------------------------------------- Our findings are in good agreement with previous studies investigating the accuracy of an USFM for MBW using SF~6~ [@pone.0017588-Pillow1]; [@pone.0017588-Fuchs1]. *Fuchs et al.* [@pone.0017588-Fuchs1] compared the USFM with MS for MBW, and 95% of differences between USFM and MS signals were within a range of \<1% of mean SF~6~ concentration. Even not focusing exclusively on end-expiratory tracer gas levels [@pone.0017588-Fuchs1], which probably reveals more stable signals, we were able to demonstrate excellent MM~USFM~ signal accuracy ([figure 4b](#pone-0017588-g004){ref-type="fig"}). In general it has to be acknowledged that the observed signal differences were well below 1% of the mean MM ([figure 4b](#pone-0017588-g004){ref-type="fig"}) and are probably due to inaccuracies of both methods, as no systematic bias was evident. Reproducibility of the single breath washout method {#s4c} --------------------------------------------------- The DTG-SBW was highly repeatable and reproducible with low inherent variation in measurements over time. The CV of AUC compares favourably to CV of other tidal breathing or vital capacity techniques. CV of slope of phase III derived from vital capacity N~2~ SBW in children and adults was 13% and 15%, respectively [@pone.0017588-Teculescu1]; [@pone.0017588-Teculescu2]. In healthy adults, intra-individual change between days in forced expiratory volume in one second (FEV~1~) of 11% is attributed to true clinical change [@pone.0017588-Pellegrino1]. For DTG-SBW, an intra-individual change in AUC of more than 12% would be unlikely due to measurement noise and thus reflect a true physiological impact with 95% probability [@pone.0017588-Chinn1]. Certainly, more data from DTG-SBW tests in a population of interest are required to estimate significant change. Possible mechanisms and physiological relevance {#s4d} ----------------------------------------------- Single tracer gas SBW tests are not specific for small airway disease as these tests do not allow separation between VI due to convective gas transport in large airways and VI due to interaction between diffusion and convection resulting in small airways [@pone.0017588-Paiva1]--[@pone.0017588-Gustafsson3]. Based on the *Paiva and Engel* lung model [@pone.0017588-Lacquet1], the diffusion front for SF~6~ arises more distal than for He. Greater structural asymmetries in lung periphery and the *Peclet* number describing the ratio of convection and diffusion transport of gases contribute to the differing distributions of SF~6~ and He. In healthy subjects these gas transport mechanisms result in a non-linear washout relationship of SF~6~ and He which may be altered in small airway disease most likely due to structural alterations in lung periphery [@pone.0017588-Estenne1]; [@pone.0017588-VanMuylem1]; [@pone.0017588-Lacquet1]; [@pone.0017588-Gustafsson3]; [@pone.0017588-Crawford1]--[@pone.0017588-Verbanck1]. We assume that these mechanisms determined the shape of the DTG-SBW curve in our study. During expiratory phase I representing absolute dead space, a "negative" MM signal reflecting relatively more He than SF~6~ washout was observed ([figure 2](#pone-0017588-g002){ref-type="fig"}). Subsequently the "positive" MM signal reflected relatively more SF~6~ than He washout during the bronchial and alveolar phases ([figures 2](#pone-0017588-g002){ref-type="fig"} and [3](#pone-0017588-g003){ref-type="fig"}). These findings are consistent with MS based SBW studies using SF~6~ and He in healthy adults [@pone.0017588-VanMuylem1]; [@pone.0017588-VanMuylem2]. Limitations and open questions {#s4e} ------------------------------ We did not apply DTG-SBW tests in children or diseased subjects. Thus further studies supporting its feasibility are needed. However, tidal breathing lung function tests have been already successfully applied in healthy and diseased young children [@pone.0017588-Aurora1]; [@pone.0017588-Latzin1]; [@pone.0017588-Thamrin1]; [@pone.0017588-Fuchs3]. As MM~USFM~ depends on humidity and temperature, sidestream sampling was applied to address this issue [@pone.0017588-Fuchs1]; [@pone.0017588-Fuchs2]; [@pone.0017588-Latzin2]. Compared to mainstream techniques, sidestream sampling introduces signal delay and slightly increases dead space [@pone.0017588-Thamrin1]. In our study, system inherent VI and up to 44% increase of technical dead space on either side of the MM sampling tube did not affect the shape of the DTG-SBW curve. This suggests that within these volume limits SF~6~ and He transport mechanisms are not significantly altered by increased dead space, *i.e.* gas bulks of SF~6~ and He may rather flow by convection than diffusion [@pone.0017588-Lacquet1]. Further data are necessary to identify potential confounders of the DTG-SBW curve such as breathing pattern, lung volume, and flow [@pone.0017588-Crawford2]; [@pone.0017588-Gronkvist1]. These issues, however, apply to MBW tests or MS and other USFM setups as well [@pone.0017588-Fuchs1]. We propose AUC as first and straightforward index to quantify the shape of the SF~6~ and He washout curve, but suggest that this complex washout relationship would be best explained by appropriate modelling allowing more information on lung physiology to be obtained. Complementary data on airway disease may be gathered comparing DTG-SBW indices with diffusion indices derived from upcoming sophisticated lung imaging techniques, *e.g.* hyperpolarized helium magnetic resonance imaging [@pone.0017588-Yablonskiy1], or additional lung function tests, such as MBW, vital capacity SBW or electrical impedance tomography [@pone.0017588-Riedel1]. Conclusion {#s4f} ---------- Relative change in SF~6~ and He washout may be viewed as a marker of functional changes in lung periphery, making it potentially sensitive to pathological processes affecting the structure of this ventilation zone. We have developed a fast, reliable, and straightforward USFM based SBW method, which provides valid information on SF~6~ and He washout patterns during tidal breathing in healthy adults. This easy SBW test has potential for widespread use in clinical and research settings. The current authors would like to thank Mr. Rudolf Isler for his extremely valuable input in the development of the method and his skilful technical assistance, and Eco Medics AG for supporting this study with their patented equipment (patent no. WO 2009/030058A1, priority date 09/07/2007). **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**The work for this report was funded by the Confederation\'s Innovation Promotion Agency (CTI; <http://www.bbt.admin.ch/kti/>) grant no. 11661.1 PFLS-LS. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: FS GS CT TR PG UF PL. Performed the experiments: FS OF PG PL. Analyzed the data: FS GS CT OF TR PG PL. Contributed reagents/materials/analysis tools: GS CT PG UF. Wrote the paper: FS GS PL.

PubMed Central

2024-06-05T04:04:19.738026

2011-3-10

PMC3053359

Introduction {#s1} ============ Harmful algal blooms (HABs) are commonly known for their detrimental impacts on aquatic organisms (including marine mammals), human health, and also local economies. *Karenia brevis* blooms, often referred to as "Florida red tide", occur in the Gulf of Mexico (Florida, USA) on a nearly annual basis [@pone.0017394-Steidinger1]. This dinoflagellate produces a suite of neurotoxins known as brevetoxins (PbTxs or BTXs), which are heat-stable, lipid-soluble, polyether compounds [@pone.0017394-Catterall1], [@pone.0017394-Lin1], [@pone.0017394-Baden1]. The proximate pharmacological target of PbTx is site 5 on the voltage-gated sodium channel [@pone.0017394-Catterall1], where they bind with high affinity (K~d~ 1--50 nM; [@pone.0017394-Poli1]. Once bound, these toxins alter the voltage sensitivity of the channel by interfering with the voltage sensor and inactivation gate, ultimately resulting in nerve inhibition [@pone.0017394-Huang1], [@pone.0017394-Ramsdell1] Human health effects of brevetoxins generally result from neurotoxic shellfish poisoning (NSP) [@pone.0017394-McFarren1] and/or respiratory illness caused by inhalation of aerosolized toxin [@pone.0017394-Kirkpatrick1]. The former can occur following consumption of contaminated shellfish that have accumulated sufficient levels of toxin while filter feeding the algal assemblage including *K. brevis*. This illness typically affects the nervous and gastrointestinal systems; however, all symptoms are reversible and to date there have been no human deaths associated with NSP [@pone.0017394-McFarren1], [@pone.0017394-VanDolah1]. Humans can also be exposed to brevetoxins in aerosolized form when the fragile, unarmored *K. brevis* cells burst due to wave action, releasing toxins into the air [@pone.0017394-Woodcock1] and causing irritation and burning of the throat and upper respiratory tract [@pone.0017394-Asai1]. In addition to brevetoxins acting as potent ichthyotoxins [@pone.0017394-Steidinger2], Bossart *et al.* [@pone.0017394-Bossart1] observed brevetoxin immunoreactivity in the lung, liver, and lymphoid tissues of manatees collected during a 1996 mortality event, suggesting that brevetoxins, at least in part, can be absorbed by mammals via the inhalation of aerosolized toxins [@pone.0017394-Bossart1], [@pone.0017394-Mase1]. Although a comprehensive understanding of brevetoxin trophic transfer is lacking, it is clear that finfish [@pone.0017394-Naar1], [@pone.0017394-Fire1], [@pone.0017394-Flewelling1] and certain types of seagrasses (i.e., *Thalassia testudinum*) can accumulate or be associated with brevetoxins and play a primary role in brevetoxin-induced marine mammal Unusual Mortality Events (UMEs) [@pone.0017394-Flewelling1], [@pone.0017394-Landsberg1]. Members of the diatom genus *Pseudo-nitzschia* are associated with production of domoic acid (DA), a neurotoxin that can cause amnesic shellfish poisoning (ASP) in humans [@pone.0017394-Wright1] and large-scale mortality of sea birds [@pone.0017394-Work1], pinnipeds [@pone.0017394-Scholin1], [@pone.0017394-Lefebvre1], and cetaceans [@pone.0017394-VanDolah2]. DA is an analog of the neurotransmitter glutamate, and a partial agonist that binds with high affinity to kainate receptors and intermediate affinity to α-amino-3-hydroxyl-5-methyl-4-isoxazole-propionate (AMPA) glutamate receptor subunits [@pone.0017394-Hampson1]. Trophic transfer of DA via zooplankton into higher organisms has been well documented in krill [@pone.0017394-Bargu1], shellfish [@pone.0017394-Blanco1], sand crabs [@pone.0017394-Ferdin1], and fish [@pone.0017394-Lefebvre1]. In the Texas region of the Gulf of Mexico, the presence of DA was first reported in phytoplankton in 1989 [@pone.0017394-Dickey1] with *Nitzschia pungens* f. *multiseries* (syn. *P. multiseries*) identified as the putative producer of DA (2.1 pg DA/cell) [@pone.0017394-Dickey1], [@pone.0017394-Fryxell1]. A consortium of *Pseudo-nitzschia* species was subsequently identified (*P. pseudodelicatissima*, *P. delicatissima*, *P. multiseries*, *P. pungens*, *P. subfraudulenta*) in the nearby Louisiana coastal waters [@pone.0017394-Parsons1]. The most dominant species, *P. pseudodelicatissima*, was shown to produce DA *in situ* (up to 4.82 pg/cell) [@pone.0017394-Parsons1] and in culture (up to 36 fg DA/cell) [@pone.0017394-Pan1]; levels that are not unlike *Pseudo-nitzschia* isolates collected from other coastal regions susceptible to DA-related UMEs and/or ASP events. A 'hot spot' of *Pseudo-nitzschia* spp. abundance has been identified in the northern Gulf of Mexico, near coastal Louisiana, near Mobile Bay, Alabama, and near Perdido Bay, Florida, that appears to be highly influenced by high nutrient, submarine waters [@pone.0017394-Liefer1]. As top predators, marine mammals such as bottlenose dolphins can serve as important sentinels for coastal environmental health [@pone.0017394-Wells1], [@pone.0017394-Schwacke1]. In the Gulf of Mexico, studies of the effects of *K. brevis* and brevetoxins on dolphins have historically focused on brevetoxin exposure and accumulation in carcasses recovered during UMEs [@pone.0017394-Mase1], [@pone.0017394-Flewelling1]. While data from carcasses are useful for determining relative tissue distributions and estimating amounts of toxin necessary to cause mortality, samples from live animals provide more accurate animal data (i.e., toxin levels, blood parameters, age class, sex, etc.) in conjunction with environmental (i.e., HAB species, fish collections) and spatiotemporal data. As such, animal data from live dolphins can provide insight into documenting non-lethal acute and/or chronic exposure and the effects of these exposures on a variety of other 'health' determinants not available from deceased animals. Dolphin health assessments in Sarasota Bay, Florida have been ongoing since the 1980s. Because they have involved sampling members of a long-term, year-round resident community of dolphins observed since 1970 and spanning at least five generations, these health assessments have provided a unique opportunity to monitor long-term trends of a dolphin population [@pone.0017394-Wells2]. Samples obtained between 2003 and 2005 have shown consistent exposure to brevetoxin when *K. brevis* was present in the surrounding water [@pone.0017394-Fire2]. The major route of exposure of these animals to brevetoxin appeared to be via consumption of finfish such as pinfish (*Lagodon rhomboides*), pigfish (*Orthopristis chrysoptera*), striped mullet (*Mugil cephalus*), and spot (*Leiostomus xanthurus*) [@pone.0017394-Fire1], which are among the primary prey items for bottlenose dolphins in the Sarasota Bay region [@pone.0017394-Barros1], [@pone.0017394-Barros2]. Although *K. brevis* blooms have been shown to affect fish population diversity and community structure [@pone.0017394-Gannon1], detectable levels of brevetoxin have been identified in shellfish and finfish for up to one year following a *K. brevis* bloom [@pone.0017394-Naar1], [@pone.0017394-Fire1], [@pone.0017394-Plakas1]. In contrast to studies involving brevetoxin, there have been no studies examining the presence or effects of DA or DA-producing *Pseudo-nitzschia* species on the dolphin population in Sarasota Bay. The primary objective of this study was to assess the exposure of live dolphins from the Sarasota Bay area to both brevetoxin and DA. Dolphin tissue/fluid and environmental samples were obtained from an intensive sampling effort in May 2008 and archived dolphin samples were retrospectively assessed back to June 2000. These data have been used to determine the extent of exposure of a marine mammal species to multiple algal toxins and assess putative health effects that may have negative impacts on this population. Methods {#s2} ======= Ethical treatment of animals {#s2a} ---------------------------- This study was carried out in strict accordance with the U.S. Marine Mammal Protection Act. Protocols for the dolphin capture-release and tagging were conducted under National Marine Fisheries Service Scientific Research Permits No. 522-1569 and No. 522-1785 issued to RSW. Research conducted under these permits was approved through the Mote Marine Laboratory annual Institutional Animal Care and Use Committee (IACUC) reviews. Phytoplankton cell abundance {#s2b} ---------------------------- Multiple-year *Karenia brevis* cell counts were determined on a nearly daily basis at two designated sampling sites in the waters adjacent to Mote Marine Laboratory (Bay Dock: lat./long. 27.33253/-82.57783 and New Pass: lat./long. 27.33382/-82.57911) ([Figure 1](#pone-0017394-g001){ref-type="fig"}) and average cell counts served as proxy indicators of HAB activity within the bay. Additional water samples were collected for cell counts adjacent to various dolphin capture-release sites in 2004 (February), 2008, and 2009. For all samples, the direct microscopic cell enumeration method used had a detection limit of 1000 cells/L. Prior to the May 2008 dolphin health assessment, much of the data for *Pseudo-nitzschia* spp. cell abundance were not quantified but limited to presence or absence only. During the May 2008 and 2009 dolphin health assessments phytoplankton were collected from surface waters either directly into sample collection tubes or using a plankton net (model 9100-10, Sea-Gear Corporation, Melbourne, FL, USA) with a 10 µm mesh size. All plankton samples were preserved with ∼1% acidified Lugol\'s iodine solution. Concentrated net tow samples and whole water samples concentrated using Utermöhl\'s sedimentation chambers [@pone.0017394-Utermhl1] were microscopically observed using an Olympus BX-51 (Center Valley, PA, USA) microscope with differential interference contrast (DIC). Samples with potential HAB species were positively confirmed using scanning electron microscopy (SEM) where samples were prepared using Simonsen\'s method for cleaning diatom frustules [@pone.0017394-Hasle1]. SEM samples were prepared following the method outlined in Morton [@pone.0017394-Morton1] and examined with a JEOL 5600LV SEM (Tokyo, Japan) operated at 15 kV. ::: {#pone-0017394-g001 .fig} 10.1371/journal.pone.0017394.g001 Figure 1 ::: {.caption} ###### Locations of live captured bottlenose dolphins (*Tursiops truncatus*) in Sarasota Bay, Florida, USA. Shown are (A) the captured locations for all sampled animals between 2000 and 2009 and (B) the specific 2008 and 2009 set sites where additional water sampling was also performed. :::  ::: Benthic sediment and snail collections {#s2c} -------------------------------------- To assess possible benthic routes of toxin trophic transfer, snails and sediment samples were collected from several health assessment sites using a 25 µm Fisher Scientific standard test sieve. Approximately 500 g of the upper 5 cm of the sediment layer was collected and drained of seawater, and whole snails (0.7--1.6 mm shell length) contained in this sediment layer were collected and stored frozen (−20°C) in 50 mL polypropylene centrifuge tubes. Fifty grams of the remaining sediment was collected and stored in similar manner. Sediment samples were subsequently centrifuged at 2500× *g* and again drained of seawater prior to toxin extraction. Fish collections {#s2d} ---------------- In 2008, various fish species were collected by cast net in the study area on 5/1/2008 (set site 1, lat./long. 27.50841/-82.69359) and 5/7/2008 (Chicago Zoological Society; lat./long. 27.33160/-82.57790) (see [Figure 1B](#pone-0017394-g001){ref-type="fig"}). In 2009, multiple fish species were collected by cast net in the study area on 5/5/2009 (*ca.* set site 4, lat./long. 27.395157/-82.629472) and 5/7/2009 (*ca.* set sites 1 and 8, lat./long. 27.47256/-82.661086) (see [Figure 1B](#pone-0017394-g001){ref-type="fig"}). Whole fish were identified, sorted according to species, weighed, and frozen at −20°C. Whole viscera (all organs in intracoelomic cavity) were dissected, weighed, and extracted for biotoxins. Dolphin samples {#s2e} --------------- Dolphin samples (urine, feces, whole blood, serum, and gastric fluid) for this study were collected from individuals sampled during health assessments in June 2000, February 2004, June 2004, February 2005, June 2005, June 2006, May 2008, and May 2009 ([Figure 1A](#pone-0017394-g001){ref-type="fig"}) according the methods of Wells *et al*. [@pone.0017394-Wells1]. Briefly, dolphins were captured by encircling them with a 500 m×4 m seine net deployed by a local commercial fisherman from a net boat in shallow (\<2 m) waters. May 2008 and 2009 capture and release sites ("set sites") are illustrated in [Figure 1B](#pone-0017394-g001){ref-type="fig"}. A team of more than 50 biologists, veterinarians, and trained dolphin handlers operating from up to eight other small boats was involved to insure the safe handling of the animals. Once secured, the dolphins were transferred one by one onto the padded, shaded deck of a 9 m veterinary examination boat where a series of length and girth measurements were taken, as well as weight. Whole blood (∼10 mL aliquot) was collected via venipuncture of the fluke vasculature and was (1) separated by centrifugation into plasma and serum for subsequent hematologic and biochemical analysis at a diagnostic laboratory as previously described [@pone.0017394-Wells1], [@pone.0017394-Schwacke2], [@pone.0017394-Schwacke3], and (2) spotted onto blood collection cards for toxin analysis as described by Maucher et al. [@pone.0017394-Maucher1]. Urine samples were collected via the urethra using a sterile catheter. Gastric samples were collected using a small tube inserted via the esophagus into the stomach, from which stomach contents were drained and stored. Feces samples were taken during processing using a sterile catheter. All samples were collected into sterile cryotubes or centrifuge tubes and frozen until biotoxin analysis was conducted. Due to the retrospective nature of this study, many pre-2008 samples were initially stored at −20°C prior to permanent storage at −80°C. Specific site and animal information for 2008 and 2009 are outlined in [Table 1](#pone-0017394-t001){ref-type="table"}. ::: {#pone-0017394-t001 .table-wrap} 10.1371/journal.pone.0017394.t001 Table 1 ::: {.caption} ###### Animal information and biotoxin tissue/fluid concentrations for the bottlenose dolphins (*Tursiops truncatus*) sampled during the May 2008 and 2009 health assessments in Sarasota Bay, Florida. ::: {#pone-0017394-t001-1} Animal ID (FB\#) Capture Date Set \# Sex Age Length (cm) Weight (kg) Brevetoxin Concentration (ng/g or ng/mL) Domoic Acid Concentration (ng/g or ng/mL) ------------------ -------------- -------- ----- ------ ------------- ------------- ------------------------------------------ ------------------------------------------- ------ ------ ------ ------ ------ ------ ------- ------- ------ ------ 133 5/1/2008 1 F 9 225 148 \<dl \<dl \<dl 5.2 \<dl \<dl \<dl \<dl 224 5/1/2008 1 M 6 213 120 \<dl \<dl \<dl 13.6 \<dl \<dl \<dl 17.6 41.5 \<dl 203 5/2/2008 2 F 8 217 135 \<dl \<dl \<dl 5.6 \<dl \<dl 1.8 \<dl 155 5/5/2008 3 F 18 240 171 \<dl \<dl \<dl 30.0 6.5 \<dl \<dl 1.6 9.4 \<dl 205 5/5/2008 3 F 2 183 72 \<dl \<dl \<dl 4.1 \<dl \<dl \<dl \<dl 197 5/5/2008 3 F 5 210 118 \<dl \<dl \<dl 3.7 \<dl \<dl \<dl \<dl 175 5/6/2008 5 F 17 na na \<dl \<dl \<dl \<dl 207 5/6/2008 5 F 3 190 98 \<dl \<dl \<dl 5.4 \<dl \<dl 4.9 13.8 209 5/8/2008 8 F 4 198 92 \<dl \<dl \<dl 4.5 \<dl \<dl 6.1 \<dl 11 5/9/2008 9 F 24 257 204 \<dl \<dl \<dl 13.6 6.7 \<dl \<dl 1.0 10.3 \<dl 250 5/9/2008 9 M 3 190 89 \<dl \<dl \<dl 3.5 10.5 \<dl \<dl 4.4 8.5 \<dl 178 5/9/2008 9 M 13 257 201 \<dl \<dl \<dl 12.3 4.6 \<dl \<dl \<dl \<dl \<dl 188 5/9/2008 9 M 12 236 173 \<dl \<dl \<dl 2.3 4.3 \<dl \<dl \<dl \<dl \<dl 151 5/3/2009 1 F 9 228 130 \<dl \<dl \<dl \<dl \<dl \<dl \<dl 1.6 37 \<dl 55 5/4/2009 4 F 23 250 176 \<dl \<dl \<dl 3 \<dl \<dl \<dl \<dl trace 4 \<dl \<dl 213 5/4/2009 4 F 2 187 71 \<dl \<dl \<dl \<dl \<dl \<dl \<dl \<dl 138 5/4/2009 4 M 17 265 240 \<dl \<dl \<dl 3 \<dl \<dl \<dl 3.6 11.5 \<dl 92 5/5/2009 6 M 21 249 193 \<dl \<dl \<dl 9 \<dl \<dl \<dl \<dl 4 \<dl 252 5/6/2009 7 M 3 216 107 \<dl \<dl \<dl \<dl \<dl \<dl \<dl trace trace \<dl 151 5/6/2009 8 F 9 na na \<dl \<dl \<dl \<dl \<dl \<dl \<dl 1.6 37 \<dl 141 5/6/2009 8 F 19 232 149 \<dl \<dl \<dl \<dl \<dl \<dl \<dl \<dl \<dl 6.5 \<dl \<dl 217 5/6/2009 8 F 2 186 76 \<dl \<dl \<dl \<dl 215 5/6/2009 8 F 9 225 na \<dl \<dl \<dl \<dl 219 5/7/2009 9 F 14 255 203 \<dl \<dl \<dl 5 \<dl \<dl \<dl 4.6 36 2.2 254 5/7/2009 9 M \>17 266 247 \<dl \<dl 9 \<dl \<dl 2.4 trace \<dl 125 5/7/2009 11 F 11 257 194 \<dl \<dl 32 \<dl \<dl \<dl 1.4 15 \<dl \<dl 256 5/7/2009 11 M 2 210 106 \<dl \<dl \<dl trace 198 5/7/2009 11 M 13 252 226 \<dl \<dl \<dl \<dl \<dl \<dl ::: Brevetoxin extraction and analysis {#s2f} ---------------------------------- Urine samples were centrifuged at 13000× *g* for 10 min at 25°C, supernatant removed, and 0.45 µm filtered prior to analysis. Fish viscera, dolphin feces and gastric samples were homogenized and extracted in acetone (3 volumes) four times, filtered via a 1 or 0.2 µm Pall syringe filter, dried under nitrogen gas, resuspended in 80% aqueous methanol (30 mL), twice solvent partitioned with hexane (30 mL), and the methanolic layer collected, dried under nitrogen gas and resuspended in 100% methanol (0.5 or 5 mL, depending on the weight of the original sample). Brevetoxins were extracted from blood cards using a phosphate buffered saline (94%)/methanol (6%) solvent (0.8 mL), followed by protein precipitation in acetonitrile (2.4 mL) [@pone.0017394-Maucher1]. Briefly, samples were centrifuged at 4000× *g* for 15 min at 4°C and the supernatant was collected. Plankton filter samples were collected by filtering ∼250 mL of seawater through Whatman GF/F filters and storing frozen at −20°C until analysis. Filters were twice extracted overnight with 2.5 mL methanol (100%) and vortexing with five short pulses [@pone.0017394-Roth1] prior to centrifugation for 30 sec at 12000× *g* through NanoSep MF (0.2 µm) spin columns. All extracts were stored at −20°C until analysis. The brevetoxin ELISA determined the presence of brevetoxin and brevetoxin-like compounds based on cross-reactivity with an antibody in a 96-well direct ELISA format [@pone.0017394-Maucher1]. In order to eliminate matrix effects, minimum assay dilutions for each sample type were: blood (1∶10), serum (1∶100), gastric fluid (1∶50), feces (1∶50), and fish viscera (1∶50). Sample calibration was performed using a brevetoxin-3 standard curve and non-linear regression analysis. The limits of detection (LOD) were: blood (0.2 ng/mL), serum (9 ng/mL), gastric fluid (2.7 ng/mL), feces (0.8 ng/g), and fish viscera (between 0.8 and 4.1 ng/g). The brevetoxin radioimmunoassay (RIA) determined the presence of brevetoxin and brevetoxin-like compounds based on a sheep antiserum prepared against a brevetoxin-2 conjugate [@pone.0017394-Woofter1], [@pone.0017394-Poli2], [@pone.0017394-Poli3]. The RIA measured the competition between radiolabeled brevetoxin-3 (^3^H-brevetoxin-3) and unknown samples for the anti-brevetoxin-2 antiserum. Antibodies were filtered onto 25 mm glass fiber filters and the radioactivity of each filter was measured to determine the amount of brevetoxin-3 equivalents in each sample. The LODs were: blood (2.3 ng/mL), gastric fluid (between 0.5 and 3.5 ng/mL), feces (between 0.5 and 3.5 ng/g), and fish viscera (between 0.7 and 3.5 ng/g). The brevetoxin receptor binding assay (RBA) measured competition between radiolabeled brevetoxin-3 (^3^H-brevetoxin-3) and unknown samples for the voltage-gated sodium channel in a rat brain crude membrane preparation, the pharmacological target of brevetoxins, to determine the total brevetoxin-3 equivalent activity of the sample. Details of this assay are described by Van Dolah *et al.* [@pone.0017394-VanDolah3]. The LOD for feces samples in this assay was ∼100 ng brevetoxin-3-equiv./g. The brevetoxin LC-MS/MS method measures the unambiguous structures of brevetoxin based on size, column retention and fragmentation patterns. Brevetoxin liquid chromatography (LC) separations were performed on a Luna C8(2) 150×2 mm column using an Agilent Technologies Model 1100 LC system. Mobile phase consisted of water (A) and acetonitrile (B), with 0.1% acetic acid additive. LC gradient: 2 min 35% B, linear gradient to 80% at 30 min, 95% B at 35 min, held at 95% B for 7 min, returned to 35% B at 43 min, and held for 7 min before the next injection. The mobile phase flow rate was 0.2 mL/min. The elutant from LC was analyzed by an Applied Biosystems/MDS Sciex 4000 QTRAP hybrid triple quadrupole/linear ion trap mass spectrometer equipped with a Turbo V™ source (Applied Biosystems, Foster City, CA, USA). The detection of brevetoxin congeners and metabolites by mass spectrometry was achieved by multiple reaction monitoring (MRM) and selected ion monitoring (SIM) [@pone.0017394-Plakas2], [@pone.0017394-Wang1]. The following brevetoxin congeners and their derivatives were analyzed by comparison to commercial and/or in house derivatized standards: brevetoxin-1, -2, -3, -7, -9, oxidized brevetoxin-2, open A-ring brevetoxin-2, -3, -7, oxidized brevetoxin-2, cysteine-brevetoxin-A(B) and its sulfoxide, open-A ring cysteine-brevetoxin-B, and glutathione-brevetoxin-A(B). S/N ratio was about 26 for brevetoxin-3 standard at 1 ng/mL, 8 for brevetoxin-7 at 5 ng/mL, 54 for open-A ring brevetoxin-3 at 5 ng/mL, 76 for cysteine-brevetoxin-B at 10 ng/mL, and 17 for cysteine-brevetoxin-B sulfoxide at 10 ng/mL. The LODs for brevetoxin in seawater were 0.01 µg/L and 1 ng/mL urine. Domoic acid extraction and analysis {#s2g} ----------------------------------- Dolphin urine and serum samples were centrifuged at 12000× *g* through a 0.22 µm filter column prior to analysis. Blood card samples (100 µL whole blood) were extracted for DA using a water (60%)/methanol (40%) solution (2.0 ml) according to the methods of Maucher and Ramsdell [@pone.0017394-Maucher2]. Briefly, samples were sonicated then extracted for 12 h at 4°C. Extracts were dried under nitrogen gas and resuspended in 10 mM PBS/0.05 Tween (100 uL). Gastric and feces samples were diluted 1∶4 with aqueous methanol (50%) prior to centrifugation for 10 min at 3000× *g* followed by filtration through a Nanosep (0.45 µm) syringe filter. Fish viscera were dissected from whole fish, homogenized and diluted 1∶4 with aqueous methanol (50%) prior to centrifugation for 10 min at 3000× *g* followed by filtration through a glass fiber filter (1 µm) followed by a Nanosep (0.45 µm) syringe filter. Fish were not allowed to thaw before the whole viscera was removed, in order to avoid leakage of DA from the digestive tract into the adjacent tissues [@pone.0017394-Lefebvre2]. Plankton filter samples were processed by filtering ∼250 mL of surface seawater through Watman GF/F filters and storing frozen at −20°C until extraction. Filters were extracted with 5 mL aqueous methanol (10%) prior to centrifugation for 1 min at 6000× *g* followed by filtration through a 0.22 µm syringe filter. All extracts were stored at −20°C until analysis. A direct competitive DA enzyme-linked immunosorbent assay (ELISA) from Biosense Laboratories (Bergen, Norway) was used to screen the dolphin serum, urine, and blood card extracts [@pone.0017394-Maucher2]. This assay measures DA in a sample through its competition with DA coated onto microplate wells for anti-DA antibodies in solution. Blood card extracts were diluted 1∶10, serum samples were diluted 1∶100, and urine diluted 1∶200 prior to analysis. The limit of detection (LOD) of this assay was 0.12 ng/mL blood, 1.2 ng/mL serum, and 1.0 ng/mL urine. Selected samples were analyzed for the presence of DA using tandem mass spectrometry coupled with liquid chromatographic separation (LC-MS/MS) [@pone.0017394-Wang2]. This method utilized reverse phase chromatography, using an Agilent 1100 HPLC coupled to an Applied Biosystems/MDS Sciex API-4000 triple quadrupole mass spectrometer equipped with a Turbo V™ source (Applied Biosystems, Foster City, CA, USA). Chromatographic separation was performed on a Luna C18(2), 5 µm, 150×2 mm column (Phenomenex, Torrance, CA, USA). Mobile phase consisted of water and acetonitrile (ACN) in a binary system, with 0.1% formic acid as an additive. The elution gradient was: 3 min of 5% ACN, with a linear gradient to 40% ACN at 16 min, 95% ACN at 18 min, held for 5 min, then returned to initial conditions at 24 min and held for 5 min before the next injection. To reduce mass spectrometer contamination, a diverter valve was used to switch the LC eluant to the waste container except for the 6 minute window of LC eluant bracketing the retention time for DA that was sent to the MS. Retention time of DA in samples was determined based on the retention time observed with a certified reference standard (NRC Canada, Halifax, Canada). The detection of domoic acid by MS was achieved by multiple reaction monitoring (MRM) method with Turbo ion spray interface in positive ion mode. Four MRM transitions from protonated domoic acid were monitored: *m/z* 312→266, *m/z* 312→248, *m/z* 312→193, and *m/z* 312→161. The limits of quantification (LOQ) for this method were ∼0.5 ng/mL urine, 2 ng/mL gastric fluid, 0.01 µg/L seawater, and 2 ng/g dolphin feces or fish viscera, with a signal to noise ratio above ten. Statistical Analysis {#s2h} -------------------- Statistical analysis was conducted to examine the relationship between toxin (brevetoxin and DA) concentration and 16 hematologic and biochemical parameters. Hematology and blood chemistry data were categorized into 3 panels of interest: Red blood cell (RBC) indices, white blood cell (WBC) differential, and liver and/or kidney associated enzymes. A generalized linear model (GLM) was applied to examine the effect of toxin concentration for each of the three panels. Each panel was first analyzed using a multivariate GLM, and when the multivariate F-statistic (Wilks lambda) was significant (p\<0.05), univariate GLMs (F-test) were conducted independently for each health parameter in the panel. Age was included as a covariate for the WBC differential and liver/kidney enzymes; sex was included as a covariate for RBC indices. The inclusion of covariates was based on previous examinations of factors influencing hematology and blood chemistry parameters [@pone.0017394-Schwacke2]. P-plots were examined to assess normality of residuals, and variables were log-transformed when necessary (specifically for neutrophil, monocyte, lymphocyte and eosinophil counts) to meet model assumptions. Toxin concentrations from both urine and feces were included but in separate analyses. Results {#s3} ======= In the Sarasota Bay region between 2000 and 2009, *K. brevis* cell densities varied between below the detection limit (1000 cells/L) to a maximum of 8×10^7^ cells/L ([Figure 2](#pone-0017394-g002){ref-type="fig"}). In each of these years, with the exception of 2008 and 2009, *K. brevis* cell densities exceeded 10^5^ cells/L. On three occasions, dolphin health assessments corresponded to periods in the bay when *K. brevis* cell counts were greater than 10^5^ cells/L (Feb 2004, Feb 2005, June 2005), whereas the remaining four health assessments occurred during periods when *K. brevis* was below the detection limit (June 2000, June 2004, June 2006, May 2008, May 2009). Unfortunately, *K. brevis* cell counts were not obtained at each set site during the health assessments. In May 2008 and 2009, seawater samples were analyzed for brevetoxin but all samples were below the limit of detection (\<dl) of ∼0.01 µg/L (data not shown). ::: {#pone-0017394-g002 .fig} 10.1371/journal.pone.0017394.g002 Figure 2 ::: {.caption} ###### Representative *K. brevis* and *Pseudo-nitzschia* sp. cell count data for 2000 to 2009 in Sarasota Bay, Florida USA. Shaded regions represent the times during which dolphin health assessments occurred and samples were obtained for analyses. :::  ::: Prior to 2008, there were essentially no data available in the Sarasota Bay area on diatom presence, specifically *Pseudo-nitzschia* spp. However, in mid-April just prior to the 2008 dolphin health assessment, a *Pseudo-nitzschia* bloom occurred in the southern region of Sarasota Bay (Bay Dock, New Pass, Longboat Key, and Lido Key) ([Figure 2](#pone-0017394-g002){ref-type="fig"}) close to Mote Marine Laboratory with at least one sample containing over 3×10^6^ cells/L (Valeriy Palubok, pers. comm.). Not constrained to just the southern region of the bay, water samples from each set site all contained *Pseudo-nitzschia* densities in excess of 10^4^ cells/L ([Figure 3A](#pone-0017394-g003){ref-type="fig"}). These same water samples were used to positively identify the *Pseudo-nitzschia* cells as *P. pseudodelicatissima* by SEM ([Figures 3C and 3D](#pone-0017394-g003){ref-type="fig"}). Water samples containing *P. pseudodelicatissima* were confirmed in set sites 2, 5, and 8. Other harmful algal bloom species identified were *Chaetoceros* spp. ([Figure 3C](#pone-0017394-g003){ref-type="fig"}) and *Prorocentrum compressum* ([Figure 3E](#pone-0017394-g003){ref-type="fig"}). All seawater samples (n = 6) collected from each set site ([Figure 1A](#pone-0017394-g001){ref-type="fig"}) during the May 2008 health assessment were confirmed positive for DA with a mean particulate concentration of 101.7±40.9 ng/L (mean ± SE; range: 16--289 ng/L) ([Figure 3A](#pone-0017394-g003){ref-type="fig"}). These corresponded to cellular DA quotas of 5.15±3.17 pg/cell (mean ± SE; range: 0.54--17.57 pg/cell). Samples from set site 8, located at the mouth of Palma Sola Bay just north of Sarasota Bay, had the highest DA particulate and cell quota concentrations. Samples from set site 9 were compromised during the cell count procedure and not included. ::: {#pone-0017394-g003 .fig} 10.1371/journal.pone.0017394.g003 Figure 3 ::: {.caption} ###### *Pseudo-nitzschia* spp. and domoic acid in Sarasota Bay, Florida, USA. Cell densities of *Pseudo-nitzschia* spp. and domoic acid water concentrations for various (A) 2008 and (B) 2009 set site locations. Scanning electron micrographs of three samples illustrating the presence of potential HAB species: C) A mixed assemblage of phytoplankton cells including *Chatocerous* spp. and *Pseudo-nitzschia* spp. from 2008 set site 8, D) *P. pseudodelicatissima* cells from 2008 sets sites 2, 5, and 8, and E) *Prorocentrum compressum* from 2008 set site 2. Note: Presence of *P. pseudodelicatissima* was also confirmed in 2008 set sites 8 and 2. 'na' represents 'not available' and '\<dl' represents 'below detection limit'. :::  ::: Throughout much of 2009, low to moderate densities of *Pseudo-nitzschia* cells were observed at New Pass and Bay Dock but not quantified ([Figure 2](#pone-0017394-g002){ref-type="fig"}). During the May 2009 health assessment, *Pseudo-nitzschia* cells were observed at set sites 1, 6, 8, and 9 ranging in densities between 6.6×10^4^ to 2.2×10^6^ cells/L with set sites 1, 6, and 8 in the northern region of the bay and set site 9 in the southern end ([Figure 3B](#pone-0017394-g003){ref-type="fig"}). Mean particulate domoic acid concentrations were 80±65.7 ng/L (mean ± SE; range: \<dl-405 ng/L) with three out of the six sites testing positive for DA (set sites 1, 8, and 9) ([Figure 3B](#pone-0017394-g003){ref-type="fig"}). Samples from set sites 1 and 8 were taken three days apart in Palma Sola Bay (northern region of the greater Sarasota Bay; see [Figure 1B](#pone-0017394-g001){ref-type="fig"}), whereas set site sample 9 was from the southern region of Sarasota Bay; suggesting patchy domoic acid distributions. These toxin concentrations result in DA cellular quotas of 0.42±0.23 pg/cell (mean ± range). At set sites 6 and 8, cells densities of *Pseudo-nitzschia* were in excess of 6.6×10^4^ cell/L but DA concentrations were \<LOD. Viscera from seven different fish species (striped mullet, pigfish, pinfish, striped mojarra, scaled sardines, mangrove snapper, sheepshead) collected in the Sarasota region in 2008 and/or 2009 were analyzed for both brevetoxin and DA concentrations ([Table 2](#pone-0017394-t002){ref-type="table"}). Based on mean weight and calculated or measured standard length, all fish were determined to be adults ([www.fishbase.org](http://www.fishbase.org)). In 2008, with the exception of the striped mojarra that were collected by the Chicago Zoological Society, all other fish were collected at set site 1 ([Figure 1B](#pone-0017394-g001){ref-type="fig"}). Mean brevetoxin concentrations for these species (striped mullet, pigfish, pinfish, striped mojarra, scaled sardines) ranged between 22 and 46 ng/g viscera and mean DA concentrations ranged between 39 and 440 ng/g viscera. Fish specimens that were of sufficient size (i.e., striped mullet, pinfish, striped mojarra; n = 15 total) for analysis of both toxins from the same homogenized tissues all tested positive for both brevetoxin (8--60 µg/g) and DA (9--171 µg/g). In 2009, striped mullet, pinfish, mangrove snapper, and sheepshead were caught over a two-day period in Sarasota Bay. Brevetoxin concentrations ranged between 3 and \>20.3 ng/g viscera and DA concentrations ranged between \<dl and 150.8 ng/g viscera. Although all individual fish tested positive for brevetoxin by ELISA and/or LC-MS/MS, only the striped mullet concurrently tested positive for DA as well. ::: {#pone-0017394-t002 .table-wrap} 10.1371/journal.pone.0017394.t002 Table 2 ::: {.caption} ###### Concentrations of brevetoxin and domoic acid found in viscera of various fish species in Sarasota Bay, Florida, USA during the May 2008 and May 2009 dolphin health assessments. ::: {#pone-0017394-t002-2} Capture Date Common Name Genus species Total No. Fish Total Fish Weight (g) Visceral Weight (g) Brevetoxin (ng/g viscera) Domoic acid (ng/g viscera) -------------- ------------------ ------------------------------- ---------------- ----------------------- --------------------- --------------------------- ---------------------------- 5/1/2008 Striped mullet *Mugil cephalus* 5 386±49 72±15 22±4 (n = 5\*) 90±29 (n = 5\*) 5/1/2008 Pinfish *Lagodon rhomboides* 5 37±4 4.9±0.5 24±5 (n = 5\*) 39±16 (n = 5\*) 5/1/2008 Pigfish *Orthopristis chrysoptera* 7 4.3±0.2 0.28±0.03 43±20 (n = 2) 76±21 (n = 5) 5/7/2008 Striped mojarra *Eugerres plumieri* 5 151±42 18±4 45±6 (n = 5\*) 65±15 (n = 5\*) 5/1/2008 Scaled sardines *Harengula jaguana* 10 14±0.7 3.5±0.6 46±4 (n = 5) 440±92 (n = 5) 5/5/2009 Striped mullet *Mugil cephalus* 3 619±65.4 72.7±5.0 all \>20.3 (n = 3\*) 54±4 (n = 3\*) 5/5/2009 Pinfish *Lagodon rhomboides* 3 86±1.9 9.3±1.2 all \>7 (n = 3\*) \<dl (n = 3\*) 5/7/2009 Mangrove snapper *Lutjanus griseus* 2 147±5.5 7±0.2 3±1 (n = 2\*) \<dl (n = 2\*) 5/5/2009 Sheepshead *Archosargus probatocephalus* 1 562 45.8 10 (n = 1\*) \<dl (n = 1\*) Note: Data shown are mean ± SE (or ± range where n = 2). An asterisk (\*) represents the same fish specimens used for both toxin analyses. ::: During the 2009 health assessment event, benthic samples were taken for biotoxin analysis. All sediment samples (n = 3 total) collected from set sites 1, 7, and 9 were negative for DA, but positive for brevetoxin (by ELISA) ranging in concentration from 1--3 ng/g ([Figure 4](#pone-0017394-g004){ref-type="fig"}). At two sites, brevetoxin was identified at low concentrations in benthic snails (3--6 ng/g whole tissue) (ELISA with LC-MS/MS confirmation) with the concurrent presence of domoic acid (range: 3--13 ng/g tissue). ::: {#pone-0017394-g004 .fig} 10.1371/journal.pone.0017394.g004 Figure 4 ::: {.caption} ###### Domoic acid and brevetoxin in Sarasota Bay, FL sediment and benthic snails. Various samples for biotoxin analysis were collected in May 2009 that correspond to dolphin sampling set sites ("Sets"). Note: 'na' represents 'not available' and '\<dl' represents 'below detection limit'. :::  ::: During the May 2008 health assessment, several dolphin samples (blood, serum, urine, feces, gastric fluid) from 13 animals were obtained for biotoxin analysis. All feces (100%; 6/6) and most gastric fluid (92%; 11/12) samples tested positive for brevetoxin (2.3--30 ng/g or ng/mL) and a majority of urine (58%; 7/12) and feces (67%; 4/6) samples tested positive for DA (up to 17.6 ng/mL and 41.5 ng/g; respectively) ([Table 1](#pone-0017394-t001){ref-type="table"}, [Figure 5](#pone-0017394-g005){ref-type="fig"}, and [Table S1](#pone.0017394.s001){ref-type="supplementary-material"}). Feces samples contained the highest concentrations of both biotoxins (brevetoxin: 2.3--30 ng/g; DA: \<dl-41.5 ng/g) relative to other samples. Only one gastric fluid sample tested positive for DA (8%; 1/12) and all blood and serum samples were below the detection limit (\<dl) for both toxins. With the exception of FB175 where only blood and serum samples were available, 7 of the remaining 12 dolphins (58%) concurrently contained both brevetoxin and DA in at least one sample type (urine, feces, and/or gastric fluid). ::: {#pone-0017394-g005 .fig} 10.1371/journal.pone.0017394.g005 Figure 5 ::: {.caption} ###### Biotoxin concentrations in urine, blood, and/or feces of Sarasota Bay, Florida dolphins between 2000--2009. \(A) Brevetoxin was analyzed by RIA, ELISA, RBA, and/or LC/MS, and (B) domoic acid was analyzed by ELISA and/or LC/MS. Brevetoxin values are reported in ng PbTx-3 equiv./g or/mL. Data are median, quartile, and minimum/maximum values indicated by the midline, box, and whisker lines; respectively. Raw data are available in [Table S1](#pone.0017394.s001){ref-type="supplementary-material"} and [Table S2](#pone.0017394.s002){ref-type="supplementary-material"}. Note: 'na' denotes samples not available. :::  ::: During the May 2009 health assessment, several dolphin tissue/fluid samples (blood, serum, urine, feces, gastric fluid, milk) from 14 animals were obtained for biotoxin analysis. Many feces (67%; 6/9) samples tested positive for brevetoxin (\<dl-32 ng/g) and a majority of urine (67%; 8/12) and feces (100%; 9/9) samples tested positive for DA (up to 4.6 ng/mL and 36 ng/g; respectively) ([Table 1](#pone-0017394-t001){ref-type="table"}, [Figure 5](#pone-0017394-g005){ref-type="fig"}, and [Table S1](#pone.0017394.s001){ref-type="supplementary-material"}). Only one gastric fluid sample tested positive for DA (9%; 1/11) and all blood, serum and milk samples were below the detection limit (\<dl) for both toxins. Six out of the 14 dolphins (43%) concurrently contained both brevetoxin and DA in at least one sample type (most often feces samples). Archived samples from dolphin health assessments in the Sarasota Bay area dating back to June 2000 (6 additional health assessments) were obtained and analyzed for both brevetoxin and DA ([Figure 5](#pone-0017394-g005){ref-type="fig"}, [Table S1](#pone.0017394.s001){ref-type="supplementary-material"}, and [Table S2](#pone.0017394.s002){ref-type="supplementary-material"}). All blood and serum samples were negative for brevetoxin and DA except for several brevetoxin-positive blood samples from February 2004. Nine out of 17 blood samples from February 2004 were positive for brevetoxin (53%; up to 1.3 ng/mL). Brevetoxin was detected in urine of animals sampled in February 2004 (1/1), February 2005 (4/5) and June 2006 (12/13) at concentrations up to 0.53, 7.1, 89.7 ng/mL; respectively, but was not detected in urine from animals in June 2000 (0/1) and June 2004 (0/9). Similar to May 2008, feces samples from June 2006 contained some of the highest brevetoxin concentrations observed (up to 101 ng brevetoxin-3 equiv./g). DA was detected in many of the urine samples collected in June 2000 (76%; 13/17), June 2004 (75%; 9/12), and June 2005 (83%; 5/6) up to concentrations of 12.3, 21.8, 16.7 ng/mL; respectively, but was not detected in urine samples from February 2005 (0/4) and June 2006 (0/15). Out of the 118 animals tested for brevetoxin over the ten-year period (including individual animals that were repeat samplings in different years), 43 were positive for brevetoxin (36%) whereas 44 out of 83 (53%) animals tested for DA were positive. Three animals in particular were exposed to brevetoxin on two separate occasions (FB118, FB133 and FB188) while two individuals were exposed to DA on two separate occasions (FB11 and FB92) with a third animal exposed to DA three times (FB155) ([Table S2](#pone.0017394.s002){ref-type="supplementary-material"}). Of the eight health assessments analyzed within this study, five of these health assessments had several animals that tested positive for brevetoxin and five of these health assessments had several animals that tested positive for DA ([Figure 5](#pone-0017394-g005){ref-type="fig"}, [Table S1](#pone.0017394.s001){ref-type="supplementary-material"}, and [Table S2](#pone.0017394.s002){ref-type="supplementary-material"}). Over the course of this study, 13 animals (all from 2008 and 2009) were exposed to both brevetoxin and DA concurrently. Multivariate GLMs for both urine and fecal DA concentration ([Table 3](#pone-0017394-t003){ref-type="table"}) showed a significant effect (p\<0.05) on WBC differential and specifically on eosinophil count (urine p = 0.031, fecal p\<0.001) ([Figure 6](#pone-0017394-g006){ref-type="fig"}). In addition, the effect of DA concentration in feces was significant for total WBC (p = 0.001). The effect of DA concentration in urine was significant for two RBC indices: red blood cell count (p = 0.002) and mean corpuscular volume (p\<0.001). However, DA concentration in feces did not have a significant effect on RBC indices (p = 0.838). Brevetoxin concentration did not have a significant effect on any of the health panel indices. ::: {#pone-0017394-g006 .fig} 10.1371/journal.pone.0017394.g006 Figure 6 ::: {.caption} ###### Domoic acid concentration in feces versus eosinophil count. Dashed green lines represent published reference thresholds [@pone.0017394-Schwacke2]. Regression analysis indicated a statistically significant increase in eosinophil count (p\<0.001), although none of the measured values were outside of established reference thresholds. :::  ::: ::: {#pone-0017394-t003 .table-wrap} 10.1371/journal.pone.0017394.t003 Table 3 ::: {.caption} ###### Relationships between toxin concentrations and various health parameters examined by p-values generated from generalized linear model (GLM). ::: {#pone-0017394-t003-3} Health Parameters Domoic Acid Brevetoxin -------------------------------------------- ------------- ------------ ----------- ----------- **RBC Indices; Covariate: Sex** **0.008** **0.838** **0.114** **0.229** Red blood cells (RBCs) 0.002 na na na Hematocrit (Hct) 0.245 na na na Hemoglobin (Hgb) 0.355 na na na Mean corpuscular volume (MCV) \<0.001 na na na **WBC Differential; Covariate: Age** **\<0.001** **0.011** **0.906** **0.773** Total white blood cells (WBCs) 0.113 0.001 na na Neutrophils 0.768 0.386 na na Lymphocytes 0.612 0.127 na na Monocytes 0.651 0.542 na na Eosinophils 0.031 \<0.001 na na **Liver & Kidney Enzymes; Covariate: Age** **0.733** **0.191** **0.055** **0.773** Blood urea nitrogen (BUN) na na na na Creatinine na na na na Alanine aminotransferase (ALT) na na na na Aspartate aminotransferase (AST) na na na na Gamma-glutamyl transferase (GGT) na na na na Lactate dehydrogenase (LDH) na na na na Creatine kinase (CK) na na na na Note: Bolded data indicate results from multivariate F-test (Wilks lambda). When multivariate p-value was less than 0.05, univariate F-tests were conducted for each dependent (health) variable. 'na' denotes not available. ::: Discussion {#s4} ========== In this study we documented the repeated exposure of bottlenose dolphins in southwest Florida to two marine algal toxins, brevetoxin and DA, over a ten year period (2000--2009) and measured toxin concentrations in whole water samples as well as in benthic samples and potential prey fish species that likely serve as important vectors for toxin trophic transfer in this region. We are the first to positively identify a known DA-producing species of *Pseudo-nitzschia* (*P. pseudodelicatissima*) in Sarasota Bay, FL, and to our knowledge, this is the first published report to describe concurrent exposure of marine mammals to multiple algal toxins in this region. *Karenia brevis* blooms and brevetoxin {#s4a} -------------------------------------- *K. brevis* blooms are a nearly annual event in the southwestern region of Florida. Since *K. brevis* blooms are often patchy in their spatial distribution, cell count data from two frequently sampled sites (New Pass and Bay Dock) located at the southern end of Sarasota Bay were used as a proxy for bloom activity within the area. Although these data correlate with the cell count data in the other parts of the bay, they cannot be used to accurately assess toxin exposure to the dolphins or their prey items. During the course of this study, every year except for 2008 and 2009 had *K. brevis* cell densities greater than 10^5^ cells/L. Most notable were 2005 and August through November 2006 when cell densities exceeded 10^5^ cells/L for months at a time. During this particular period there was an ongoing multi-species UME (dolphins, manatees and sea turtles) that began in September 2005 and continued to July 2006 ([www.nmfs.noaa.gov/pr/health/mmume/](http://www.nmfs.noaa.gov/pr/health/mmume/)). During the eight dolphin health assessments in southwestern Florida that were used in this study, three of them occurred when *K. brevis* cell densities were considered 'bloom' conditions (≥10^5^ cells/L), whereas during the remaining five assessments *K. brevis* cell densities were at levels below detection (\<10^3^ cells/L). The exposure of marine mammals from the Sarasota Bay region to brevetoxin is not unusual or unexpected considering the intense *K. brevis* blooms that have occurred in the area on a nearly annual basis. In total, over a third of the animals (43 out of 118; 36%) collected between 2000 and 2009 were brevetoxin-positive, at concentrations similar to those published by Fire *et al.* [@pone.0017394-Fire2]. The occurrence of brevetoxin in dolphin samples often correlated directly with the presence (February 2004, February 2005) or absence (June 2000, June 2004) of *K. brevis*. However, there were three occasions where low to moderate levels of brevetoxin were observed in the animals (June 2006, May 2008, May 2009) even though there was an absence of *K. brevis*. In fact, for each of these occasions it had been over 4 months since blooms of *K. brevis* had been detected in the area. This was in contrast to a previous study of Sarasota dolphins that consistently found an absence of brevetoxin in live dolphins during non-bloom conditions [@pone.0017394-Fire3]. Nonetheless, some of the highest brevetoxin concentrations observed in this study were from animal samples collected in June 2006 (urine and feces), and the highest proportion of brevetoxin-positive animals were from May 2008 (92% animals positive) followed by May 2009 (43%). However, it should be noted that more sample types were collected and tested during the May 2008 assessment relative to the other assessments. Nonetheless, there appeared to be additional factors contributing to brevetoxin trophic transfer and bottlenose dolphin exposure beyond the presence or absence of a concurrent *K. brevis* bloom. Although some dolphin populations are migratory in nature, the population within Sarasota Bay is quite endemic. The current year-round resident population of dolphins in Sarasota Bay spans five generations and includes two individuals initially identified in 1970--71 and still routinely observed in the Sarasota Bay region [@pone.0017394-Wells3]. The distribution of resightings of Sarasota dolphins suggests long-term residency and a definitive localized home range to Sarasota Bay [@pone.0017394-Wells4]. Although the ichthyotoxic effects of brevetoxin are still being described [@pone.0017394-Steidinger2], [@pone.0017394-Naar1], [@pone.0017394-Davis1], [@pone.0017394-Baden2], it has been shown that blooms of *K. brevis* reduce fish density and species richness in Sarasota Bay [@pone.0017394-Gannon1]. Finfish compose the majority of the diet of bottlenose dolphins from the eastern Gulf of Mexico [@pone.0017394-Barros1], [@pone.0017394-Barros2] and have been postulated to play a role in brevetoxin trophic transfer to these mammals. Undigested fish collected from the stomachs of dolphins that died during the 2004 Florida Panhandle UME supported the link between many species of planktivorous fish (presumably feeding on *K. brevis*) and dolphin mortalities [@pone.0017394-Flewelling1]. A subsequent study in the same area has shown that over 40 species of live planktivorous, omnivorous, and piscivorous fish contained brevetoxin in their tissues; some for up to one year following the termination of a *K. brevis* bloom [@pone.0017394-Naar1]. Adult striped mullet (benthic omnivores that can feed on decaying plant and algal material) [@pone.0017394-Odum1] and scaled sardines (pelagic omnivores feeding at least partially on copepods) [@pone.0017394-Moyle1], [@pone.0017394-Motta1] contained some of the highest brevetoxin values, particularly in the liver and gastrointestinal contents [@pone.0017394-Naar1]. Similarly, viscera from many of the fish in Sarasota Bay that are known to be prey for resident dolphins [@pone.0017394-Barros1] contained toxin concentrations ranging between 68 and 190 ng brevetoxin-3 equiv./g in the absence of *K. brevis* [@pone.0017394-Fire1]. When *K. brevis* was present, brevetoxin concentrations were much higher (pinfish\>pigfish\>mullet), averaging between 81 and 1313 ng brevetoxin-3 equiv./g viscera. During controlled laboratory exposures of striped mullet to *K. brevis*, blood brevetoxin concentrations varied according to the *K. brevis*/brevetoxin exposure concentrations where maximal blood brevetoxin concentrations were ∼19 ng/mL at 8 to 12 hr post-exposure with a protracted retention time (T~1/2~) of over 11 days [@pone.0017394-Woofter2]. Studies subsequently performed with Atlantic menhaden (*Brevoortia tyrannus*) also demonstrated slow elimination rates with a half-life of 24 days [@pone.0017394-Hinton1]. Although in our study the sample size was relatively small (n = 1, 2 or 5 per fish species), these slow rates of elimination probably cannot account for the observation that brevetoxin was found in every fish species and each individual collected in May 2008 and May 2009, many months since the previous *K. brevis* bloom. As such, secondary routes of brevetoxin trophic transfer might have been involved in these protracted residency times of at least 5 months. Omnivorous and piscivorous fish may also accumulate brevetoxins via indirect routes of trophic transfer [@pone.0017394-Landsberg1]. As observed in St. Joseph Bay, Florida, non-planktivorous fish accumulate moderate to high concentrations of brevetoxin in their tissues, particularly the liver tissues of piscivorous fish such as red snapper (*Lutjanus campechanus*; up to 16483 ng brevetoxin-3 equiv./g) [@pone.0017394-Naar1]. One trophic route was demonstrated in laboratory studies where pinfish and croakers (*M. undulates*) accumulated brevetoxin via feeding on contaminated shellfish [@pone.0017394-Naar1]. Brevetoxin trophic transfer may also begin with herbivorous fish feeding on contaminated seagrass. In 2002, 34 manatees died due to brevetoxicosis during a UME in southwest Florida following consumption of brevetoxin-contaminated seagrass such as *Thalassia testudinum* [@pone.0017394-Flewelling1]. Organisms feeding on (i.e., manatees, sea turtles) (D. Fauquier, pers. comm.) living within (i.e., macrobenthic epifauna) [@pone.0017394-Virnstein1], or associated with (i.e., epiphytes, macroalgae) [@pone.0017394-Dawes1] seagrass may also be susceptible to brevetoxin exposure. In addition, a benthic brevetoxin sink [@pone.0017394-SekulaWood1] may play a role in toxin (re)cycling into benthic organisms or into the water column during resuspension events as moderate amounts of toxin can be found in coastal sediments at concentrations up to 9.7 ng brevetoxin/g dry sediment [@pone.0017394-Mendoza1]. This finding is supported by the present study, which observed similar levels in sediment (1--3 ng/g) and benthic snails (3--6 ng/g) providing evidence that benthic brevetoxin is biologically available. Based on the comparable concentrations of brevetoxin for each of the five fish species collected in May 2008 (22--46 ng brevetoxin equiv./g viscera), it appears that in the absence of any recent *K. brevis* bloom activity, these particular fish species pose an approximately equal degree of risk for bottlenose dolphins. This is unusual since the feeding behaviors of these fish species, all adults based on size, are quite distinct. Scaled sardines are the most likely to feed directly on *K. brevis*. Striped mullet feed on decaying plant detritus and sediment which may include epiphytic and benthic microalgae (such as dead or dying *K. brevis*), [@pone.0017394-Odum1]. Pigfish, pinfish, and striped mojarra are more likely to be exposed to brevetoxin via indirect routes such as feeding on crustaceans, gastropods and/or polychaetes [@pone.0017394-Moyle1], [@pone.0017394-Nelson1]. Pinfish in particular are well known to associate with seagrass beds, and much of their diet can be composed of vegetation [@pone.0017394-Darcy1]. Interestingly, scaled sardines as well as other members of the clupeidae family (i.e., herrings, shads, sardines, anchovies, menhadens) were shown to be positively associated with *K. brevis* density while all other guilds significantly decreased during *K. brevis* bloom events in Sarasota Bay making clupeids potentially an even more vital dietary prey item for bottlenose dolphins as well as an effective vector for brevetoxin exposure [@pone.0017394-Gannon1]. The potential importance of striped mullet as a source for biotoxin trophic transfer the Sarasota Bay region was again identified in May 2009. Although all fish species tested were positive for brevetoxin (striped mullet, pinfish, mangrove snapper, sheepshead) (range: 3-\>20.3 ng/g), the striped mullet also concurrently contained DA (mean ± SE: 54±4 ng/g; n = 3). In the Florida Panhandle, scaled sardines accumulate very high concentrations of brevetoxin (\>2400 ng/g liver and GI contents) [@pone.0017394-Naar1] and are not as susceptible to the ichthyotoxic effects of *K. brevis*. *Pseudo-nitzschia* blooms and domoic acid {#s4b} ----------------------------------------- In mid-April 2008, a *Pseudo-nitzschia* bloom was observed in the southern region of Sarasota Bay with densities exceeding 1.5 million cells/L. *P. pseudodelicatissima* and DA were concurrently identified in samples collected from three set sites within Sarasota and Palma Sola Bays in early May 2008. Cell quotas for these samples (5.15 pg/cell) were similar to field estimates of *P. pseudodelicatissima* from Louisiana (4.82 pg/cell) [@pone.0017394-Parsons1] and *P. multiseries* from Texas (2.1 pg/cell) [@pone.0017394-Dickey1], but an order of magnitude lower than some *P. australis* cells from California (7--75 pg/cell) obtained during the 1998 sea lion UME [@pone.0017394-Scholin1]. In 2009, high densities of *Pseudo-nitzschia* spp. were observed in Palma Sola Bay (up to 2.2×10^6^ cells/L) with most samples in Sarasota Bay also containing significant densities of *Pseudo-nitzschia* (\>10^4^ cells/L). However, DA concentrations in the surface water varied greatly (\<dl to 405 ng/L) with low corresponding cell quotas (0.42 pg/cell). Although no ASP events have been documented in the Gulf of Mexico, clearly species of *Pseudo-nitzschia* have been established in the northern region for many years, and we have now demonstrated that toxigenic species of *Pseudo-nitzschia* are also present in southwestern Florida. In light of the fact that prior to May 2008 there were no documented *Pseudo-nitzschia* blooms in southwestern Florida, this study has demonstrated that since at least June 2000, dolphins in this area have been exposed to DA. DA was detected in at least one animal during five out of the eight health assessments. Over our ten-year study period, more than half of the animals (44 out of 83; 53%) tested were positive for DA. Considering the intensity and frequency of *K. brevis* blooms, it seems unusual that there were a greater proportion of DA exposed animals than there were brevetoxin exposed animals (36%). The pharmacokinetic properties of DA in mammals (i.e., rats) suggest that relative to brevetoxin, DA is short-lived and quickly eliminated [@pone.0017394-Suzuki1], [@pone.0017394-Cattet1]. For DA to be observed so often in southwestern Florida dolphins, DA-producing *Pseudo-nitzschia* spp. are likely present with greater frequency and abundance than previously believed or cryptic sources of DA (e.g., benthic) may re-emerge during non-bloom periods. In three animal samples collected in May 2009, milk from lactating females were all DA negative even in light of finding DA in corresponding feces of those same mothers. The absence of DA suggests that at that particular point in time, the potential for maternal transfer of DA was minimal. The trophic transfer of DA has been well documented with growing evidence that both pelagic and benthic routes are important. In the Sarasota Bay region, all fish collected in 2008 (n = 5 individuals of 5 species) were positive for DA. In 2009, only the striped mullet were found to contain DA, whereas the other fish species (pinfish, mangrove snapper, sheepshead) were all negative. The benthic-feeding striped mullet (omnivore; *M. cephalus*) and scaled sardines (pelagic filter feeder; *Harengula jaguana*) contained the highest concentrations of DA, suggesting that toxin content via direct exposure was at least partially explained by the feeding behavior of these fish species. Although the DA concentrations observed in the scaled sardines were still 2 orders of magnitude less than what would be considered unfit for consumption by FDA guidelines (20 µg/g or 20 ppm) and 1--3 orders of magnitude less than levels observed in anchovies and sardines off the coast of California associated with UMEs (0.27--223 µg/g viscera or gut) [@pone.0017394-Scholin1], [@pone.0017394-Lefebvre1], it has been suggested that Gulf of Mexico clupeid species become even more important dolphin dietary items during and following *K. brevis* events [@pone.0017394-Gannon1]. This is postulated to occur as the abundance of the dolphin\'s normal primary prey items (i.e., pigfish, pinfish, mullet) become significantly reduced. As such, clupeids may play a significant role in the trophic transfer of both DA and brevetoxin to southwestern Florida dolphins. Vertical fluxes of DA from blooms of DA-producing *Pseudo-nitzschia* blooms into sediments may in turn contaminate benthic organisms [@pone.0017394-SekulaWood1], [@pone.0017394-Kvitek1] such as demersal fish [@pone.0017394-Lefebvre3], shellfish [@pone.0017394-Blanco1], and crustaceans [@pone.0017394-Ferdin1]. Benthic recycling or reemergence of DA in the Sarasota Bay region may be a significant initiation point for new trophic transfer pathways exposing dolphins to this toxin. In this study, benthic snails collected from two sites contained 3--13 ng DA/g. On the west coast of the US, DA has been observed in other benthic organisms [@pone.0017394-Lefebvre3], [@pone.0017394-Vigilant1] presumably following rapid downward transport of *Pseudo-nitzschia* cells and toxin from ongoing pelagic blooms [@pone.0017394-SekulaWood1] and provide new dynamic routes of exposure to marine mammals. Although sediments from Sarasota Bay were below the limit of detection, DA has been shown to adsorb to sediments. However, since the desorption rates of DA have been shown to be quite rapid (i.e., minutes), in turn maintaining equilibrium with the surrounding water [@pone.0017394-Burns1], unless DA was continually supplied, DA\'s half-life in the sediment will be relatively short. Bottlenose dolphin health {#s4c} ------------------------- Relative to brevetoxin concentrations reported for dolphins from UMEs occurring in other regions of Florida [@pone.0017394-Flewelling1], the concentrations of brevetoxins observed in live Sarasota animals were generally 1--2 orders of magnitude less. No Sarasota Bay resident dolphins were known to have died as a result of exposure to brevetoxin in 2005, even in the midst of a UME that encompassed Sarasota Bay and surrounding waters. During these UMEs, dolphin stomach contents typically contained the highest concentrations of brevetoxin, ranging from \<dl to \>54000 ng/g (n = 53 animals) during the 2005/06 central west Florida UME (unpublished data) and 136 to 6176 ng/g (mean = 2126; n = 30 animals) during the 2004 Panhandle UME [@pone.0017394-Flewelling1]. Similarly, marine mammals that have died during DA-related UMEs have had much higher concentrations of DA within their tissues relative to live Sarasota animals. In 1998, deceased California sea lions had in excess of 1 µg DA/mL feces (up to 182 µg/mL) and between 0.12 and 3.7 µg DA/mL urine; 6- to 40-fold higher than any concentration observed in this study. Nonetheless, the exposure of these animals to either DA or brevetoxin at low concentrations over long periods of time raises issues regarding potential chronic health effects. In this study we examined the relationship between measured low-level biotoxin concentrations and panels of hematologic and biochemical parameters. No relationship was found between brevetoxin concentration and any of the health panels. However, a weak but statistically significant effect of DA concentration was determined for WBC differential; specifically there were increases in total WBC and eosinophil counts. The effect of DA on eosinophil counts was significant regardless of the matrix (urine or feces) from which DA was measured. Yet even the highest eosinophil count (7500 eosinophils/µL; DA concentration = 36.8 ng/g feces) was still lower than the established reference threshold (8100 eosinophils/µL) [@pone.0017394-Schwacke2] suggesting that the low levels of DA are not perturbing hematological parameters outside of the "normal" biological range. Nonetheless, these findings were significant in that the results were consistent with prior studies in mammals. A relatively large percentage of Florida Panhandle dolphins (23%) that were exposed to DA (\<dl - 201 ng/mL urine) had eosinophil counts that were well above the established reference threshold [@pone.0017394-Schwacke3]. The study further found that the elevated eosinophils were associated with more sensitive indicators of functional immune status, such as decreased T-lymphocyte proliferation and increased neutrophil phagocytosis. Similarly, California sea lions with clinical symptoms of DA toxicity had elevated eosinophil counts [@pone.0017394-Gulland1], which is further supported by the observed immunomodulatory effects of DA on mammalian WBCs [@pone.0017394-Levin1], [@pone.0017394-Levin2]. As such, the results from this study support the hypothesis that DA caused an immunomodulatory response in these dolphins, although the implications for an individual survival outcome and/or a population level health effect require further elucidation. Over the course of this study, 13 animals (14%) (all from 2008 and 2009) were concurrently exposed to both brevetoxin and DA. Although neither of the toxin concentrations approached concentrations observed in UME animals, the presence of more than one toxin in individual animals raises questions involving potential synergistic effects. Preliminary data investigating the effects of both DA and brevetoxin suggested that there are no synergistic effects of these toxins on mouse lethality following acute, single dose exposures (J. Naar, pers. comm.) or neuroblastoma cytotoxicity (Y. Bottein, pers. comm.). Nonetheless, stomach content, urine and/or blood samples collected from 9 stranded animals during the 2004 Panhandle UME all contained low concentrations of DA (\<10 ng DA/g or mL) in addition to high concentrations of brevetoxin (M. Twiner, unpublished data. One possible scenario is that the presence of multiple toxins may first result in reduced health such as immunosuppression, which in turn makes these animals more sensitive to secondary stressors leading to death. In addition to well-described neurological toxicities of DA and brevetoxin, there is a growing amount of evidence that suggests that both DA [@pone.0017394-Levin2], [@pone.0017394-Lefebvre4] and brevetoxin [@pone.0017394-Walsh1], [@pone.0017394-Murrell1] also have immunomodulatory effects that may aid in this sensitization. It has recently been shown UME bottlenose dolphins off the coast of Texas (USA) were concurrently exposed to both DA and okadaic acid [@pone.0017394-Fire4] and that the endangered North Atlantic right whales (*Eubalaena glacialis*) are annually exposed to both DA and saxitoxin via consumption of contaminated copepods (*Calanus fimmarchicus*) [@pone.0017394-Leandro1] with postulated long-term effects on reproductive dysfunction. It is generally acknowledged that there is an apparent increase in the incidence and intensity of HABs [@pone.0017394-VanDolah1], [@pone.0017394-Hallegraeff1] potentially influenced by natural and anthropogenic factors, including eutrophication [@pone.0017394-Anderson1] and global climate change [@pone.0017394-Paerl1]. If this apparent increase represents an actual increase in bloom events involving different HAB species producing unique toxic molecules, there will likely be a greater chance of concurrent exposure of aquatic organisms such as dolphins and humans to multiple HAB toxins. It is therefore imperative that future studies attempt to gain a better understanding of the potential effects of exposure to multiple toxins. Supporting Information {#s5} ====================== Table S1 ::: {.caption} ###### **Concentration of brevetoxin (ng/mL or ng/g) in various animal samples.** (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S2 ::: {.caption} ###### **Concentration of domoic acid (ng/mL or ng/g) in various animal samples.** (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Sample collection during bottlenose dolphin health assessments in Sarasota Bay were made possible by the efforts of the staff, students, and volunteers of the Chicago Zoological Society\'s Sarasota Dolphin Research Program along with collaborating scientists and veterinarians from a number of institutions. Funding for health assessments and fish collections was provided by the National Marine Fisheries Service, Dolphin Quest, and Disney\'s Animal Programs. The authors would like to thank Dr. Gary Kirkpatrick of the Phytoplankton Ecology Program, Mote Marine Laboratory, Sarasota, FL for generously supplying cell count data, Hannah Giddens, Jen Maucher, Tod Leighfield, Seana Powers, Liz Symon, and Valeriy Palubok for technical help, and Larry Fulford for dolphin capture-release and fish collections. Disclaimer: This publication does not constitute an endorsement of any commercial product or intend to be an opinion beyond scientific or other results obtained by the National Oceanic and Atmospheric Administration (NOAA). No reference shall be made to NOAA, or this publication furnished by NOAA, to any advertising or sales promotion which would indicate or imply that NOAA recommends or endorses any proprietary product mentioned herein, or which has as its purpose an interest to cause the advertised product to be used or purchased because of this publication. **Competing Interests:**Funding was received from Disney\'s Animal Programs to help conduct this study; however, there are no restrictions on sharing of data and materials. **Funding:**The authors would like to acknowledge their funding sources, including the NOAA Marine Biotoxins Program ([www.chbr.noaa.gov/habar/default.aspx](http://www.chbr.noaa.gov/habar/default.aspx)), NOAA Fisheries ([www.nmfs.noaa.gov](http://www.nmfs.noaa.gov)), the Batchelor Foundation, Dolphin Quest ([www.dolphinquest.com](http://www.dolphinquest.com)), the Georgia Aquarium ([www.georgiaaquarium.org](http://www.georgiaaquarium.org)), Disney\'s Animal Programs (disney.go.com/index), and the International Whaling Commission ([iwcoffice.org](http://wwww.iwcoffice.org)). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: MJT SF LS TKR RSW. Performed the experiments: MJT SF LS LD ZW SM BB TKR RSW. Analyzed the data: MJT SF LS ZW SR TKR RSW. Contributed reagents/materials/analysis tools: MJT SF LS LD SR BB TKR RSW. Wrote the paper: MJT SF LS RSW. [^2]: ¤ Current address: Department of Natural Sciences, The University of Michigan-Dearborn, Dearborn, Michigan

PubMed Central

2024-06-05T04:04:19.740519

2011-3-10

PMC3053360

Introduction {#s1} ============ Dyslipidemia, a major systemic disorder, is one of the most important risk factors for cardiovascular diseases which are a major cause of morbidity and a leading contributor to mortality worldwide [@pone.0017326-Murray1]. It also includes developing countries, and among them in particular China [@pone.0017326-Murray1], [@pone.0017326-Wu1]. Over the next 20 years, cardiovascular disease morbidity and mortality in China has been projected to increase both in absolute number and as a proportion of total disease burden [@pone.0017326-Wu1]. The marked increase in cardiovascular diseases in economically developing countries has resulted from the economic growth and associated sociodemographic changes that have occurred over recent decades. During this period, the burden of illness from infectious disease has fallen. Parallel changes in lifestyle and diet have led to an increase in life expectancy and a greatly increased burden of cardiovascular disease and other chronic diseases [@pone.0017326-Wu1]--[@pone.0017326-Yusuf1]. Dyslipidemia is one of the most important modifiable risk factors for cardiovascular diseases [@pone.0017326-Gordon1]--[@pone.0017326-Verschuren1]. Previous population-based studies form China [@pone.0017326-Wu2]--[@pone.0017326-Kang1], such as the Sino-MONICA Study from 1983 to 1993 [@pone.0017326-Wu1], [@pone.0017326-Wu2], [@pone.0017326-Yao1], the International Collaborative Study of Cardiovascular Diseases in Asian (InterASIA) from 2000 to 2001 [@pone.0017326-He1], and The Fourth Chinese National Nutrition and Health Survey from 2002 [@pone.0017326-Li1], revealed that the Chinese population as compared with Western societies previously had lower concentrations of serum lipids and a lower prevalence of dyslipidemia. The studies also showed a trend over the recent decades towards an increase in the prevalence of dyslipidemia in China, parallel to a change in the lifestyle. The changes predominantly took place in the urban areas, while the situation changed to a lesser degree in rural communities. There have been only few recent studies addressing the medically and socio-economically important question of the prevalence of dyslipidemia and its associated factors in rural and urban areas of China. We, therefore, assessed in our population-based investigation the frequency of dyslipidemia, its associated factors, and the awareness, treatment, and control of dyslipidemia in the urban and rural region of Greater Beijing. Results {#s2} ======= Demographic Parameters {#s2a} ---------------------- The study included 2951 (90.8%) subjects (1674(56.7%) women) for whom serum lipids measurements were available. The mean age was 60.4±10.0 years (median: 60 years; range: 45--89 years). Out of the 2951individuals, 1407(47.7%) subjects (834 women) came from the rural region, and 1544 (52.3%) subjects (840women) came from the urban region. The subjects from the rural region compared with the subjects from the urban region were significantly younger (56.9±9.0 years versus 63.6±9.9 years; *P*\<0.001), and had a significantly lower monthly income (399±310 Yuan versus 2177±594 Yuan; *P*\<0.001) and a lower level of education (*P*\<0.001). The participants of the survey 2006 compared with the non-participants were significantly younger (55.3±10.1 years versus 58.6±11.6 years; *P*\<0.001), came more often from the rural region than from the urban region (1500/1751 versus 473/714; *P*\<0.001), and had a higher level of education (*P* = 0.001). There were no significant differences in gender (*P* = 0.84) and reported income (1067±827 versus 1093±1052; *P* = 0.41). Laboratory Results {#s2b} ------------------ Mean levels of total cholesterol, HDL (high-density lipoprotein) cholesterol, LDL (low-density lipoprotein) cholesterol, and triyglcerides were 4.92±1.01 mmol / L, 1.61±0.36 mmol/L, 2.88±0.85 mmol/L, and 1.76±1.29 mmol/L, respectively ([Fig. 1](#pone-0017326-g001){ref-type="fig"}). A hypercholesterolemia (total cholesterol concentration ≥5.72 mmol/L (220 mg/dL)) was found in 19.0±0.7%, a hypertrigylceridemia (triglyceride concentration ≥1.70 mmol/L (150 mg/dL)) in 35.5±0.9%, and abnormally low high-density lipoprotein-cholesterol (HDL-C concentration ≤0.91 mmol/L (35 mg/dL)) in 1.7±0.2% of the study population. Dyslipidemia was found for 1332 (45.1±0.9%) subjects (95%CI: 43.3%, 46.9%). A positive history for dyslipidemia was described by 842 (28.5%) subjects, so that the total prevalence of dyslipidemia was 1654 / 2951 or 56.1±0.9% (95%CI: 54.3%, 57.8%). For all further statistical analysis, the total prevalence of dyslipidemia (abnormal blood examinations results and / or positive history for dyslipidemia) was taken. Adjusted for age and gender on the basis of Chinese census 2000, the prevalence of dyslipidemia was 54.1%. ::: {#pone-0017326-g001 .fig} 10.1371/journal.pone.0017326.g001 Figure 1 ::: {.caption} ###### Boxplots showing the distribution of total cholesterol, triglycerides, high-density lipoprotein cholesterol and low-density lipoprotein cholesterol in the Beijing eye Study. :::  ::: Univariate Analysis {#s2c} ------------------- In univariate analysis, all four blood laboratory parameters were significantly (*P*\<0.001) correlated with each other. In univariate analysis, the prevalence of dyslipidemia was significantly associated with age, female gender, urban region, body mass index (all *P*\<0.001), lower body height (*P* = 0.02), body weight (*P*\<0.001), level of education (*P*\<0.001), income (*P*\<0.001), fasting serum concentration of glucose (*P*\<0.001), systolic blood pressure (*P* = 0.02), mean arterial blood pressure (*P* = 0.03), and less consumption of wine (*P = *0.001) and less smoking (*P = *0.002) ([Table 1](#pone-0017326-t001){ref-type="table"}). The prevalence of dyslipidemia was not significantly associated with consumption of beer (*P* = 0.79). ::: {#pone-0017326-t001 .table-wrap} 10.1371/journal.pone.0017326.t001 Table 1 ::: {.caption} ###### Associations between the presence of dyslipidemia and other parameters in the Beijing Eye Study 2006. ::: {#pone-0017326-t001-1} Univariate analysis: --------------------------------------------------- --------- ------ ------------ Age \<0.001 1.02 1.02, 1.03 Female gender 0.001 1.29 1.11, 1.49 Urban region \<0.001 2.09 1.81, 2.43 Body mass index \<0.001 1.11 1.08, 1.13 Body height 0.02 0.99 0.98, 0.99 Body weight \<0.001 1.02 1.02, 1.03 Level of education \<0.001 1.15 1.08, 1.23 Income \<0.001 1.20 1.15, 1.24 Fasting serum concentration glucose concentration \<0.001 1.14 1.08, 1.20 Systolic blood pressure 0.02 1.01 1.01, 1.02 Mean arterial blood pressure 0.03 1.01 1.01, 1.02 Consumption of wine 0.001 0.86 0.78, 0.94 Smoking 0.002 0.79 0.67, 0.92 Consumption of beer 0.79 Binary regression analysis: ----------------------------- --------- ------ ------------ Age \<0.001 1.02 1.01, 1.03 Gender \<0.001 1.51 1.25, 1.83 Urban region 0.001 1.82 1.30, 2.55 Body mass index \<0.001 1.13 1.10, 1.15 Income 0.01 1.11 1.02, 1.21 Blood concentr. of glucose \<0.001 1.10 1.05, 1.16 Diastolic blood pressure 0.02 1.02 1.01, 1.03 Smoking 0.04 1.23 1.01, 1.51 Level of education 0.40 Systolic blood pressure 0.56 Consumption of wine 0.24 ::: Multivariate Analysis {#s2d} --------------------- In a binary multivariate logistic regression analysis, with the presence of dyslipidemia as dependent variable and all parameters, which were significantly associated with dyslipidemia in univariate analysis, as independent variables, revealed that the presence of dyslipidemia was still significantly associated with increasing age (*P*\<0.001), female gender (*P*\<0.001), urban region (*P* = 0.001), body mass index (*P*\<0.001), income (*P* = 0.01), blood concentration of glucose (*P*\<0.001), diastolic blood pressure (*P* = 0.02), and smoking (*P* = 0.04) ([Table 1](#pone-0017326-t001){ref-type="table"}). In the binary logistic regression analysis with adjustment for age, gender, urban versus rural region, body mass index, income, blood glucose concentration, diastolic blood pressure and smoking, the prevalence of dyslipidemia was no longer associated with level of education (*P* = 0.40), systolic blood pressure (*P* = 0.56) and consumption of wine (*P* = 0.24). Awareness of Dyslipidemia {#s2e} ------------------------- Out of the 2951 study participants, 842 (28.5%) subjects (28.5%±0.8%; 95%CI: 26.9%, 30.2%) or 842 subjects out of the 1654 subjects with dyslipidemia (50.9%±1.2%; 95%CI: 48.5%, 53.3%) reported of any previous diagnosis of dyslipidemia by a healthcare professional. Within the group of subjects with dyslipidemia, the prevalence of the awareness of dyslipidemia was significantly associated (univariate analysis) with increasing age (*P*\<0.001), male gender (*P* = 0.001), urban region (*P*\<0.001), lower body mass index (*P* = 0.001), higher income (*P*\<0.001), lower blood concentration of glucose (*P* = 0.02), lower diastolic blood pressure (*P*\<0.001), and less smoking (*P*\<0.001) ([Table 2](#pone-0017326-t002){ref-type="table"}). ::: {#pone-0017326-t002 .table-wrap} 10.1371/journal.pone.0017326.t002 Table 2 ::: {.caption} ###### Associations between the awareness of dyslipidemia and other parameters in the Beijing Eye Study 2006. ::: {#pone-0017326-t002-2} ----------------------------- --------- ------ ------------ Univariate analysis: Age \<0.001 1.04 1.03, 1.05 Male gender 0.001 0.72 0.59, 0.88 Urban region \<0.001 10.1 7.96, 12.8 Body mass index 0.001 0.96 0.93, 0.98 Income \<0.001 1.67 1.58, 1.77 Blood conc. of glucose 0.02 0.94 0.89, 0.99 Diastolic blood pressure \<0.001 0.94 0.92, 0.95 Smoking \<0.001 0.67 0.54, 0.84 Binary regression analysis: Urban region 0.001 6.50 4.10, 10.3 Body mass index 0.001 1.06 1.02, 1.09 Income 0.02 1.14 1.02, 1.27 Age 0.56 Gender 0.21 Blood conc. Glucose 0.24 Diastolic blood pressure 0.11 Smoking 0.97 ----------------------------- --------- ------ ------------ ::: In a multivariate binary logistic regression analysis, with the presence of awareness of dyslipidemia as dependent variable and all parameters, which were significantly associated with dyslipidemia in univariate analysis, as independent variables, revealed that the presence of dyslipidemia was still significantly associated with urban region (*P* = 0.001), body mass index (*P* = 0.001), and higher income (*P* = 0.02), while it was no longer significantly associated with age (*P* = 0.56), gender (*P* = 0.21), blood concentration of glucose (*P* = 0.24), diastolic blood pressure (*P* = 0.11), and smoking (*P* = 0.97) ([Table 2](#pone-0017326-t002){ref-type="table"}). Treatment of Dyslipidemia {#s2f} ------------------------- Out of the 1654 subjects with dyslipidemia 393 (23.8%±1.0%; 95%CI: 22.0%, 26.0%) reported to be under treatment of the disorder. Out of the 393 subjects under treatment, 236 (60.1%) still had abnormally high levels of total cholesterol or of triglycerides or abnormally low high-density lipoprotein-cholesterol concentrations. Discussion {#s3} ========== Our study aimed to examine the prevalence of dyslipidemia, its associated factors, awareness, treatment, and control in the urban and rural region of Greater Beijing. It revealed that the prevalence of dyslipidemia was about 56%. Transferred onto the total population of China, this figure may imply that approximately 175 million Chinese may be affected by dyslipidemia. It shows the tremendous importance, dyslipidemia may get in the next future in China. The results of our study agree with previous populations-based studies from China. If the results of our study were compared with the findings from the InterAsia Collaborative Study [@pone.0017326-He1], the mean concentrations of total cholesterol (4.92 mmol/L or 190.2 mg/dL versus 186.1 mg/dL), HDL cholesterol (1.61 mmol/L or 62.3 mg/dL versus 51.7 mg/dL), LDL cholesterol (2.88 mmol/L or 111.4 mg/dL versus 109.5 mg/dL), and triyglcerides (1.76 mmol/L or 155.3 mg/dL versus 128.1 mg/dL) did not vary markedly between both studies [@pone.0017326-He1]. Also other regional studies have previously examined serum lipid concentrations in Chinese populations [@pone.0017326-Li1], [@pone.0017326-Johnson1]. The PRC-USA Collaborative Study in Cardiovascular and Cardiopulmonary Epidemiology, which was performed in 1983 to 1984, reported that age-adjusted mean serum total cholesterol level was higher in urban than in rural samples and generally higher in Beijing than in Guangzhou [@pone.0017326-Johnson1]. In a repeated survey conducted in the same populations during 1993 to 1994, the mean total cholesterol level increased in Guangzhou but decreased in Beijing [@pone.0017326-Li1]. However, this study was conducted in an occupational population sample of convenience rather than in a representative sample of the general population. A rapid increase in total serum cholesterol level has also been observed in residents living in Shanghai, China [@pone.0017326-Jia1]. The differences in dietary nutrient intake between north and south as well as between rural and urban China may contribute to the observed regional differences in serum lipid levels [@pone.0017326-Zhou1]. It is confirmed in our study, in which the prevalence of dyslipidemia was significantly higher in the urban region than in the rural region, after adjustment for age, gender, body mass index, income, blood glucose concentration, diastolic blood pressure and smoking. It is in agreement with the general increase in the prevalence of dyslipidemia in China with increasing urbanization and change in lifestyle. Correspondingly, in the regions with fast economic growth, such as Guangzhou and Shanghai, the mean level of serum cholesterol previously increased markedly [@pone.0017326-Li1], [@pone.0017326-Jia1]. To mention an example, over a 10-year period the serum total cholesterol level for men and women in Guangzhou increased 13.9% and 21.5% in urban areas and 18.9% and 24.9% in rural areas, respectively [@pone.0017326-Li1]. The prevalence of dyslipidemia as found in our study (total dyslipidemia: 56.1±0.9%; dyslipidemia without taking into account history of dyslipidemia: 45.1±0.9%) and the prevalence of hypercholesterolemia (19.0±0.7%), hypertrigylceridemia (35.5±0.9%) and hypo HDL-cholesterol (1.7±0.2%) was partially comparable to figures reported in previous investigations. To cite an example, in the InterAsia Collaborative Study, a cross-sectional survey in a nationally representative sample of 15,540 Chinese adults 35 to 74 years of age, 23.8% of the subjects had borderline high total cholesterol (200 to 239 mg/dL), and 9.0% had high total cholesterol (≥240 mg/dL). The population estimates for borderline high (130 to 159 mg/dL), high (160 to 189 mg/dL), and very high ≥190 mg/dL) LDL- cholesterol were 17.0%, 5.1%, and 2.7%, respectively. In addition, 19.2%, had a low HDL-cholesterol (\<40 mg/dL). In a recent study on 19,003 suburban Beijing residents aged 18 to 76 years, the age-standardized prevalence of dyslipidemia was 30.3% [@pone.0017326-Zhang1]. Our findings may have public health implications. Traditionally, mortality from coronary heart disease in China was infrequent and was estimated to be only 10% of that in Western populations [@pone.0017326-Tao1]. A low serum total cholesterol level related to a low habitual dietary intake of fat and cholesterol was considered to be the main underlying reason for the low coronary heart disease mortality in China [@pone.0017326-Johnson1]. In the recent InterASIA study [@pone.0017326-He1], a relatively high mean level of serum cholesterol but a low rate of hypercholesterolemia control was noted. These findings were confirmed by our study which was conducted 5 years after the InterASIA study. It might explain the recent rapid increase in coronary heart disease mortality in China. Furthermore, the findings from the InterASIA study as well as our results suggest that without a national emphasis on prevention, treatment, and control of dyslipidemia, the societal burden of cardiovascular diseases in China will continue to increase in the near future. Factors associated with dyslipidemia in our study were higher age, female gender, urban region, higher body mass index, income, blood concentration of glucose and diastolic blood pressure, and smoking. In contrast, the level of education, systolic blood pressure and alcohol consumption were not related with dyslipidemia. The association between dyslipidemia and higher age agrees with all previous population-based and hospital-based studies on the same topic [@pone.0017326-Wu1], [@pone.0017326-Gordon1]--[@pone.0017326-Johnson1]. In a similar manner, the previous studies as our study reported on relationships between dyslipidemia and female gender, urban region, higher body mass index, income, blood concentration of glucose and diastolic blood pressure, and smoking. It may suggest that the factors leading to dyslipidemia and the consequences of dyslipidemia (such as increased body mass index) are similar across ethnic borders. It may imply that, as already pointed out above, the increasing rate of dyslipidemia in China may lead to similar cardiovascular and cerebrovascular consequences as it did in Western countries several decades ago. The rate of awareness among the subjects with dyslipidemia in the present investigation was 50.9%, which was unexpectedly high. In previous studies from China such as the InterASIA study, the awareness rate was less than 10%. The reason for the unexpectedly high awareness rate may be differences in the regions included into the studies. The average income in the urban regions included into the Beijing Study was considerably higher than in other urban regions of Beijing (1688 RMB (Yuan) versus 866 RMB). In addition, the health care system in the urban study regions which also included quarters with government employees was better developed than in other urban regions of Beijing. In parts of the urban study regions, there was a relatively high standard with some communities supplying free health care examinations, and in these areas, the cost for medical care was covered by the government. Correspondingly, the rate of awareness of dyslipidemia was significantly associated with living in the urban region. Without doubt, the relatively high income level and the relatively well developed health care system in the urban regions of the present study may have artificially increased the awareness rate of dyslipidemia in our study. In a similar manner, the rural regions included in our study had a better developed health care system than other rural regions in the vicinity so that also the data on the awareness of dyslipidemia from the rural regions may have a bias and may be artificially high. In addition, the first survey performed in 2001 included a questionnaire with questions on the presence, duration and treatment of dyslipidemia. Although the blood lipids were not measured in 2001, the questionnaire might have made the study participants aware of the risks of dyslipidemia, so that they got their blood examined between 2001 and 2006. It may have led to a bias with an artificially high rate of awareness of dyslipidemia in the survey of 2006. The data on the awareness of dyslipidemia as found in the present study may, therefore, not be representative for the whole country. Potential limitations of our study should be mentioned. First, a major concern in any prevalence study is non-participation. In the present study, out of 5324 eligible individuals, 4439 subjects participated in the examination in the year 2001. Out of these 4439 subjects, 3251 (73.2%) returned for the follow-up examination in the year 2006 in which the blood pressure measurements were performed. Based on the originally eligible number of 5324 individuals from the year 2001, the number of 3251 participants in the survey of 2006 indicates a participation rate of 61.1%. The participants of the survey 2006 compared with the non-participants of the survey 2006 were significantly younger, came more often from the rural region, and had a higher level of education. They did not, however, vary in significantly income and gender. One may, therefore, infer that the participation rate may still be acceptable to draw conclusions about the prevalence and associated factors of dyslipidemia in the Greater Beijing area. Second, another limitation of the study may be the question how representative the rural region and the urban regions of Greater Beijing are for the remaining provinces of China. The unexpectedly high rate of awareness of dyslipidemia particularly in the urban region may suggest that one may be cautious in transferring the data of awareness of the diseases to other less developed regions in China. This may, however, not account for the prevalence of dyslipidemia and its associated factors. In conclusion, dyslipidemia was present in about 56% of the study population (aged 45+ years) in Greater Beijing. Factors, which were likely associated with dyslipidemia, were higher age, female gender, urban region, higher body mass index, higher income, higher blood concentration of glucose, higher diastolic blood pressure, and smoking. In the examined study population, the treatment rate of dyslipidemia was 24% with about 60% of the treated subjects still having uncontrolled dyslipidemia. Methods {#s4} ======= Study Participants {#s4a} ------------------ The Medical Ethics Committee of the Beijing Tongren Hospital approved the study protocol and all participants gave written informed consent. The eligibility criterion for the study was an age of 40 or more years. The Beijing Eye Study is a population-based study performed in the region of Greater Beijing. The study was carried out in 7 communities. Three of the communities were located in the Daxing District in the village area of Yufa situated south of Beijing area about 50 to 100 km from the center of Beijing. Four communities were located in the Haidian urban district of the Northern part of Central Beijing. The reason to perform the study in a rural area and in an urban area was that both areas differed markedly in the level of education, access to medical care, mobility, frequency of hereditable diseases, and way of life. In the rural areas, health care services and a referral system to ophthalmologists were often not available, and the cost for medical care was usually not covered by the government. In the urban areas, health care was at a relatively high standard with some communities supplying free ophthalmic examinations, and in these areas, the cost for medical care were covered by the government. All people residing in the communities were officially registered by name, gender and age at the local mayor\'s office. Using this register as the sampling frame, all subjects living in the 7 communities and fulfilling the inclusion criterion of an age of 40+ years were eligible for the study. Home visits were performed according to the registration list, and the eligibility criteria for the study, an age of 40 or more years, was confirmed by door-to-door enrollment. Out of 5324 eligible individuals, 4439 individuals (2505 women) participated (response rate: 83.4%). The mean age was 56.2±10.6 years (range: 40--101 years). The study was described in detail recently [@pone.0017326-Xu1]--[@pone.0017326-Xu3]. To assess the public health impact and to address how representative the sample of the study was for the total population of China, the demographic data of the study population was compared with census data available for the population of China and of Beijing [@pone.0017326-Beijing1]. The average monthly income per capita in the rural areas of Beijing was similar in the .census (391 RMB (Yuan)) and in the rural part of the study population (393 RMB (Yuan)). The income figures of the urban regions were higher in the Beijing Eye Study population (1688 RMB (Yuan)) than in the census (866 RMB (Yuan)). We repeated the study in the year 2006 by re-inviting all participants from the previous survey, and 3251 subjects (1838 (56.5%) women) participated (response rate: 73.2%; 1500 (46.1%) subjects from the rural region). The mean age was 60.4±10.0 years (range: 45--89 years). For the present study, only data measured in the survey of 2006 were taken, and only subjects with blood pressure measurements were considered. Methods {#s4b} ------- All examinations were carried out in the communities, either in schoolhouses or in community houses. Trained research technicians asked the study participants questions from a standard questionnaire providing information on demographic variables such as age, gender, level of education, occupation, family income, known diagnosis of dyslipidemia, and current treatment of dyslipidemia. In the year 2006, blood samples were additionally taken. The serum concentrations of total cholesterol, triglycerides, high-density lipoprotein-cholesterol (HDL-C) and low-density lipoprotein-cholesterol (LDL-C) were measured using the enzymatic method with a Hitachi 7600 auto-analyzer (Hitachi, Tokyo, Japan). Reagents of the same batch were used. Dyslipidemia was defined as any of hypercholesterolemia (total cholesterol concentration ≥5.72 mmol/L (220 mg/dL)) or hypertriglyceridemia (triglyceride concentration ≥1.70 mmol/L (150 mg/dL)) or low high-density lipoprotein-cholesterol (HDL-C concentration ≤0.91 mmol/L (35 mg/dL)) [@pone.0017326-The1]. Total prevalence of dyslipidemia was defined dyslipidemia and / or a positive history for dyslipidemia. For the statistical analyses, the total prevalence of dyslipidemia (abnormal blood examinations results and / or positive history for dyslipidemia) was taken. Body weight and height and the arterial blood pressure were also measured {#s4c} ------------------------------------------------------------------------- Awareness of dyslipidemia was defined as self-report of any previous diagnosis of dyslipidemia by a healthcare professional. Treatment of dyslipidemia was defined as the use of a pharmacological treatment to lower blood lipids during the previous 2 weeks. Participants were considered to have controlled dyslipidemia if their total cholesterol concentration was 5.2 mmol/l (\<200 mg/dL), or to have controlled LDL cholesterol concentration if their LDL cholesterol was \<3.38 mmol/L (130 mg/mL) [@pone.0017326-He1]. Statistical Analysis {#s4d} -------------------- Inclusion criterion for the present study was the availability of serum measurements of total cholesterol, triglycerides, and HDL-C measurements. The statistical analysis was performed using a commercially available statistical software package (SPSS for Windows, version 17.0, SPSS, Chicago, IL). The data were given as mean ± standard deviation. Logistic regression was used to investigate the associations of the binary dependent variable "presence of dyslipidemia" with the continuous or categorical independent variables, such as age, gender, area, body mass index, the concentrations of glucose, blood pressure, education and income. Confidence intervals were presented. All *P*-values were two-sided and were considered statistically significant when the *P*-values were less than 0.05. Based on the data of the 2000 China population census, the mean data and the prevalences were adjusted to calculate the total number of subjects with dyslipidemia in China [@pone.0017326-The2]. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**The authors have no support or funding to report. [^1]: Conceived and designed the experiments: SW LX JBJ YXW QSY HY. Performed the experiments: SW LX JBJ YXW QSY HY. Analyzed the data: SW JBJ. Contributed reagents/materials/analysis tools: LX JBJ. Wrote the paper: SW JBJ.

PubMed Central

2024-06-05T04:04:19.747274

2011-3-10

PMC3053361

Introduction {#s1} ============ There is widespread scientific agreement that coral reef ecosystems worldwide are being rapidly degraded [@pone.0017516-Knowlton1], [@pone.0017516-Hughes1]. Substantial declines in coral abundance are thought to have occurred in most coral reef regions [@pone.0017516-Pandolfi1] and coral decline is frequently described as ongoing with the integrity and persistence of the reef system threatened by a number of different stressors [@pone.0017516-Bellwood1]. Given the important ecosystem services of corals to reef ecosystems, including reef accretion and provision of habitat structure, accurately quantifying the dynamics of coral is important to both scientists and managers in understanding the causes and consequences of reef degradation. Climate change is widely regarded as the single greatest threat to coral reef ecosystems. There are clear links between increases in ocean temperature and coral bleaching. Global patterns of coral loss [@pone.0017516-Wilkinson1] suggest that areas closest to urban centres are most degraded, implying that chronic stressors are compounding the effects of warming. Most studies on how climate change will impact coral reefs predict changes in the frequency or severity of existing stressors. There are, however, few studies that have directly compared the relative effects of different disturbances and data on the propensity, severity, or spatial patterns of impacts across large spatial scales. Severe storms can strip the entire substrate [@pone.0017516-Ayling1], [@pone.0017516-Fabricius1], [@pone.0017516-Done1] and affect large reef tracts while less intense storms may affect only fragile growth forms [@pone.0017516-Cheal1] and promote fragmentation that leads to rapid regeneration [@pone.0017516-Ayling2]. *Acanthaster planci* predation, coral disease, and bleaching also have differential impacts depending on severity [@pone.0017516-Marshall1], [@pone.0017516-McClanahan1], [@pone.0017516-Moran1]. Understanding the patterns of change in coral cover in relation to different sources of disturbance should provide a broader understanding of risks to reef resilience. Coral decline can result from disturbance that is too frequent or intense, or from depressed recovery due to stress or recruitment failure, as has been documented in the Caribbean [@pone.0017516-Somerfield1]. Mass mortality of coral on the Great Barrier Reef (GBR) has been associated with *A. planci* [@pone.0017516-Miller1], bleaching [@pone.0017516-Berkelmans1], disease [@pone.0017516-Willis1] and storms [@pone.0017516-Fabricius1], [@pone.0017516-Halford1]. Recent research on disturbance and recovery of corals on the GBR has mostly examined responses to specific impacts such as bleaching [@pone.0017516-Baird1], outbreaks of *A. planci* [@pone.0017516-Pratchett1], [@pone.0017516-Wilson1], storms and cyclones [@pone.0017516-Fabricius1], [@pone.0017516-Emslie1] or has documented changes in coral communities in relation to specific environmental gradients [@pone.0017516-DeVantier1], [@pone.0017516-Fabricius2], [@pone.0017516-Burgess1]. Globally, there are only 26 records for coral bleaching prior to 1982 and the scale and extent of bleaching on the GBR since 1998 is unprecedented [@pone.0017516-Oliver1]. Coral disease is an emerging stressor that was first recorded on the GBR in the early 1990s [@pone.0017516-Willis1], [@pone.0017516-Lough1]. Cyclone intensity is also predicted to increase in a warming climate and since 1995, three high intensity systems have crossed the GBR [@pone.0017516-Lough2]. *A. planci* outbreaks peaked in 2003 but reefs on the southern GBR have been experiencing continuous *A. planci* predation since monitoring began [@pone.0017516-Sweatman1]. Disturbances appear to be increasing in frequency and severity and it is not known whether coral growth will be able to keep pace with increased disturbance. Rapid recovery of disturbed reefs has been recorded with some reefs taking less than 10 years to recover their previous communities from low coral cover [@pone.0017516-Halford1], [@pone.0017516-Emslie1]. However modelling studies based on current rates of disturbance and recovery predict long-term declines on both inshore and offshore reefs [@pone.0017516-Thompson1], [@pone.0017516-Wakeford1]. Percent cover of live coral is by far the most widely used metric of coral reef condition and is universally used in studies that document coral reef decline and recovery across large spatial scales [@pone.0017516-Pandolfi1], [@pone.0017516-Bellwood2], [@pone.0017516-Bruno1]. Determining regional trends in coral cover is difficult due to the large spatial and temporal scales involved. The variability and stochastic nature of disturbance events that shape reef communities mean that small scale studies can easily miss or over represent the impact of localised disturbance events. The data record for coral reefs prior to the 1980s is very limited. On the GBR, regional studies that include data prior to mid 1980\'s are all based on metadata and suggest reef condition has declined. Bellwood et al [@pone.0017516-Bellwood1] show ongoing decline up to 2004, while Bruno et al [@pone.0017516-Bruno1] estimate there were substantial losses of coral just before the mid 80\'s. There are few reef regions where systematically collected large scale data are available. For the GBR, we are fortunate to have data from a dedicated large-scale monitoring program. Regional estimates of coral cover using the manta-tow method are available from the mid 80\'s and found average reef-wide coral cover across the GBR declined from 28.1% to 21.7% between 1986 and 2004 [@pone.0017516-Sweatman2], a substantially lower figure than estimates made from metadata. Since 1992, the monitoring of permanent sites, designed specifically to assess change in coral cover, has avoided many of the short-comings associated with metadata and allows for a powerful analysis of coral cover change over time and multiple spatial scales. The aims of this study were threefold. Firstly, to examine the patterns of change and estimate the extent of decline in total hard coral, Acroporidae, and non-Acroporidae coral on the GBR from 1995 to 2009 across multiple spatial scales. Secondly, to identify the agents of disturbance that have caused coral decline. Thirdly, assess the extent of recovery in coral cover after periods of decline. Our results indicate that, from 1995 to 2009, GBR-wide coral cover did not decline. Rather, there have been contrasting and uncorrelated temporal trends in coral cover at subregional scales (10--100 km), driven mostly by changes in fast-growing Acroporidae as a result of localized disturbance events, mainly *A. planci* predation and cyclones. Results {#s2} ======= Spatial and temporal trends in total hard coral cover, Acroporidae, and non-Acroporidae coral {#s2a} --------------------------------------------------------------------------------------------- At the scale of the whole GBR, there was no net decline in hard coral cover between 1995 and 2009. Average cover was 30% over this period, peaking at 33% in 1999 and lowest at 27% in 1995 ([Fig. 1A](#pone-0017516-g001){ref-type="fig"}). However, this apparent overall stability was a product of variable and asynchronous increases and decreases in coral cover at the scale of subregions ([Fig. 2A](#pone-0017516-g002){ref-type="fig"}, [Fig. 3](#pone-0017516-g003){ref-type="fig"}, B--P). Coral cover varied dramatically in some subregions (e.g., [Fig. 3D, P](#pone-0017516-g003){ref-type="fig"}) while changing relatively little in others (e.g., [Fig. 3C,G, K, N](#pone-0017516-g003){ref-type="fig"}). The linear trend in total hard coral cover was positive for 6 subregions and negative for 7 ([Table S1](#pone.0017516.s002){ref-type="supplementary-material"}). For two subregions ([Fig. 3D, P](#pone-0017516-g003){ref-type="fig"}) the trend in total cover was clearly non-linear and linear trends were not assessed. Only one subregion ([Fig. 3I](#pone-0017516-g003){ref-type="fig"}) had a statistically significant linear trend, which was negative ([Table S1](#pone.0017516.s002){ref-type="supplementary-material"}). ::: {#pone-0017516-g001 .fig} 10.1371/journal.pone.0017516.g001 Figure 1 ::: {.caption} ###### Temporal trends in percent cover of hard coral on the Great Barrier Reef (1995--2009). \(A) Average coral cover for the whole GBR; (B--P) Average coral cover in each subregion. Dots and dashed lines show the average subregion temporal profile with 95% confidence intervals. For the whole GBR there was a non-significant linear trend of −0.27%, (−0.68, 0.14 95%CI) as a result of asynchronous increases and decreases in each subregion. The grey shaded area indicates periods of coral decline associated with disturbance. :::  ::: ::: {#pone-0017516-g002 .fig} 10.1371/journal.pone.0017516.g002 Figure 2 ::: {.caption} ###### Linear trends in percent cover of hard coral, Acroporidae and non-Acroporidae families on the Great Barrier Reef (1995--2009). \(A) Hard Coral, (B) Acroporidae coral, (C) Non-Acroporidae coral. Straight lines show the linear trend and confidence intervals (dashed lines). Individual reef profiles are in grey. Non-Acroporidae coral declined significantly ([Table S1](#pone.0017516.s002){ref-type="supplementary-material"}). :::  ::: ::: {#pone-0017516-g003 .fig} 10.1371/journal.pone.0017516.g003 Figure 3 ::: {.caption} ###### Temporal trends in percent cover of hard coral on the Great Barrier Reef (1995--2009). (B--P) Average annual coral cover in each subregion. Dashed lines show the subregion temporal profile and the straight lines show the average linear trend. Individual reef profiles are in grey. Disturbances associated with coral decline are represented by a dot for each reef where that type of disturbance occurred. :::  ::: Temporal trends in the cover of Acroporidae corals were, in most cases, similar to those for total coral cover ([Fig. S1](#pone.0017516.s001){ref-type="supplementary-material"}, [Table S1](#pone.0017516.s002){ref-type="supplementary-material"}). Overall, change in the cover of Acroporidae accounted for 68% of the change in total hard coral cover. The GBR wide trend in Acroporidae cover was also slightly negative (−0.13%, −0.47, 0.20 95%CI) per year but non-significant ([Fig. 2B](#pone-0017516-g002){ref-type="fig"}). At subregional scales, the cover of Acroporidae had qualitatively similar positive and negative trends to hard coral ([Table S1](#pone.0017516.s002){ref-type="supplementary-material"}). For both total hard coral and Acroporidae, reefs within subregions had similar temporal profiles in all but one subregion ([Fig. 3N](#pone-0017516-g003){ref-type="fig"}, [Fig. S1N](#pone.0017516.s001){ref-type="supplementary-material"}), where *A. planci* predation reduced coral cover on two surveyed reefs, while three reefs were unaffected. At the scale of the whole GBR, the cover of non-Acroporidae declined −0.16% (−0.26, −0.06 95%CI) per year, which was statistically significant ([Fig. 2C](#pone-0017516-g002){ref-type="fig"}). At subregional scales, the cover of non-Acroporidae increased in 4 and decreased in 11 subregions ([Fig. S1](#pone.0017516.s001){ref-type="supplementary-material"}; [Table S1](#pone.0017516.s002){ref-type="supplementary-material"}). The cover of Acroporidae was also more variable over time than that of non-Acroporidae corals. The average coefficient of variation (CV) across all reefs was twice as large for Acroporidae (CV = 0.48) compared to non-Acroporidae (CV = 0.22). Impacts of different disturbance agents {#s2b} --------------------------------------- Storms and *A. planci* predation were the most important agents of decline, both in terms of their prevalence and severity ([Table 1](#pone-0017516-t001){ref-type="table"}, [Fig. 4](#pone-0017516-g004){ref-type="fig"}). Storms and *A. planci* predation each accounted for approximately one third of coral loss and affected 24 and 23 individual reefs respectively ([Table 1](#pone-0017516-t001){ref-type="table"}). *A. planci* outbreaks were somewhat more severe than storms, with reefs losing 42.2% of cover during *A. planci* outbreaks compared with 31.5% following storms ([Table 1](#pone-0017516-t001){ref-type="table"}). Additionally, *A. planci* outbreaks typically affected reefs for 2--3 consecutive years or more, resulting in more reef-years of impact and a lower remaining absolute cover level than that following storms ([Table 1](#pone-0017516-t001){ref-type="table"}). ::: {#pone-0017516-g004 .fig} 10.1371/journal.pone.0017516.g004 Figure 4 ::: {.caption} ###### Prevalence and severity of disturbance events. *A.planci*, storms and multiple disturbances were associated with the greatest coral decline and the largest distribution of coral loss. Despite bleaching being widespread in 1998 and 2002, coral mortality was relatively low. (A) the loss in live coral cover as a percentage of the pre-event cover. (B) Relative magnitude of loss of pre-event cover summed for all reefs and years. :::  ::: ::: {#pone-0017516-t001 .table-wrap} 10.1371/journal.pone.0017516.t001 Table 1 ::: {.caption} ###### The effects of various agents of disturbance on live coral cover on the Great Barrier Reef from 1995 to 2009. ::: {#pone-0017516-t001-1} Disturbance type Storms *A.planci* Disease Multiple Bleaching Unknown -------------------------------------------------- -------- ------------ --------- ---------- ----------- --------- \% of summed proportional loss 33.8 36.7 6.5 6.1 5.6 11.2 No. reefs affected 24 23 11 5 10 21 Reef-years of effect 32 71 22 6 12 29 Mean proportion of existing cover lost per event −31.5 −42.2 −12.2 −36.6 −13.9 −12.9 Mean pre-event coral cover 38.4 32.2 53.0 43.2 36.9 39.8 Mean post-event coral cover 25.5 18.1 46.9 28.1 31.4 35.0 ::: Bleaching and disease had far less impact than *A. planci* and storms in prevalence and severity. Eleven reefs were affected by disease, with a mean of 12.2% of pre-existing cover lost per outbreak, accounting for just 6.5% of total overall proportional losses ([Table 1](#pone-0017516-t001){ref-type="table"}). Observed bleaching events were consistent with the overall pattern described for the GBR [@pone.0017516-Berkelmans2] with coral losses of 30--35% of the existing coral cover on the worst affected reef ([Fig. 3H](#pone-0017516-g003){ref-type="fig"}). However, while bleaching was widespread in 1998 and 2002 and locally severe on the southern GBR in 2006 [@pone.0017516-Berkelmans3] only ten of the sampled reefs had coral decline. Mean coral cover loss attributed to bleaching was 13.9%, just 5.6% of total losses across all reefs. Pre-event coral covers were similar for bleaching, *A. planci* and storm events (means of 37, 32, 38% respectively) but higher for disease, multiple and unknown agents of decline (means of 53, 43, 40% respectively) ([Table 1](#pone-0017516-t001){ref-type="table"}). Spatial patterns of disturbances shifted over the survey period from the inner and mid-shelf to the outer shelf. Most inshore and mid-shelf subregions had periods of continuous disturbance up to 2002 but have been in recovery phases since the mid 2000s ([Fig. 1](#pone-0017516-g001){ref-type="fig"}, [Fig. 3](#pone-0017516-g003){ref-type="fig"}). On outer shelf subregions disturbance has been more intermittent ([Fig. 1](#pone-0017516-g001){ref-type="fig"}). This was also the case in the Whitsunday sector across all shelf positions ([Fig. 1K, L, M](#pone-0017516-g001){ref-type="fig"}). Ten reefs had intense disturbance where coral cover declined to less than 10%, including two of the three reefs in the Townsville mid-shelf where there was a significant decline in coral ([Fig. 3I](#pone-0017516-g003){ref-type="fig"}). Coral cover less than 10% occurred on at least one reef in 7 subregions and included inner, mid and outer shelf reefs. Most reefs (8 of 10) had *A. planci* or *A. planci* plus other impacts. Storms were responsible for low coral cover on two outer shelf reefs. On average there were 4 years between disturbances. Small disturbances (\<5% annual change) to coral cover were common. Forty one reefs had between one and five small disturbances within the sampling period that were associated with all disturbance types. Moderate disturbances, defined as declines of between 5--10%, occurred on 27 reefs. Moderate declines were associated with all disturbance types. Large disturbances associated with declines of 10--17% occurred on 18 reefs and were associated with all disturbance types. Very large disturbances with coral declines of greater than 17% occurred on 13 reefs and were associated with *A. planci* predation, storms, and multiple disturbances. There was an average of six years between moderate disturbances and 11 years between large disturbances. Recovery from disturbance {#s2c} ------------------------- Disturbance type did not affect recovery rate with the exception of *A. planci*, after which estimated intrinsic growth rate was 0.22 (0.16, 0.28 95%CI) compared with an average of 0.14 (0.09, 0.19 95%CI) for all other disturbance types ([Fig. 5](#pone-0017516-g005){ref-type="fig"}). Average growth rate in disturbance free years was 3.9% (3.5,4.2 95%CI). The fastest recovery rates were observed on reefs dominated by tabulate *Acropora spp* ([Fig. 3D, P](#pone-0017516-g003){ref-type="fig"}). The seven reefs in these two subregions had low coral at the start of sampling in 1995, with all reefs increasing to 50--75% coral cover. The average growth rate during this period of increase, which lasted between 7 and 16 years, was between 5--11% per year with a maximum annual increase of 19% in one year. Growth periods at other reefs had mean values of 2--5%. The median length of recovery between disturbances was 5 years. ::: {#pone-0017516-g005 .fig} 10.1371/journal.pone.0017516.g005 Figure 5 ::: {.caption} ###### Intrinsic growth rates following different disturbance types. The intrinsic growth rate *r* estimated by fitting a logistic growth model to periods of coral recovery following disturbance events. Growth rates were similar following all disturbance types with the exception of *A. planci*. Carrying capacity *K* was estimated at 80.1%. :::  ::: We identified 36 reefs where either coral cover did not decline or cycles of disturbance and recovery were completed. We found coral cover on 33% (12 reefs) of reefs did not decline, 57% (21 reefs) of reefs declined and recovered, and 8% (3 reefs) of reefs declined without recovery. Eleven reefs were excluded as having had recent disturbance or low starting values for coral cover and low recovery. Of the three reefs with declines, two reefs had additional disturbance before recovery from a prior disturbance was completed. One mid shelf reef (Reef 19131) in the Whitsundays lost coral due to a storm in 1997, and then coral bleaching in 2002 took coral cover to a new low before recovery had been completed. In 2009 coral cover had recovered to 72% of its maximum value of 60% coral cover. One outer shelf reef in the Townsville sector (Myrmidon) was recovering from the 2002 bleaching event when it was damaged by storms in 2007. In 2009 coral cover was 74% of its maximum of 39% coral cover. A number of reefs were still in recovery phases in 2009 and have not reached their pre-disturbance cover. The primary impact on the third reef, Havannah Island, was coral bleaching, which was followed by storms and sub-outbreak numbers of *A. planci*. A shift to macroalgae occurred and coral cover was still below 10% in 2010. Discussion {#s3} ========== This study indicates that at the scale of the whole GBR there was no net decline in live hard coral cover between 1995 and 2009. Rather there have been contrasting and uncorrelated temporal trends in coral cover, driven mostly by Acroporidae corals, at subregional scales (10--100 km) resulting from localized disturbance events. Despite two category 5 cyclones and increased incidence of coral disease and bleaching since 1998, the level of hard coral cover throughout the system as a whole has changed little since 1995. We contend that 27--33% cover of hard coral represents a meaningful and accurate baseline range for average coral cover on the GBR since 1995. The patterns of temporal change within subregions were diverse but the overall pattern is of dynamic stability with the number of increases and decreases being similar both in number of subregions and numbers of reefs. The limited data on coral cover prior to the mid-1980s suggest that coral cover on the GBR declined before our surveys began, and that we are reporting stability in a 'shifted baseline' [@pone.0017516-Pauly1], [@pone.0017516-Knowlton2]. A shifted baseline is a likely outcome if disturbance is too frequent or intense, or if recovery is too slow [@pone.0017516-Bellwood1], [@pone.0017516-Bruno1]. Data from elsewhere in the Pacific prior to the 1980\'s also suggest higher coral cover to what we report here [@pone.0017516-Gomez1]. Regional stasis has been reported for the Caribbean from 1982 to 2006 [@pone.0017516-Schutte1] whereas on the GBR there has been a small decline in coral cover since 1986 [@pone.0017516-Sweatman2]. Estimates of coral cover from entire reef perimeters (using the manta-tow technique) found the GBR mean had declined from 28-22% between 1986 and 2004 [@pone.0017516-Sweatman2]. The lower average from reef perimeter surveys as compared to our result is due to large areas of non-coral habitat (e.g., sandy back-reef lagoons) being included in estimates. The sites we sampled were located on the flank of the reef, an area of active reef accretion and good circulation. We would expect our results to be a 'best case' scenario for the GBR as a whole. This study represents a robust baseline from which to assess future changes, but there is also an obvious need to know whether there is any evidence of changes in the frequency of disturbances. During the period 1995--2009 there were several cycles of the Southern Oscillation Index, which is an important driving factor of disturbance events on the GBR [@pone.0017516-Lough2] but estimating changes in the periodicity of particular disturbance types is not possible with only 16 years of data. We have established a baseline for disturbance of every 4 years. The frequency of moderate (every 6 years) and large disturbances (\>11 years) can be used for assessing relative disturbance pressure at a reef or for modelling changes in disturbance and response through time. The variability of coral decline associated with disturbances of each type indicates the complexity of factors influencing disturbance and response. These include disturbance history and level of coral cover, the type of coral community, as well as environmental factors such as exposure and circulation. A detailed understanding of the variability of these factors is needed to determine if current disturbance regimes are sustainable. The stability of the GBR average coral cover and overall balance of increases and decreases in coral cover at subregional scales suggests the disturbance intervals we have defined are meaningful in the current environment. Since 1995, *A. planci* and storms impacted the largest number of reefs and caused the largest declines in coral cover on the GBR. Since regular monitoring of *A. planci* started in the mid-1980s the periodicity of *A. planci* outbreaks has been approximately 15 years on the central GBR [@pone.0017516-Miller2], with outbreaks in the Townsville sector peaking in 1986 and 2003. Recovery of reef-wide coral cover between outbreaks was variable and model results suggest that the outbreak frequency is too high for adequate coral recovery between outbreaks [@pone.0017516-Lourey1], [@pone.0017516-Fabricius3]. Since the most recent outbreaks, disturbance frequency was high with recovery most likely impaired by bleaching in 2002. This is especially likely on the Townsville mid-shelf where coral declined significantly ([Fig. 3I](#pone-0017516-g003){ref-type="fig"}). Up to 2009, the majority of declines in coral cover were associated with particular disturbances. We attribute the relatively low contribution of coral disease and bleaching to coral declines to relatively low levels of stress (e.g., over-fishing and pollution) in the GBR system. Stressors may be more patchy and localised due to the low anthropogenic pressure relative to other reef regions and the dynamics of the reef matrix which promote water circulation [@pone.0017516-Bruno2]. The number of declines in coral cover associated with 'unknown' agents has increased since 2000, but even if these could be attributed to disease or bleaching, the magnitude of decline would still be low compared to *A. planci* and storms. Surveys in 1998 and 2002 identified large areas of reef where coral was bleached [@pone.0017516-Berkelmans2]. Our results indicate that mortality was not widespread or severe at depths of 6--9 meters. Since the last widespread bleaching event on the GBR in 2002, summer sea surface temperatures have been high in some locations. Both disease and bleaching have clear links with increased temperature and are likely to be important causes of chronic mortality on stressed reefs [@pone.0017516-Lesser1]. The long-term persistence of reefs requires coral communities to recover between episodic disturbance events. While our dataset is unique in its spatial and temporal coverage, 16 years is still a short period in which to document disturbance and recovery cycles, and the specific time frame influences our observations. For reefs where we were able to document cycles of disturbance and recovery, a high percentage (92%) either did not decline or declined and recovered. We found two reefs where additional disturbance interrupted recovery before the pre-disturbance coral cover was reached. Both reefs were back to 70% of their maximum in 2009 and increasing. Recovery was similar across disturbance types except on reefs recovering from *A.planci* that had higher intrinsic growth rates. Similar results were found for other reef regions [@pone.0017516-Graham1] and have been attributed to the maintenance of structural complexity following *A.planci*. While bleaching and disease also leave an intact skeleton, thermal stress associated with these disturbance types impairs reproductive success in *Acropora spp* [@pone.0017516-Negri1], [@pone.0017516-McClanahan2]. The fine-scale complexity of tabulate *Acropora* skeletons may provide protection from grazing in the early stages of growth when pressure can be intense (A. Thompson, pers comm.). Alternately the maintenance of fish diversity in habitats with higher structural complexity [@pone.0017516-Cheal1] may help create suitable substrate for coral recruitment. Multiple disturbances were associated with declines on inner and mid-shelf reefs where coral cover dropped below 10%. The spatial and temporal aggregation of disturbances in the Cairns and Townsville sectors up to 2002 resulted in four inshore reefs and three mid-shelf reefs having low coral cover. Recovery was slow until 2007. On Cairns inshore reefs, shallower sites on the inshore reefs recovered to high coral cover sooner and had a higher proportion of *Acropora spp* driving coral cover change [@pone.0017516-Sweatman1], [@pone.0017516-Thompson2]. Light limitation associated with turbidity is likely to be a factor inhibiting recovery on inshore reefs. Wet season rainfall has been high in recent years and has included flood events that have led to persistent turbidity in inshore environments [@pone.0017516-Schaffelke1]. At one reef (Green Island), recovery from *A. planci* outbreaks prior to 1995 was already poor on deeper sites and slow recovery may be due to recruitment failure [@pone.0017516-Done2]. On Havannah Island, a Townsville inshore reef, persistent macroalgae dominance, indicative of a 'phase shift', is possibly due to low diversity and abundance of herbivorous fishes (e.g., Acanthurids and Siganids) that can prevent the establishment of macroalgae [@pone.0017516-Cheal2]. A better understanding of disturbance risk and recovery potential is required to better manage reefs at local scales. There is substantial knowledge of individual risks for bleaching, cyclones, disease and *A. planci*. Areas least at risk of multiple or intense events should be identified, especially on the mid- and outer shelf where anthropogenic effects not related to climate change should be minimal. It is not known to what extent low coral cover compromises ecosystem functioning, but there are indications that the amount of coral cover influences other species either directly or indirectly. While several studies have found declines in fish diversity when coral declines [@pone.0017516-Jones1], [@pone.0017516-Wilson2], fish counts from AIMS survey reefs indicate that there was no long term loss of fish diversity or evenness, though fish abundance declined following loss of coral cover [@pone.0017516-Cheal3]. The shift from coral to macroalgae at Havannah Island has not resulted in fish species loss, but low starting values for species diversity and abundance may have been a factor in the long-term persistence of macroalgae [@pone.0017516-Cheal2]. Storms destroy the physical substrate, which has more severe consequences for fish abundance and diversity [@pone.0017516-Connell1]. *A. planci*, bleaching and disease have less effect on fish populations, presumably due to the maintenance of topographic complexity [@pone.0017516-Emslie1], but see [@pone.0017516-Wilson1]. The effects of Acroporidae loss are well documented for specialist coral feeders [@pone.0017516-Graham2] but the wider consequences to reef communities are not well known. With the exception of Havannah Island, all reefs that had low coral cover were in recovery phases in 2009, suggesting critical 'tipping points' have not been exceeded [@pone.0017516-Mumby1]. Climate change is expected to cause changes in relative abundance of species due to differential mortality and recovery rates [@pone.0017516-Hughes1]. The susceptibility of Acroporidae (and especially *Acropora* spp) to *A. planci* [@pone.0017516-Pratchett1], bleaching [@pone.0017516-Loya1], and disease [@pone.0017516-Willis1] is well documented. Only one subregion had a significant decline in Acroporidae, suggesting that on most reefs, recruitment, growth and mortality are keeping up with the recent disturbance regime. For non-Acroporidae families, 11 of the 15 subregions had declines, three of which were statistically significant. All but one inshore subregion and all outer shelf subregions had negative trends for non-Acroporidae. Persistent declines are a 'red flag' for managers and researchers to elucidate where functional failures are occurring. The pressures on inshore reefs of the GBR have been well documented [@pone.0017516-Fabricius4], [@pone.0017516-Thompson1] with existing research suggesting coral bleaching and elevated nutrients are adding unsustainable pressure to inshore reefs [@pone.0017516-Death1], [@pone.0017516-Wooldridge1]. Our results suggest that outer shelf reefs are also at risk due to intense disturbance pressure in recent years and large disturbance size, both spatially and in intensity of storms. For slower growing coral, evidence suggests that the current disturbance regime is unsustainable. Ecological health indices calculated from species richness estimates on GBR reefs found that species loss was a feature of depauperate coral communities, rather than a shift in community composition [@pone.0017516-DeVantier1]. Loss or decline of less tolerant species would be consistent with our results and needs further investigation. In conclusion, precise estimates of coral cover from a dedicated monitoring program revealed that system-wide coral cover changed very little on the GBR between 1995 and 2009. Although coral cover averaged 29% across the whole GBR, previous studies indicate that coral cover was higher prior to when our surveys began. Nonetheless, there appears to be no evidence of continued system-wide decline since 1995. During this 16 year period, storms and *A. planci* predation had the largest impact on coral cover, especially at subregional scales (10--100 km), in terms of reefs affected, summed coral lost at all reefs, and amount of decline at individual reefs. The impact of bleaching and coral disease, to date, was not severe on our sites. There are a number of factors however, that suggest that the current disturbance regime may not be sustainable. One inshore reef, for example, had a phase shift from hard coral to macroalgae, similar to that which has occurred at much larger scales in the Carribbean [@pone.0017516-Hughes2]. Corals with less capacity for growth and recruitment than Acroporidae had widespread negative trends. However, the abundance of Acroporidae species and relatively low anthropogenic agents of disturbance appears to place the GBR in a healthier state than the global average. Methods {#s4} ======= This study was approved as part of ongoing research of the Australian Institute of Marine Science. Data were collected under Great Barrier Reef Marine Park Authority Permit G06/19994.1 Sampling {#s4a} -------- Coral communities were surveyed annually between 1995 and 2009, on 47 reefs in six latitudinal sectors (Cooktown-Lizard Is, Cairns, Townsville, Whitsunday, Swain and Capricorn-Bunker) and in three positions on the continental shelf (inshore, mid-shelf and outer shelf) of the GBR. Each combination of latitudinal sector and continental shelf position is referred to as a 'subregion'. There are no substantial reefs inshore in the Swains sector or in inshore and mid-shelf positions in the Capricorn-Bunker sector. Between two and five (usually three) reefs were surveyed in each subregion. Three sites on the north-east flank of each reef were surveyed. Each site consisted of five 50 m transects at depths between 6 m and 9 m that were marked with steel rods. Percent cover of live hard coral was estimated from a randomly selected sequence of images taken along the transects using a point-sampling technique in a quincunx pattern [@pone.0017516-Adbo1]. Corals were identified to a minimum taxonomic resolution of family and to genus where possible. Only 29 reefs were surveyed in the first two years and only 10 were surveyed in the 14^th^ and 16^th^ years due to changes in the scope of the monitoring program. Analyses {#s4b} -------- ### Spatial and temporal trends in total hard coral cover, Acroporidae, and non-Acroporidae {#s4b1} Where appropriate, linear trends in percent live hard coral cover were estimated using mixed-effects linear models with random effects to account for the repeated observations on reefs. Models were built in R (R Development Core Team 2008) with the NLME package [@pone.0017516-Pinheiro1], [@pone.0017516-Pinheiro2]. The GBR-wide trend in coral cover was estimated using a repeated-measures design with a fixed linear time parameter and random time and intercept effects for reefs. Trends for individual subregions were estimated using a model with a fixed linear time parameter for each subregion and random time and intercept effects for reefs. The addition of further random effects to model the spatial arrangement of reefs was explored but did not substantially influence the fitted parameters or the overall goodness of fit. ### Identifying agents of disturbance {#s4b2} Declines in coral cover were attributed to *A. planci* predation, storm and cyclone damage, bleaching, and disease. The attribution of coral decline to a particular agent was based on evidence gathered by trained and experienced personnel during manta tow and SCUBA surveys [@pone.0017516-Miller3]. Each of the four disturbance agents have distinctive and identifiable effects on corals. For example, the presence of *A. planci* and feeding scars indicated *A. planci* predation on live coral, while dislodged and broken coral indicated storm damage[@pone.0017516-Fabricius1]. Coral bleaching also has distinctive effects on live coral compared to disease [@pone.0017516-Miller3]. If no evidence was found, the agent of decline was recorded as unknown. If evidence of more than one agent was present, the cause was recorded as multiple. Although correlative, this approach represents our best estimate of the proximate causes of coral decline and is consistent with approaches used in other studies [@pone.0017516-Fabricius1] [@pone.0017516-Pratchett1]. It is of course possible that other chronic and sublethal disturbances occurred that are not reflected in total coral cover. ### Assessing the impacts of different disturbance agents {#s4b3} The relative importance of each disturbance agent was assessed by comparing their prevalence and severity. Prevalence was assessed by counting the number of reefs affected and the number of reef-years of impact for each agent. For example, if a reef had bleaching for two years, then reef-years is recorded as two for that reef. Severity was estimated as the decline in the percent cover of live coral from pre-event cover. Since absolute loss in coral cover is positively correlated with the initial coral cover, coral losses were expressed as the proportion of pre-existing cover. Individual storms and mass bleaching events last from a few hours to a few weeks. Outbreaks of *A. planci* or diseases can cause coral mortality at a reef over several years. Where a disturbance lasted multiple years, cover loss was the total change in cover during that period as a proportion of the initial cover. The overall importance of each agent of decline was determined by summing the proportional losses across reefs and years. This sum of proportional losses from individual events gives a good representation of the relative importance of the agents and should not be biased by the sequence in which impacts occur. However, proportional coral loss may still be related to pre-event cover, as the highest cover values may be associated with more developed coral structures that, depending on the growth form of the corals, can be more susceptible to wave damage [@pone.0017516-Madin1]. Accordingly, pre-event levels of coral cover were also examined to see whether they differed across disturbance types. Mean rates of change for hard coral were estimated from annual rates of change from each reef and year. Disturbance frequency was estimated from single year impacts. Small, moderate and large disturbances were defined by percentiles of the mean rate of annual coral decline (50^th^  = 5%, 75^th^ = 10%, 90^th^ = 17%). Change under 2% was excluded because this is within the range of sampling error [@pone.0017516-Davidson1]. ### Recovery from disturbance {#s4b4} We used the same criteria as Connell [@pone.0017516-Connell1] to look at how many reefs have declined or recovered. Regression trees [@pone.0017516-Death2] were used to identify periods of consistent increase or decrease in hard coral cover, constrained in a way so that years in each grouping must be a sequence (e.g., years 2, 3 and 4 can be grouped but 2, 3 and 7 cannot). Tree sizes (number of periods of consistent change identified) were determined by cross-validation using data from the three sites at each survey reef. If coral cover at a reef decreased proportionally by 33% from its pre-disturbance cover, a decline was recorded. If coral cover recovered to at least 50% of its pre-disturbance level it was recorded as recovered. Reefs that had recent disturbances, or had low starting values and had not recovered to the GBR average of 29% were excluded. To determine if disturbance type influenced subsequent recovery a logistic growth model was fitted to individual annual changes in total coral cover for 5 years post disturbance, allowing the intrinsic growth rate to vary according to the previous disturbance type. The logistic model was fit as a non-linear mixed effects model using the NLME package for R, with fixed effects on the intrinsic growth rate, *r,* according to previous disturbance type and a random effect on growth rate at the level of individual reefs. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### **Temporal trends in percent cover of Acroporidae and non-Acroporidae on the Great Barrier Reef (1995--2009).** Average annual cover of Acroporidae and non-Acroporidae coral in each subregion (B--D). Straight lines show the average linear trend. Disturbances associated with coral decline are represented by a dot for each reef where that type of disturbance occurred, as in [Figure 3](#pone-0017516-g003){ref-type="fig"}. (EPS) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S1 ::: {.caption} ###### Linear trends in coral cover (1995--2009). (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: We thank all past and present members of the Australian Institute of Marine Science Long-term Monitoring Program who contributed to the collection of the data used here. We thank John Bruno and an anonymous reviewer for comments that improved the manuscript. We also thank Hugh Sweatman for editorial assistance and Angus Thompson for assistance with improving the figures. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This study was supported by the Australian Institute of Marine Science and by The Australian Federal Government\'s Marine and Tropical Scientific Reseach Facility (Project 1.1.2; [www.rrrc.org.au](http://www.rrrc.org.au)). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: KO AD SB. Performed the experiments: KO SB KJ. Analyzed the data: KO AD. Contributed reagents/materials/analysis tools: KO AD SB KJ. Wrote the paper: KO AD SB. [^2]: ¤a Current address: Department of Freshwater Conservation, Brandenburg University of Technology, Brandenberg, Germany [^3]: ¤b Current address: School of Biological Sciences, University of Queensland, Brisbane, Australia

PubMed Central

2024-06-05T04:04:19.750006

2011-3-10

PMC3053362

Introduction {#s1} ============ Global warming is a classic "wicked problem." [@pone.0017571-Rittel1] Wicked problems have no easy solutions in that they are beyond the capacity of any one organization to solve, and there is disagreement among organizations about both the causes and the best means by which to solve the problem [@pone.0017571-Australian1]. Managing wicked problems requires working successfully within and across organizational boundaries, engaging citizens and other stakeholders in policy-making and implementation of those policies, and ultimately changing the behavior of groups of citizens or all citizens [@pone.0017571-Australian1], [@pone.0017571-Conklin1]. Successfully mitigating and adapting to global warming will require significant modifications in public policy and population behavior [@pone.0017571-Leiserowitz1]. Public engagement campaigns are an important strategy to encourage population behavior change and build support for appropriate public policies [@pone.0017571-Maibach1]--[@pone.0017571-Ockwell1]. Many factors limit the success of engagement campaigns, however, some of them inherent (e.g., the myriad influences on human behavior that are largely beyond the reach of a communication campaign) [@pone.0017571-Maibach1], [@pone.0017571-Moser1], [@pone.0017571-Ockwell1] and others situational (e.g., the tendency of governments to prematurely terminate public engagement campaigns) [@pone.0017571-Moser1], [@pone.0017571-Akerlof1]. Although the research literature on global warming communication campaigns is relatively new and not yet well developed [@pone.0017571-Maibach1], [@pone.0017571-Moser1], other fields including commercial marketing [@pone.0017571-Kotler1], social marketing [@pone.0017571-Kotler2], public health [@pone.0017571-Hornik1] and political science [@pone.0017571-Sosnik1] offer considerable research on the attributes of effective public engagement campaigns. Audience segmentation is one of the methods widely supported in all of these diverse research literatures. Audience segmentation is a process of identifying groups of people within a larger population who are homogeneous with regard to critical attributes (e.g., beliefs, behaviors, political ideology) that are most relevant to the objectives of a public engagement campaign (e.g., product sales, consumer boycotts, political participation) [@pone.0017571-Slater1]. Audience segmentation research -- conducted insightfully -- provides organizations with an important strategic planning asset: empirical information about how best to focus the organization\'s limited resources, both human and financial, to advance its objectives [@pone.0017571-Dibb1]. For example, a smaller audience segment whose members are willing to behave in ways sought by the organization may be a more productive target than a larger, less predisposed audience segment. The principal aim of our current research was to identify audience segments within the American adult population that could be considered as potential targets for global warming public engagement campaigns. The nature of the global warming public engagement challenge -- i.e., the need to build public understanding and support for appropriate public policies, and to change the behavior of large numbers of people -- necessitated that we adapt and extend previously used segmentation methods. Specifically, there is strong precedent in the research literature for segmenting audiences based on what people are doing (i.e., behaviors) and why (i.e., motivations) [@pone.0017571-Maibach3]--[@pone.0017571-McDonald1]. That method is well suited to population behavior change campaigns (e.g., smoking cessation campaigns), but it largely ignores a second potential focus for global warming public engagement campaigns: building public understanding of and support for appropriate public policies. Here, we extend the method of segmenting audiences based on what people are doing and why to also include people\'s policy preferences as an additional dimension in the analysis. The other aim of our research was to develop an easily implemented, survey-based identification tool that can be used to identify the audience segments in independent population samples with acceptable levels of accuracy. Such a tool will enable social science researchers and public engagement campaign planners to further study the audience segments identified in our research, and to test public engagement methods with them. We believe that both aims of our research were achieved. Results {#s2} ======= We conducted a nationally representative survey of adults (n = 2,164) and used three major categories of variables as inputs into a segmentation analysis: global warming motivations, behaviors, and policy preferences. The global warming motivations category included two distinct sub-categories: beliefs about global warming and degree of involvement in the issue. We measured a total of 36 variables across these four categories ([Tables 1](#pone-0017571-t001){ref-type="table"}, [2](#pone-0017571-t002){ref-type="table"}, [3](#pone-0017571-t003){ref-type="table"} and [4](#pone-0017571-t004){ref-type="table"}). To maximize the practical value of the segmentation findings, we limited the analysis to five, six and seven segment solutions. As described in the [Methods](#s4){ref-type="sec"} section below, we determined that the six-segment solution was optimal. ::: {#pone-0017571-t001 .table-wrap} 10.1371/journal.pone.0017571.t001 Table 1 ::: {.caption} ###### Global Warming Beliefs by Audience Segment. ::: {#pone-0017571-t001-1} Survey Questions Audience Segment Scale Points --------------------------------------------------------------------------------- ------------------ -------------- ------ ------ ------ ------ ------- 1\. & 1a. Certainty global warming is occurring 8.70 7.92 6.54 5.91 5.06 3.06 9 2\. Human causation (% agree) 88 79 49 39 8 1 \-\-- 3\. Scientific consensus (% agree) 80 64 37 23 11 8 \-\-- 4\. Personal risk 3.09 2.59 1.90 2.75 1.29 1.02 4 5\. Risk to future generations 3.98 3.78 2.96 4.00 1.89 1.04 4 6\. Risk to plant & animal species 3.97 3.78 3.00 3.40 1.94 1.12 4 7\. Timing of harm to Americans 5.46 4.83 3.53 3.85 1.77 1.01 6 8\. Ability of humans to successfully mitigate warming 3.90 3.74 3.45 3.38 2.33 1.57 5 9\. Actions of individual can make a difference 3.36 3.07 2.69 2.76 2.35 1.86 4 10\. Technological optimism 1.70 2.05 2.32 2.03 2.38 2.33 4 11\. Perceived impact of own mitigation actions 2.94 2.72 2.31 2.41 1.53 1.02 4 12\. Impact of own actions if widely adopted in United States 3.69 3.48 3.01 2.90 1.94 1.10 4 13\. Impact of own actions if widely adopted in modern industrialized countries 3.84 3.76 3.34 3.24 2.27 1.18 4 (*p*\<.001 for all differences). ::: ::: {#pone-0017571-t002 .table-wrap} 10.1371/journal.pone.0017571.t002 Table 2 ::: {.caption} ###### Global Warming Issue Involvement by Audience Segment. ::: {#pone-0017571-t002-2} Survey Questions Audience Segment Scale Points ----------------------------------------------------- ------------------ -------------- ------ ------ ------ ------ --- 14\. Rating of global warming (good = 1 to bad = 6) 5.72 5.31 4.35 4.04 3.66 3.19 6 15\. Worry about global warming 3.65 3.08 2.44 2.31 1.56 1.12 4 16\. Thought given to global warming 3.65 2.75 2.22 1.71 2.19 2.82 4 17\. Need for information (4  =  low need) 2.74 2.16 1.89 1.60 2.50 3.58 4 18\. Personal importance of issue 4.44 3.39 2.59 2.54 1.81 1.38 4 19\. Unwilling to change opinion 3.77 2.95 2.41 2.16 3.02 3.69 5 20\. Personally experienced global warming 2.92 2.26 1.95 1.96 1.52 1.19 4 21\. Global warming discussion frequency 3.02 2.36 1.86 1.29 1.88 2.05 4 22\. Friends share views on global warming 3.59 2.71 2.21 1.65 2.85 3.61 5 (*p*\<.001 for all differences). ::: ::: {#pone-0017571-t003 .table-wrap} 10.1371/journal.pone.0017571.t003 Table 3 ::: {.caption} ###### Global Warming and Energy Use Behaviors by Audience Segment. ::: {#pone-0017571-t003-3} Survey Questions Audience Segment Scale Points ------------------------------------------------------------------ ------------------ -------------- ------ ------ ------ ------ ---- 14\. Contacted govt. officials re mitigation 1.53 1.11 1.07 1.07 1.06 1.00 5 15\. Rewarded companies that reduced emissions 3.34 2.18 1.50 1.38 1.31 1.19 5 16\. Intend to reward companies that reduce emissions 2.76 2.51 2.17 2.14 2.06 1.92 3 17\. Punished companies that are not reducing emissions 3.14 1.92 1.32 1.28 1.18 1.08 5 18\. Intend to punish companies that are not reducing emissions 2.73 2.51 2.13 2.18 2.03 1.79 3 19\. Stage of change for lowering thermostat in winter 7.02 6.50 5.99 5.74 6.21 6.18 10 20\. Stage of change for using public transportation or car pool 3.92 3.06 2.74 3.14 2.11 2.27 10 21\. Stage of change for walking/biking instead of driving 4.73 3.49 3.14 2.59 2.68 2.72 10 22\. Stage of change for CFL use 3.49 3.26 2.86 2.97 2.71 2.40 4 (*p*\<.001 for all differences). ::: ::: {#pone-0017571-t004 .table-wrap} 10.1371/journal.pone.0017571.t004 Table 4 ::: {.caption} ###### Preferred Societal Responses by Audience Segment. ::: {#pone-0017571-t004-4} Survey Questions Audience Segment Scale Points --------------------------------------------------------------------------------------------------------- ------------------ -------------- ------ ------ ------ ------ ----- 23\. Priority of global warming for president & Congress 3.54 2.89 2.29 2.57 1.54 1.11 4 24\. Corporations should do more/less to reduce warming 4.81 4.37 3.93 3.62 3.07 2.01 4 25\. Citizens should do more/less to reduce warming 4.75 4.23 3.74 3.58 3.03 1.97 4 26\. Desired US effort to reduce warming, given associated costs 3.78 3.33 2.89 2.83 2.01 1.37 4 27\. Contingent int\'l conditions for US mitigation action (% regardless of actions of other countries) 98 93 74 84 59 40 \-- (*p*\<.001 for all differences). ::: The six identified segments -- each of which was given a concise name to summarize its essential qualities -- differ dramatically with regard to what they believe about global warming, how engaged they are with the issue, what they are doing about it, and what they would like to see American government officials, businesses, and citizens do about it. The six segments also differ dramatically with regard to size: the largest represents 33% of the U.S. adult population, and the smallest only 7% ([Figure 1](#pone-0017571-g001){ref-type="fig"}). These six audience segments represent a spectrum of concern and action about global warming, ranging from the Alarmed (18% of the population), to the Concerned (33%), Cautious (19%), Disengaged (12%), Doubtful (11%) and Dismissive (7%). ::: {#pone-0017571-g001 .fig} 10.1371/journal.pone.0017571.g001 Figure 1 ::: {.caption} ###### Proportion of the U.S. adult population in the Six Americas. :::  ::: Mean values for (or in the case of three variables, percent agreement with) each of the variables used in the segmentation analysis, by segment, are presented in [Tables 1](#pone-0017571-t001){ref-type="table"}, [2](#pone-0017571-t002){ref-type="table"}, [3](#pone-0017571-t003){ref-type="table"} and [4](#pone-0017571-t004){ref-type="table"}. The between-segment differences on all of these variables, as ascertained by ANOVA or chi-square tests, were significant at p\<.001. Additional profiling information about the audience segments -- i.e., how the six segments differ with regard to a range of additional relevant beliefs, behaviors (including media use), values, and demographics -- is available at: <http://environment.yale.edu/climate/publications/global-warmings-six-americas-2009/>. In brief, the *Alarmed* are the segment most engaged in the issue of global warming. They are very convinced it is happening, human-caused, and a serious and urgent threat. The Alarmed are already making changes in their own lives and support an aggressive national response. The *Concerned* are also convinced that global warming is a serious problem, but while they support a vigorous national response, they are distinctly less involved in the issue, and less likely than the Alarmed to be taking personal action. The *Cautious* also believe that global warming is a problem, although they are less certain that it is happening than the Alarmed or the Concerned. They don\'t view it as a personal threat, and don\'t feel a sense of urgency to deal with it through personal or societal actions. The *Disengaged* haven\'t thought much about the issue. They are the segment most likely to say that they could easily change their minds about global warming, and they are the most likely to select the "don\'t know" option in response to every survey question about global warming where "don\'t know" was presented as an option. The *Doubtful* are evenly split among those who think global warming is happening, those who think it isn\'t, and those who don\'t know. Many within this group believe that if global warming is happening, it is caused by natural changes in the environment, that it won\'t harm people for many decades into the future, if at all, and that America is already doing enough to respond to the threat. Finally, the *Dismissive*, like the Alarmed, are actively engaged in the issue, but on the opposite end of the spectrum. The large majority of the people in this segment believe that global warming is not happening, is not a threat to either people or non-human nature, and is not a problem that warrants a personal or societal response. To validate the predictive utility of these audience segments, we conducted four regression analyses using demographics (i.e., age, household income, gender, marital status, employment status, and race/ethnicity), political ideology, and segment membership as predictors of an outcome measure. A scale measuring support for nine specific potential federal greenhouse gas emission reduction policies was used as the outcome measure; these specific policy support measures are distinct from the preferred societal response measures used in the segmentation analysis, which are more general in nature (see [Table 5](#pone-0017571-t005){ref-type="table"}). As shown in [Table 6](#pone-0017571-t006){ref-type="table"}, demographics (Model 1, F = 2.8; p\<.01), political ideology (Model 2, F = 267; p\<.001) and segment status (Model 3, F = 1,411; p\<.001) are each significant predictors of policy support when assessed in isolation of each other. Conversely, when assessed simultaneously (Model 4), demographic variables are not significant predictors, political ideology is a significant predictor with a moderately sized beta coefficient (B = .10; p\<.001) and audience segment status is a significant predictor with a large beta coefficient (B = .60; p\<.001). Audience segment alone explains as much variance in policy preferences (41%), as do demographics, political ideology and audience segment combined. We interpret these findings as validation of the predictive validity of the audience segmentation. ::: {#pone-0017571-t005 .table-wrap} 10.1371/journal.pone.0017571.t005 Table 5 ::: {.caption} ###### Support for Emission Reduction Policies by Audience Segment. ::: {#pone-0017571-t005-5} Survey Questions Audience Segment --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ------------------ ------ ------ ------ ------ ------ 1\. Establish a special fund to help make buildings more energy efficient and teach Americans how to reduce their energy use. This would add a \$2.50 surcharge to the average household\'s monthly electric bill. 3.25 2.91 2.48 2.54 2.09 1.56 2\. Provide a government subsidy to replace old water heaters, air conditioners, light bulbs, and insulation. This subsidy would cost the average household \$5 a month in higher taxes. Those who took advantage of the program would save money on their utility bills. 3.44 3.07 2.81 2.79 2.23 1.78 3\. Regulate carbon dioxide (the primary greenhouse gas) as a pollutant. 3.67 3.22 2.93 2.86 2.43 1.84 4\. Require electric utilities to produce at least 20% of their electricity from wind, solar, or other renewable energy sources, even if it cost the average household an extra \$100 a year. 3.50 3.14 2.76 2.60 2.36 2.10 5\. Sign an international treaty that requires the United States to cut its emissions of carbon dioxide 90% by the year 2050. 3.51 3.07 2.64 2.68 1.98 1.49 6\. Require automakers to increase the fuel efficiency of cars, trucks, and SUVS, to 45 mpg, even if it means a new vehicle will cost up to \$1,000 more to buy. 3.64 3.32 3.12 2.73 2.68 2.33 7\. Fund more research into renewable energy sources, such as solar and wind power. 3.84 3.57 3.31 3.16 3.14 2.96 8\. Provide tax rebates for people who purchase energy-efficient vehicles or solar panels. 3.60 3.33 3.12 2.78 2.91 2.60 9\. Increase taxes on gasoline by 25 cents per gallon and return the revenues to taxpayers by reducing the federal income tax. 2.50 2.14 2.00 1.97 1.69 1.37 10\. Policy support index (mean of 9 measures; α  = .86) 3.44 3.09 2.80 2.68 2.39 2.00 (All items measured on 4-point scales, where 1  =  strongly oppose & 4  =  strongly support; *p*\<.001 for all differences). ::: ::: {#pone-0017571-t006 .table-wrap} 10.1371/journal.pone.0017571.t006 Table 6 ::: {.caption} ###### Policy Support Predicted by Socio-Demographics, Political Orientation & Audience Segment. ::: {#pone-0017571-t006-6} Model 1:Socio-demographics Model 2:Political orientation Model 3:Audience segment Model 4:Full model --------------------------------------------- ---------------------------------------- -------------------------------------------- ---------------------------------------------- -------------------------------------------- Age .01 .01 Education .06[\*](#nt106){ref-type="table-fn"} .00 Household Income .00 .01 Gender (2  =  F) .05[\*](#nt106){ref-type="table-fn"} -.02 Marital status (2  =  married or w/partner) -.02 .01 Work status (2  =  working) -.02 -.02 Race: white -.14 -.07 Race: black -.06 -.06 Race: Hispanic -.01 -.04 Race: other -.04 -.05 Political ideology (5  =  very liberal). .33[\*\*\*](#nt108){ref-type="table-fn"} .10[\*\*\*](#nt108){ref-type="table-fn"} Audience segment(6  =  Alarmed) .64[\*\*\*](#nt108){ref-type="table-fn"} .60[\*\*\*](#nt108){ref-type="table-fn"} Adjusted R^2^ .01 .12 .41 .41 F 2.8[\*\*](#nt107){ref-type="table-fn"} 266.8[\*\*\*](#nt108){ref-type="table-fn"} 1,411.7[\*\*\*](#nt108){ref-type="table-fn"} 120.8[\*\*\*](#nt108){ref-type="table-fn"} N 2,067 2,052 2,062 2,052 \*p\<.05; \*\*p\<.01; \*\*\*p\<.001. Note: Cell entries are standardized regression weights. For dummy variables, the excluded race category was "mixed race, non-Hispanic." ::: To enable identification of segment status with new, independent samples, we created an identification tool based on a linear discriminant function of all 36 variables used in the segmentation analysis. This identification tool -- termed the "full discriminant model tool" -- correctly classified 90.6% of the sample (ranging from 79 to 99% in the six segments; see [Table 7](#pone-0017571-t007){ref-type="table"}). We also developed a shorter, more practical 15-item identification tool by eliminating the 20 least predictive variables from the discriminant function. This short identification tool -- termed the "reduced discriminant model tool" -- when applied to our dataset, correctly classified 83.8% of the sample (ranging from 60 to 97% in the six segments). ::: {#pone-0017571-t007 .table-wrap} 10.1371/journal.pone.0017571.t007 Table 7 ::: {.caption} ###### Prevalence of Audience Segments in 2008 Based on Three Methods of Identification. ::: {#pone-0017571-t007-7} Segment Latent Class Analysis FullDiscriminant Model ReducedDiscriminant Model ---------------- ----------------------- ------------------------ --------------------------- ------- ------- 1\. Alarmed 18.0% 18.0% 92.6% 17.1% 85.6% 2\. Concerned 33.3% 33.4% 91.3% 33.5% 85.8% 3\. Cautious 18.7% 17.6% 87.5% 18.0% 80.9% 4\. Disengaged 12.2% 13.6% 98.9% 14.9% 96.7% 5\. Doubtful 10.6% 9.5% 79.2% 8.0% 60.1% 6\. Dismissive 7.2% 8.0% 93.2% 8.5% 89.9% ::: Discussion {#s3} ========== With this research, we set out to identify and validate an audience segmentation system that can be used to inform global warming public engagement campaigns, and to develop easy-to-use survey-based identification tools that can be used to replicate our results with acceptable levels of accuracy. Both aims were achieved with a large representative sample. To be useful in supporting public engagement campaigns, a market segmentation scheme must demonstrate five attributes: (1) segments must be distinct from one another, and members of each segment must be sufficiently similar to be effectively targeted by the same marketing strategy; (2) segments must have direct relevance to the campaign objectives being pursued; (3) segments must be large enough to justify the time and effort required to target them; (4) the segment status of individuals in the market must be identifiable; (5) the campaign organization -- or organizations -- must be capable of targeting one or more of the identified segments (which may involve making the necessary changes to its structure, information and decision-making systems) [@pone.0017571-McDonald1]. The audience segments we identified possess the first four of these five attributes. The six segments -- all of which are substantial in size, and whose members can be identified with the tools we developed -- are distinct from one another in ways that have direct bearing on efforts to promote global warming mitigation and adaptation. The last of these five attributes, ultimately, is demonstrated by whether or not campaign organizations find value in making campaign decisions using the segmentation system. In the following paragraphs, we briefly elaborate on how global warming campaign organizations might select among the six audiences identified. Members of the *Alarmed* segment are a highly engaged and active audience, at least in their capacity as consumers (with the exception of their travel behavior, which is more-or-less similar to that of other segments). They have a strong demonstrated tendency to use their consumer purchasing power to reward businesses they believe are contributing to solutions, and punish businesses they believe are not. They are markedly less active in their role as citizens, however; only about one in four had contacted an elected official in the past year to urge them to take action to reduce global warming. Organizations seeking to promote policy advocacy -- and possibly those seeking to modify people\'s travel behavior \-- should consider targeting this audience. Members of the *Concerned* segment are moderately engaged in the issue, but they are less active than are the *Alarmed*. As a result of their high prevalence in the population (1 out of every 3 adults), and their high stated intention to use their consumer purchasing power more frequently in the future to reward businesses they believe are contributing to solutions, organizations seeking to promote change through markets -- rather than, or in addition to, change through public policy -- should consider targeting this audience. Members of the *Cautious* segment are only modestly engaged in the issue, and they don\'t appear ready to take action either as consumers or citizens. Organizations that are interested in expanding the number of Americans who are actively considering the issue of climate change (rather than attempting to change people\'s behavior, or develop support for policy responses) should consider targeting members of this audience. Narrative-based communication [@pone.0017571-Dahlstrom1], and reframing the issue in terms of human health may be productive approaches [@pone.0017571-Maibach5]. Members of the *Disengaged* segment currently have no involvement in the issue. The *Disengaged* stand apart from other segments in that they are less educated and have lower household incomes, both of which place them at higher than average risk of being harmed by global warming [@pone.0017571-Protecting1]. This is a difficult segment to reach using news media and other traditional science communication channels, both due to their current lack of interest and their financial challenges. Organizations seeking to engage members of the *Disengaged* must think creatively about how to make the issue more relevant for them. As with the *Cautious* segment, narrative-based communication, and reframing the issue in terms of human health may be productive approaches. Activating new voices to explain the relevance of climate change -- such a health professionals [@pone.0017571-Maibach5], members of the faith community [@pone.0017571-Hitzhusen1], and organizations serving low-income families -- may be helpful as well. Members of the *Doubtful* segment are important because -- although they currently doubt that global warming is real or harmful, and are disinclined to support actions to address it -- they remain open to learning more about this issue. Because the *Doubtful* tend to be politically conservative, organizations that have the ability to work effectively across the political spectrum should consider developing activities to further engage the *Doubtful*. As a result of their strongly held belief that global warming is not happening or is not human caused, members of the *Dismissive* segment are highly involved in the issue as adamant opponents to taking any form of action against global warming. Like members of the *Alarmed* segment, however, they are supportive of taking both personal and societal actions to reduce energy use. Thus, while they are likely not a productive audience for a global warming public engagement campaigns per se, they may be an attractive audience for energy-efficiency campaigns because they are receptive to such appeals. It is important to note that the three classes of variables included in our segmentation -- motivations, behaviors, and policy preferences -- did not include structural and contextual factors (e.g., the availability of public transportation options, and local or state government incentives to reduce energy use) that previous research has shown to be important in influencing adoption of energy efficiency and conservation actions [@pone.0017571-Dietz1]. The implications of this decision are evident in the fact that the between segment differences on energy use and conservation actions are relatively small (albeit significant), whereas the between segment differences on global warming advocacy actions are more pronounced (see [Table 3](#pone-0017571-t003){ref-type="table"}). Thus, this segmentation system is optimized for efforts to educate or engage the public about global warming per se, and less optimized for campaigns intended to promote changes in energy use behavior. An integral part of strategic planning for a public engagement campaign involves selecting the target audiences that are the best fit for the organization\'s public engagement goals and resources [@pone.0017571-Smith1]. Depending on their goals and resources, some organizations might be well served to focus their entire effort on a single target audience. Other organizations might be best served by targeting several audiences, if feasible. Regardless, campaigns that target specific audiences and tailor their materials accordingly are more likely to achieve their public engagement objectives than campaigns that do not [@pone.0017571-Noar1]. For any given organization, the optimal target audiences are those that are likely to maximize the return on investment in campaign planning and execution. The three most relevant considerations in making that determination are the size of the audience segment, the likelihood that members of the segment will respond in the intended manner, and the organization\'s ability reach that segment with its current resources [@pone.0017571-Andreasen1]. It remains to be seen whether or not organizations involved in global warming public engagement campaigns will be capable of -- or interested in -- targeting one or more of the identified segments in the ways we describe. A number science-based organizations -- including science academies [@pone.0017571-The1], science museums [@pone.0017571-Phipps1], and natural resource and conservation organizations -- are currently considering their targeting and tailoring options using the audience segments identified. These developments may be evidence that the method possesses the final necessary attribute of utility: organizations must be capable of targeting one or more of the identified segments [@pone.0017571-McDonald1]. The "one size fits all" approach to global warming communication appears to be the default mode for most organizations, despite the fact that non-targeted approaches are at odds with best practices in campaign management [@pone.0017571-Smith1]. National global warming education campaigns, for example, tend not to target well-defined audiences, but focus instead on the general public [@pone.0017571-Akerlof1]. That non-targeted approaches remain common suggests that many organizations can\'t -- or aren\'t willing to -- bear the added costs of a targeted approach. Non-targeted campaigns are, without question, easier to implement than targeted campaigns. Segmenting and targeting multiple audiences can involve making changes to the organization\'s structure, information and decision-making systems [@pone.0017571-McDonald1]. At very least, a sustained effort to understand and engage more than one distinct target audience requires a campaign team to divide its planning and program development activities among each audience under consideration. A more intensive approach involves creating a campaign team to focus on each targeted audience [@pone.0017571-Andreasen1]. These more intensive methods are common in consumer marketing organizations, yet they remain largely unknown or underutilized outside of the for-profit sector. To monitor the stability of the audience segments identified in this research over time, we used the 36-item tool on three subsequent national surveys conducted in 2009 and 2010. Pronounced shifts in the size of the segments were evident across the three years of these surveys; for example, the Alarmed segment contracted sharply and the Dismissive segment grew markedly between fall 2008 and late 2009, but both regressed somewhat toward their prior sizes by mid-2010 [@pone.0017571-Leiserowitz2]. We are currently exploring the reasons for these shifts, but our preliminary investigations suggest that meaningful exogenous factors -- including a pronounced downturn in the economy, negative media coverage associated with the illegal release of email between climate scientists which became known as "Climategate," and escalation of industry funded global warming denial campaigns [@pone.0017571-Leiserowitz3] -- were responsible for the shifts rather than inherent instability in the segmentation method. Indeed, two of these national surveys were conducted within one month of each other [@pone.0017571-Leiserowitz2], [@pone.0017571-Leiserowitz4]. Results from these surveys show only small differences between segment sizes when they are measured more-or-less contemporaneously: Alarmed, 13 vs. 14%; Concerned, 28 vs. 31%; Cautious, 24 vs. 23%; Disengaged, 10 vs. 10%; Doubtful, 12 vs. 12%; and Dismissive, 12 vs. 11%. Our 15- and 36-item survey-based audience segment identification tools -- as well as SAS & SPSS syntax to process the data -- are available at: <http://www.climatechangecommunication.org/SixAmericasManual.cfm>. We encourage global warming campaign organizations and social science researchers to examine and evaluate them for their potential utility. To assess the robustness of this method to cultural and other contexts, we particularly encourage social science researchers to adapt these tools and assess their validity in nations other than the U.S. Materials and Methods {#s4} ===================== Survey Method {#s4a} ------------- In September through October of 2008, we conducted a nationally representative survey of American adults using KnowledgePanel, an online panel operated by Knowledge Networks. Recruited nationally using random-digit dialing (RDD) telephone methodology, KnowledgePanel is representative of the U.S. population. The panel tracks closely the December 2007 Current Population Survey (published jointly by the U.S. Census Bureau and the Bureau of Labor Statistics) on age, race, Hispanic ethnicity, geographic region, employment status, and other demographic variables. The length of our questionnaire -- a 50-minute completion time for the average respondent -- exceeded what most respondents are willing to answer in a single session. As a result, we divided the content of the instrument into two separate questionnaires. An invitation to participate in the first survey was emailed to 3,997 randomly selected panel members. A total of 2,496 completed the questionnaires, a 62% cooperation rate. Two weeks after administration of the first survey was ended, respondents to the first survey received an invitation to participate in the second survey. Completed questionnaires were received from 2,164 respondents, an 87% cooperation rate, leading to an overall 54% within panel completion rate for the study. The period of administration for each survey -- from invitation to termination of data collection -- was approximately 10 days, during which one reminder email was sent to non-respondents. To reduce the effects of any non-response and non-coverage bias in the overall panel membership, a post-stratification adjustment was applied using demographic distributions from the most recent data from the Current Population Survey (CPS). Benchmark distributions for Internet Access among the U.S. population of adults are obtained from KnowledgePanel recruitment data since this measurement is not collected as part of the CPS. The post-stratification variables were: Gender (Male/Female); Age (18--29, 30--44, 45--59, and 60+); Race/Hispanic ethnicity (White/Non-Hispanic, Black/Non-Hispanic, Other/Non-Hispanic, 2+ Races/Non-Hispanic, Hispanic); Education (Less than High School, High School, Some College, Bachelor and beyond); Census Region (Northeast, Midwest, South, West); Metropolitan Area (Yes, No); Internet Access (Yes, No). Measures {#s4b} -------- We measured a total of 306 variables with the two instruments; 36 of those variables were used in the audience segmentation analysis. Specifically, the 36 items were developed to assess four categories of global warming- and energy-related constructs: global warming beliefs ([Table 1](#pone-0017571-t001){ref-type="table"}), global warming issue involvement ([Table 2](#pone-0017571-t002){ref-type="table"}), global warming and energy efficiency and conservation behaviors ([Table 3](#pone-0017571-t003){ref-type="table"}), and preferred societal response to global warming ([Table 4](#pone-0017571-t004){ref-type="table"}). An index of support for nine specific federal greenhouse gas reduction policies was constructed and used to assess the validity of the segmentation results ([Table 5](#pone-0017571-t005){ref-type="table"}). Segmentation Analysis {#s4c} --------------------- To identify the audience segments, the 36 variables were subjected to Latent Class Analysis using LatentGold 4.5 software [@pone.0017571-Magidson1], [@pone.0017571-Magidson2]. LCA is a modeling technique for analyzing case level data with the objective of identifying groups of respondents (segments or latent classes) with similar characteristics. LCA assigns cases into clusters using model-based posterior membership probabilities estimated by maximum likelihood methods. One advantage of LCA is it can handle nominal, ordinal, and continuous variables as well as any combination of these [@pone.0017571-Magidson1]. In addition, unlike cluster analysis, LCA is not highly sensitive to missing data. Respondents with 80% or more complete data on the 36 variables were included in the analysis; this resulted in a sample size of 2,129 for modeling purposes. The 36 variables in our model were a mixture of ordinal and nominal variables. We submitted five, six, and seven segment solutions to the analyses. One potential problem in estimating latent class models is the possibility of obtaining a local maximum solution rather than a globally-based solution: an estimation algorithm may converge on a local maximum solution, which is the best solution in a neighborhood of the parameter space, but not necessarily the best global maximum. As models become more complex this problem increases. To guard against local maximum solutions, the estimation algorithm should be run several times with different parameter start values. Thus, to address this issue and to ensure the validity and stability of the findings, we conducted the analyses using 5,000 random sets of start values and replicated each solution to ensure model stability. All three models (5-, 6-, and 7-segments) replicated exactly. The three models had similar fit statistics (see [Table 8](#pone-0017571-t008){ref-type="table"}). We examined the profile data for all three models and determined that the six-segment solution offered the highest face validity. Although the seven-segment solution had slightly lower fit statistics (which indicates a better model fit), the difference was small and the six segments described above were more interpretable. ::: {#pone-0017571-t008 .table-wrap} 10.1371/journal.pone.0017571.t008 Table 8 ::: {.caption} ###### Model Fit Statistics. ::: {#pone-0017571-t008-8} L^2^ BIC(L^2^) Npar P(L^2^) ----------- ------------ ------------ ------ --------- 5 classes 146560.858 133330.136 402 \<.0001 6 classes 145443.384 132695.443 465 \<.0001 7 classes 144595.960 132330.799 528 \<.0001 ::: To create an easy-to-use tool for others to categorize survey respondents in new, independent samples, we created a linear discriminant function using the output from the Latent Class Analysis. In contrast to Latent Class Analysis, discriminant analysis does not permit missing data. We therefore used mean substitution for missing data, and then applied this linear function using all 36 variables to our data set. The 36 variable linear function correctly classified 90.6% of the sample (ranging from 79 to 99% in the six segments; see [Table 7](#pone-0017571-t007){ref-type="table"}). Elsewhere in this paper we refer to the 36-variable linear function as the "full discriminant model tool." Brief Screening Tool Development {#s4d} -------------------------------- To develop a shorter -- and therefore more easily used -- screening questionnaire capable of classifying members of independent samples into the six audience segments with 80% accuracy or better, we eliminated the 21 least predictive variables from the discriminant function. The resultant 15-item "brief" screening instrument, when applied to our dataset, correctly classifies 83.8% of the sample (ranging from 60 to 97% in the six segments; see [Table 7](#pone-0017571-t007){ref-type="table"}). Elsewhere in this paper we refer to the 15-variable linear function as the "reduced discriminant model tool." Validation of the Segments {#s4e} -------------------------- To validate that the segments account for variance in important measures above and beyond variance accounted for by other common explanatory measures, we conducted a series of linear regressions. The dependent measure for these analyses was an index of support for a series of nine federal policies that, if enacted, should reduce national levels of greenhouse gas emissions. Responses to each of these questions were combined into a simple index; the Cronbach\'s alpha for this policy support scale was 0.86. In the first analysis, the demographic variables of age, gender, income, education, marital status, work status, and race were regressed against the GHG reduction policy support measure. In the second analysis, a five-point political ideology scale (very liberal, somewhat liberal, moderate, somewhat conservative, very conservation) was added into the regression. In the final analysis, audience segment status was added into the regression. Human Subjects Approval and Informed Consent {#s4f} -------------------------------------------- This research was approved by the Human Subjects Review Board at George Mason University and Yale University. Written informed consent was obtained from all participants involved in this research. The authors wish to thank Matthew Nisbet for conceptual input, Guoqing Diao and Katya Seryakova for statistical advice, Russell Shaddox for graphic design support, Karen Akerlof for data and editorial assistance, and Colleen Redding, Norbert Mundorf, Saffron O\'Neill and an anonymous reviewer for reviewing and providing useful input on the manuscript. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was funded by the Yale Center for Environmental Law and Policy; the Betsy and Jessie Fink Foundation; the 11th Hour Project; the Pacific Foundation; and by a Robert Wood Johnson Foundation Investigator Award in Health Policy. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: EWM AL CRR. Performed the experiments: EWM AL CRR. Analyzed the data: EWM AL CRR CKM. Wrote the paper: EWM AL CRR CKM.

PubMed Central

2024-06-05T04:04:19.753467

2011-3-10

PMC3053363

Introduction {#s1} ============ HIV/AIDS emerged late in Thailand compared to other countries worldwide [@pone.0016902-Ruxrungtham1]. The first case was reported in 1984, although this was a returned emigrant who developed AIDS elsewhere [@pone.0016902-Thailand1]. A few more cases were reported in 1984--1988 between men who had sex with men (MSM) and injecting drug users (IDU) [@pone.0016902-Wirachsilp1]. In 1989 AIDS hit Thailand hard after HIV spread very quickly through the IDU community, and a year later entered the commercial sex worker (CSW) networks [@pone.0016902-Ruxrungtham1], [@pone.0016902-Weniger1]. In subsequent years, prevalence rates among these high-risk groups grew explosively from almost zero to 30 to 50% [@pone.0016902-Saengwonloey1]--[@pone.0016902-UNAIDS1]. Since then, the Thai HIV epidemic has been largely driven by CSW and IDU [@pone.0016902-Ruxrungtham1], whose epidemics appear to be linked [@pone.0016902-Tovanabutra1], [@pone.0016902-Tovanabutra2]. Over the last 12 years, for example, heterosexual (HT) and IDU transmissions accounted for 80--85% and 5% of the infections, respectively [@pone.0016902-Wirachsilp1]--[@pone.0016902-Saengwonloey1]; and although the former has now diminished considerably, the latter has remained high [@pone.0016902-UNAIDS1]. In 1991, AIDS prevention became a national priority in Thailand and between 1993 and 1997 the government increased the national budget, launched several campaigns to control and inform about AIDS spread (Ministry of Public Health, Thailand; eng.moph.go.th) and initiated the "100 percent condom program" [@pone.0016902-Hanenberg1]. All these policies slowed down the spread of AIDS and the national prevalence rate was reduced by 0.6% points [@pone.0016902-UNAIDS1]. These AIDS campaigns were mostly successful at reducing HIV infections in CSW, whose prevalence rate is now only 5%, but older HIV epidemics in IDU and MSM continue unabated (prevalence rates are \>25%; [@pone.0016902-UNAIDS1]), fueling epidemics in CSW [@pone.0016902-Saidel1] and causing new outbreaks [@pone.0016902-Ruxrungtham1]. Almost 1.5% adults are still infected with HIV in Thailand (∼610,000 infected individuals) making AIDS a leading cause of death (30,000 deaths in 2007; [@pone.0016902-UNAIDSWHO1], [@pone.0016902-UNAIDS1]). The Thai HIV epidemic has become now more heterogeneous [@pone.0016902-Over1] and it is increasingly affecting people traditionally considered to be at lower risk of infection [@pone.0016902-UNAIDS1]. Of even more concern, there are already signs that the epidemic could grow in coming years: prevalence rates among high-risk groups have increased, condom use has decreased, and risky sexual behavior is on the rise [@pone.0016902-UNAIDS1], [@pone.0016902-UNAIDS2]. In the early years of the AIDS epidemic in Thailand HIV-1, subtypes were segregated by risk group. Subtype B was predominant in IDU; while CRF01\_AE (a recombinant between subtypes A and E) was predominant in HT and MSM [@pone.0016902-Tovanabutra2], [@pone.0016902-Ou1]. As the Thai epidemic progressed, CRF01\_AE increased in frequency across all high-risk groups [@pone.0016902-Kalish1], [@pone.0016902-Wasi1] and by 1995 it became also the predominant subtype in IDU [@pone.0016902-Subbarao1]. Thus, between 1995 and 2004 CRF01\_AE accounted for 80--97% (depending on the study) of the new HIV-1 infections [@pone.0016902-Wirachsilp1], [@pone.0016902-Subbarao2]. But the Thai molecular epidemiology has been gradually growing in complexity and now it seems to be entering a new phase [@pone.0016902-Wirachsilp1], [@pone.0016902-Tovanabutra2]. New recombinant CRF15\_01B, CRF01\_AE/B and CRF01\_AE/C isolates are constantly being identified both within HT and IDU [@pone.0016902-Wirachsilp1], [@pone.0016902-Tovanabutra2], [@pone.0016902-Tovanabutra3]--[@pone.0016902-deSilva1] and some may be increasing their frequency rapidly (13% CRF01\_AE/B in IDU; [@pone.0016902-Tovanabutra2]). Under this new epidemic scenario, molecular surveillance becomes crucial to monitor emerging trends in HIV transmission, assess intervention strategies, and evaluate vaccine efficiency and design [@pone.0016902-Tovanabutra2], [@pone.0016902-deSilva1]--[@pone.0016902-Salomon1]. HIV spreads through often complex contact networks or transmission (infection) chains [@pone.0016902-Rothenberg1], [@pone.0016902-Kretzschmar1]. The characteristics of such networks play a crucial role in determining short- and long-term disease dynamics [@pone.0016902-Eames1]; hence, understanding those networks may translate into more efficient prevention measures and treatment interventions [@pone.0016902-Pybus1], [@pone.0016902-Lewis1]. Several phylogenetic studies suggest that transmission chains associated with acute (early) HIV-1 infection may greatly contribute to viral transmission and spread of the epidemic [@pone.0016902-Yerly1]. Data from both sexually- and drug-related acute infections in Europe [@pone.0016902-Yerly1]--[@pone.0016902-Thomson1], Canada [@pone.0016902-Brenner1], [@pone.0016902-Brenner2], and Panama [@pone.0016902-AhumadaRuiz1] have reported clustering in 24--65% of HIV-1 sequences. However, in a genetic analysis of 130 early diagnosed HIV-1 infections in IDU from Bangkok only 7.4% of the subtype B and 16.5% of the CRF01\_AE isolates formed transmission clusters [@pone.0016902-Nguyen1]. Similarly, in a recent study of sexually infected HIV-1 patients (mostly MSM) in North America, clustering was detected in only 11% of the isolates [@pone.0016902-PrezLosada1]. Therefore, it seems like the extent to which acute transmission of HIV-1 is clustered remains open. Thailand is one of the key international partners in the HIV vaccine efficiency trials with three trials already completed [@pone.0016902-Migasena1]--[@pone.0016902-Vanichseni1]. In 2003, the first phase III placebo-controlled trial (VAX003) of a candidate HIV-1 vaccine (AIDSVAX B/E) was completed in individuals at high risk for HIV-1 infection [@pone.0016902-Pitisuttithum2]--[@pone.0016902-vanGriensvan1]. The study enrolled 2,546 uninfected IDU from and around Bangkok of which we obtained clinical samples for 215 who became infected with HIV-1 between 1999 and 2003 despite intensive risk reduction counseling. Plasma samples from these individuals were obtained within the first 13 months after infection, and envelope glycoprotein (gp120) viral sequences were generated. These sequences (the "VAX003 dataset") have recently been released to the scientific community through the Global Solutions for Infectious Diseases HIV sequence Database ([www.GSID.org](http://www.GSID.org)). Here, we analyze the VAX003 data to assess HIV-1 variation as a function of treatment (vaccine or placebo), viral load, and CD4^+^ counts. Moreover, we perform a molecular surveillance of the VAX003 gp120 dataset to identify HIV-1 circulating subtypes in Bangkok and infer transmission networks in IDU. Finally, we combine the VAX003 dataset with other Thai sequences available in the HIV Los Alamos database ([www.hiv.lanl.gov](http://www.hiv.lanl.gov)) to investigate the timescale and molecular population dynamics of HIV-1 in Thailand. The VAX003 dataset is the largest collection of gp120 sequences from infections resulting from new and recent transmissions in Thailand and one of the few datasets collected from a large IDU cohort. These data provide a unique opportunity to study HIV-1 evolution in an epidemiological context and we anticipate it will contribute to the analysis and interpretation of the results from the RV144 Phase III HIV vaccine trial recently completed in Thailand [@pone.0016902-RerksNgarm1], [@pone.0016902-Vaccari1]. Results {#s2} ======= Molecular surveillance and subtype diversity {#s2a} -------------------------------------------- We indentified 182 CRF01\_AE (84.7%), 29 subtype B (13.4%), and 4 discordant isolates that presumptively are CRF15\_AE (1.9%). This latter recombinant type is mostly CRF01\_AE but also includes most of gp120 (except for approximately the first 36 nucleotides) and the external portion of gp41 from subtype B [@pone.0016902-Tovanabutra1]. Full genome sequencing of these discordant HIV isolates are needed to confirm this result. Number of isolates (as percentages) per year (1999 to 2003) was similar within each subtype ([Table S1](#pone.0016902.s002){ref-type="supplementary-material"}). Estimates of genetic diversity (θ) were similar (∼0.11) across subtypes ([Table 1](#pone-0016902-t001){ref-type="table"}). Selection estimates (ω) were generally below 1 although subtype B \[ω~PAML~ = 0.777; ω~omegaMap~ = 0.778 (0.673--0.901)\] showed higher (and significant for omegaMap) average *d* ~N~/*d* ~S~ rates than CRF01\_AE \[ω~PAML~ = 0.580; ω~omegaMap~ = 0.404 (0.366--0.443)\]. Population recombination rates (ρ~omegaMap~), however, were significantly lower for subtype B \[3.95 (3.45--4.53)\] than for CRF01\_AE \[15.56 (14.65--16.65)\]. DNA genetic divergence (±SD) was higher for subtype B (0.096±0.019) than for CRF01\_AE (0.067±0.011). θ, ω~PAML~ and genetic divergence were also estimated in the North American VAX004 gp120 dataset [@pone.0016902-Flynn1] for comparison between B subtypes ([Table 1](#pone-0016902-t001){ref-type="table"}). θ estimates were again similar between datasets, but ω~PAML~ estimates were lower for the VAX004 dataset (0.432) than for the VAX003 dataset (0.777), while genetic divergence was significantly higher for the VAX004 dataset (0.112±0.015). ::: {#pone-0016902-t001 .table-wrap} 10.1371/journal.pone.0016902.t001 Table 1 ::: {.caption} ###### Overall subtype diversity estimates. ::: {#pone-0016902-t001-1} HIV-1 θ ω~PAML~ ω~omegaMap~ ρ~omegaMap~ GD ------------------------ ------- --------- ---------------- ---------------- ---------------- VAX003-CRF01\_AE (182) 0.110 0.581 0.404 15.56 0.067 (0.366--0.443) (14.65--16.65) (0.067--0.067) VAX003-Subtype B (29) 0.112 0.777 0.778 3.95 0.096 (0.673--0.901) (3.45--4.53) (0.094--0.098) VAX004-Subtype B (345) 0.105 0.432 \- \- 0.112 (0.112--0.112) Genetic diversity (θ), selection in PAML (ω~PAML~) and omegaMap (ω~omegaMap~), population recombination rate in omegaMap (ρ~omegaMap~), and genetic divergence (GD). Number of isolates analyzed is indicated between parentheses in the first column. Estimates from the North American VAX004 subtype B trial were included for comparison. ::: Phylogenetic analysis {#s2b} --------------------- The GTR+Γ+I model [@pone.0016902-Tavar1] was chosen as the best-fit model for both the VAX003 gp120 dataset and for all their corresponding codon-position partitions. ML and Bayesian phylogenies did not show any obvious structure based on treatment, VL or CD4^+^ categories ([Fig. 1](#pone-0016902-g001){ref-type="fig"}). Individuals within each factor seemed to be randomly distributed across the phylogeny. ::: {#pone-0016902-g001 .fig} 10.1371/journal.pone.0016902.g001 Figure 1 ::: {.caption} ###### HIV-1 subtype phylogenetic trees. Maximum likelihood phylogenetic inference of Bangkok HIV-1 CRF01\_AE and subtype B population structuring as a function of treatment \[placebo (P) and vaccine (V)\], viral load (VL), and CD4^+^ counts. Branch lengths are shown proportional to the amount of change along the branches. Clades supported by bootstrap proportions ≥70% and posterior probabilities ≥0.95 in the Bayesian analysis (transmission chains) are shown in red color and their terminals in bold. Only one clone per isolate (numbered) is represented for simplicity. :::  ::: Transmission clusters {#s2c} --------------------- ML and Bayesian phylogenetic analyses of HIV-1 subtype B and CRF01\_AE showed 3 and 31 well-supported clades (bootstrap proportions ≥70% and posterior probability ≥0.95), respectively ([Fig. 1](#pone-0016902-g001){ref-type="fig"}). These transmission networks involved 10 (34.4%) subtype B and 91 (50%) CRF01\_AE IUD isolates distributed in 2 small (\<5 isolates)/1 large (≥5 isolates) and 26/5 clusters, respectively ([Table S2](#pone.0016902.s003){ref-type="supplementary-material"}). Attendance to a particular clinic and estimated date of seroconversion (considered as a time window of ≤6 months) were found to be associated with 1 and 1 subtype B clusters, respectively, and 6 and 13 CRF01\_AE clusters, respectively. Some overlap between factors was observed in some clusters (e.g., clade 2 in subtype B and clade 7 in CRF01\_AE). Nonetheless, these results suggest that between 1999 and 2003, the estimated date of infection seemed to play a larger role than geographic location at establishing transmission chains in CRF01\_AE IDU from Bangkok ([Table S2](#pone.0016902.s003){ref-type="supplementary-material"}). Viral evolution and patient factors {#s2d} ----------------------------------- Average θ, ρ, and ω intra-patient estimates within categories were very low for both subtypes ([Table 2](#pone-0016902-t002){ref-type="table"}). For most CRF01\_AE datasets ω\>1, while for the rest ω≈1. On the contrary, for most subtype B datasets ω\<1, but ω\>1 was also found in several cases. These intersubtype differences, nonetheless, were non-significant. CRF01\_AE sequences from individuals with lower VL and higher CD4^+^ counts showed lower θ values (0.005--0.006) than individuals with higher VL (*P* = 0.016) and lower CD4^+^ (*P* = 0.007) counts (0.007--0.009). These two factors were inversely correlated (Pearson correlation coefficient = −0.218, *P* = 0.003). ::: {#pone-0016902-t002 .table-wrap} 10.1371/journal.pone.0016902.t002 Table 2 ::: {.caption} ###### Mean patient diversity estimates. ::: {#pone-0016902-t002-2} HIV-1 θ ρ~LDhat~ ω~PAML~ ω~HYPHY~ ------------------------------- ------- ---------- --------- ---------- Subtype CRF01\_AE Treatment  Placebo (92) 0.007 4.8 1.045 1.208  Vaccine (87) 0.007 3.0 1.057 1.229 VL Categories (RNA copies/mL)  1: \<1×10^4^ (29) 0.005 6.1 0.999 1.150  2: 1×10^4^--5×10^4^ (48) 0.006 3.7 1.030 1.198  3: 5×10^4^--10×10^4^ (40) 0.007 2.8 0.958 1.110  4: 10×10^4^--25×10^4^ (45) 0.008 4.4 1.265 1.451  5: \>25×10^4^ (21) 0.009 2.5 0.905 1.082 CD4^+^ counts (cells/mm^3^)  1: \<3×10^2^ (31) 0.009 2.4 0.921 1.082  2: 3×10^2^--5×10^2^ (79) 0.007 3.8 1.100 1.250  3: 5×10^2^--7×10^2^ (39) 0.006 3.7 1.003 1.187  4: \>7×10^2^ (34) 0.006 5.9 1.119 1.317 Subtype B Treatment  Placebo (17) 0.008 5.5 0.775 0.869  Vaccine (14) 0.008 4.4 0.978 0.997 VL Categories (virions/mL)  1: \<1×10^4^ (8) 0.004 3.8 0.653 0.707  2: 1×10^4^--5×10^4^ (9) 0.009 2.3 0.873 0.887  3: 5×10^4^--10×10^4^ (7) 0.011 11.4 0.998 1.199  4: \>10×10^4^ (7) 0.007 1.9 1.100 1.047 CD4^+^ counts (cells/mL)  1: \<3×10^2^ (31) 0.011 12.4 0.870 0.899  2: 3×10^2^--5×10^2^ (79) 0.006 5.2 0.847 0.950  3: 5×10^2^--7×10^2^ (39) 0.010 2.0 0.762 0.818  4: \>7×10^2^ (34) 0.006 4.3 1.121 1.078 Genetic diversity (θ), population recombination rate in LDhat (ρ~LDhat~), and selection estimates in PAML (ω~PAML~) and HYPHY (ω~HYPHY~). Number of isolates analyzed is indicated between parentheses. ::: Population dynamics {#s2e} ------------------- BEAST\'s estimate of the substitution rate was 0.0055 (0.0050--0.0060) for CRF01\_AE and 0.0027 (0.0015--0.0038) for subtype B. The Most Recent Common Ancestor (MRCA) was dated in 1984.5 (1983--1986) for CRF01\_AE and in 1965 (1950--1979) for subtype B. The BSP analysis of CRF01\_AE sequences ([Fig. 2](#pone-0016902-g002){ref-type="fig"}) suggested that the relative genetic diversity increased exponentially between 1984 and 1991, moderately between 1992 and 1995, decreased between 1996 and 2004 with a spike in 1999--2000, and then increased slightly between 2005 and 2006 (the age of our most recent sample). ::: {#pone-0016902-g002 .fig} 10.1371/journal.pone.0016902.g002 Figure 2 ::: {.caption} ###### HIV-1 CRF01\_AE past population dynamics. Bayesian skyline plot of the HIV-1 CRF01\_AE subtype in Thailand. Solid black lines show the median estimate and dashed black lines the 95% high posterior density limits. The estimated incidence and prevalence rate are indicated in red and blue, respectively (see text for details). :::  ::: Discussion {#s3} ========== Molecular surveillance and subtype diversity {#s3a} -------------------------------------------- The predominant HIV-1 subtype circulating in IDU (215 patients) from Bangkok during 1999--2003 was CRF01\_AE (85%). Subtype B accounted for 13% of the infections and CRF15\_AE for 2%. Two early (1995 to 1998) molecular surveys in Bangkok [@pone.0016902-Subbarao1], [@pone.0016902-Nguyen1] including 102 and 130 IDU, respectively, and using the C2-V4 *env* region (345 bp), reported also high percentages of subtype B isolates (20--21%) but no CRF15\_AE recombinants. Additional surveillances among IDU in Bangkok [@pone.0016902-Ramos1] during 1997--1998 (111 patients) using *env* (530 bp) and protease (297 bp) still detected high percentages of subtype B isolates (23%), but also 3.6% CRF15\_AE isolates. Interestingly in Northern Thailand, a near full HIV-1 genome study (1999--2002; 38 patients) among IDU detected an increasing proportion of CRF15\_AE (13% total) infections but no pure subtype B isolates, suggesting that the latter subtype became extinct in this region [@pone.0016902-Tovanabutra2]. Although full HIV-1 genome surveys increase the probability of finding intersubtype recombinants, our surveillance from 1999 to 2003 suggests that subtype B is declining and CRF15\_AE is increasing among IDU from Bangkok as previously predicted by others [@pone.0016902-Xiridou1] and observed in other high-risk groups across the country [@pone.0016902-Wirachsilp1], [@pone.0016902-deSilva1], [@pone.0016902-Watanaveeradej1]. Nonetheless, considering subtype B prevalence rate and genetic diversity ([Table 1](#pone-0016902-t001){ref-type="table"}), it may remain circulating in Thailand for many years. This information is important to ensure that the virus diversity upon which vaccines are designed matches the circulating viral population. Fortunately, vaccine candidates used in the RV144 Phase III HIV vaccine trial largely contain both subtype B and CRF01\_AE viruses [@pone.0016902-RerksNgarm1], [@pone.0016902-Vaccari1]. Our estimate of gp120 genetic divergence in CRF01\_AE viruses (0.067±0.011) from IDU was higher than previously reported for *env* (0.059±0.011) in Northern Thai IDU patients [@pone.0016902-Tovanabutra2], but much lower than those reported for the C2-V4 *env* region (mean: 0.109--0.150) in other Thai regions among mostly (95%) sexually infected individuals [@pone.0016902-Wirachsilp1]. These comparisons must be considered with caution since the *env* regions compared are not exactly the same and we removed many of the variable sites after the GBlocks analysis. Genetic divergence estimates using the full VAX003 gp120 alignment were of 0.100±0.015. But independently of what dataset we consider, both Tovanabutra et al. [@pone.0016902-Tovanabutra2] and our own estimates are higher than those reported in other Asian IDU groups in, for example, China [@pone.0016902-Piyasirisilp1] or Vietnam [@pone.0016902-Liao1], [@pone.0016902-Kato1], and closer to those observed in sexual transmission cohorts [@pone.0016902-Wirachsilp1], where diversity is generally higher [@pone.0016902-McCutchan1]. This result then suggests that the IDU epidemic in Thailand is likely to be mature and that extensive exchange between sexual and IDU exposures and transmissions has been ongoing for years [@pone.0016902-Tovanabutra2], which is also supported by our phylogenetic results below. gp120 subtype B sequences from Bangkok are significantly less divergent than those from the North American VAX004 vaccine trial ([Table 1](#pone-0016902-t001){ref-type="table"}). This and previous gp120 CRF01\_AE estimates indicate that Thai HIV-1 populations are more homogeneous than those observed in other areas like Vietnam (see below) or North America. The increased homogeneity of viruses in Bangkok has been attributed to the relatively recent introduction of HIV in Thailand (1984) and a pronounced founder effect resulting from the rapid spread of the virus [@pone.0016902-Ruxrungtham1], [@pone.0016902-Weniger1]. This result then suggests a greater opportunity to overcome the challenge of HIV diversity [@pone.0016902-Taylor1] and to detect protective immunity induced by candidate vaccines in Thailand compared to North America or Africa, where viral genetic diversity is much higher. Indeed, the outcome of the RV144 vaccine testing in Thailand seems to have had greater success by better coverage of this limited diversity with vaccinated volunteers showing 31.2% fewer infections than placebo recipients [@pone.0016902-RerksNgarm1], [@pone.0016902-Vaccari1]. American subtype B viruses also appear to be under stronger purifying selection (ω = 0.43--0.58) than the Thai subtype B viruses (0.78). This suggests that differences could exist in the intrinsic immune response among ethnicities [@pone.0016902-PrezLosada2] or transmission type (i.e., IDU vs MSM) [@pone.0016902-Choisy1]. Our genetic estimators indicate that CRF01\_AE experienced almost four times more recombination than subtype B ([Table 1](#pone-0016902-t001){ref-type="table"}). Consequently, one could also expect that higher recombination rates would inflate ω~PAML~ estimates [@pone.0016902-Anisimova1]--[@pone.0016902-Wilson1], but that does not seem to be the case, since subtype B showed significantly higher levels of selection than CRF01\_AE for all estimators. Similarly, CRF01\_AE presented a mean substitution rate per site twice as high as that observed for subtype B. Significant differences in adaptive selection and substitution rate between HIV-1 subtypes have been reported before [@pone.0016902-Choisy1], [@pone.0016902-Abecasis1] and were attributed to differences in immune selective pressure from the host and in mutation rate or generation time of the virus. Phylogenetic structure of HIV-1 in Thailand {#s3b} ------------------------------------------- Our CRF01\_AE and subtype B phylogenetic trees suggest that HIV populations in IDU from Bangkok are not structured by any of the epidemiological and clinical factors studied ([Fig. 1](#pone-0016902-g001){ref-type="fig"}). Moreover, our BEAST analyses of both VAX003-LA gp120 subtypes did not show phylogenetic structuring based on transmission type either ([Fig. S1](#pone.0016902.s001){ref-type="supplementary-material"}). These results agree with previous CRF01\_AE star-like phylogenies reported in IDU from Bangkok [@pone.0016902-Nguyen1]. Geographically broader CRF01\_AE phylogenetic studies in Central [@pone.0016902-Utachee1], Northern [@pone.0016902-Tovanabutra2] and across Thailand [@pone.0016902-Wirachsilp1], [@pone.0016902-deSilva1] also showed lack of structuring based on transmission type, sociodemographic factors and geographic location. In the Wirachsilp et al. [@pone.0016902-Wirachsilp1] study, for example, sequences from Bangkok clustered together with sequences from other regions. Similarly, Keele et al. [@pone.0016902-Keele1] also showed that viral *env* genes evolving from individual transmitted or founder HIV-1 subtype B viruses generally exhibited a star-like phylogeny, such as the one observed in North American viruses [@pone.0016902-PrezLosada1]. Given the age of the HIV-1 epidemic in Thailand and the fact that the virus is thought to mutate at a rate of 1% per year [@pone.0016902-Shankarappa1], [@pone.0016902-Korber1], the possibility existed that different clades would have emerged in different regions or high-risk groups in Thailand. Indeed phylogenetic structuring based on these factors has been observed before between subtypes in, for example, Africa [@pone.0016902-Papathanasopoulos1] and Asia [@pone.0016902-Oelrichs1] and within subtypes in, for example, Vietnam [@pone.0016902-Liao1] and China [@pone.0016902-Cheng1]. But contrary to what happened in those HIV/AIDS epidemics, the Thai epidemic spread exponentially across the whole country and risk types [@pone.0016902-Ruxrungtham1], which could erase early genetic differentiation and results in star-like gene genealogies [@pone.0016902-Marjoram1], [@pone.0016902-Rosenberg1]. Moreover, both molecular (this study and [@pone.0016902-Tovanabutra1], [@pone.0016902-Tovanabutra2]) and Thai behavioral [@pone.0016902-Saidel1], [@pone.0016902-Family1] data indicate that bridging between drug and sexual epidemics through CSW has been ongoing for years, which again reduces the opportunity for differentiation. Phylogenetic clusters in acute transmissions {#s3c} -------------------------------------------- The extent to which acute transmission of HIV-1 is clustered is not clear. Some studies [@pone.0016902-Yerly1]--[@pone.0016902-Brenner2], [@pone.0016902-AhumadaRuiz2] report high clustering (24 to 65%) levels, while others [@pone.0016902-Nguyen1], [@pone.0016902-PrezLosada1] show much lower values (7 to 17%) for the same subtypes and transmission routes. Our more comprehensive phylogenetic analyses of IDU from Bangkok show higher proportions of early subtype B (35%) and CRF01\_AE (50%) infections falling into clusters, confirming that transmission chains associated with acute infection play a key role in HIV-1 transmission and spread [@pone.0016902-Yerly1]. Transmission clusters in Nguyen et al. [@pone.0016902-Nguyen1] were inferred using the C2-V4 *env* region (345 bp). This gene region, although broadly used in HIV genetic studies, is less informative than the gp120 (∼1.5 kb) region used here for estimating phylogenetic clustering. As for Pérez-Losada et al. [@pone.0016902-PrezLosada1], that study covered North America, while our Thai study and others before, focus on a single city, a small country or a recently infected area. This suggests that the size and population structure of the studied area affect our ability to identify HIV-1 transmission chains. Moreover, differences in clustering have been also observed between subtypes, transmission routes and regions [@pone.0016902-Cuevas1], [@pone.0016902-Chalmet1], [@pone.0016902-PrezLosada1]. Hence future HIV vaccine trials should pay attention to potential sources of clustering that can effectively render samples non-independent. Viral evolution and patient factors {#s3d} ----------------------------------- No significant differences in recombination, mutation, and selection rates were observed among vaccinated and placebo individuals in both subtype B and CRF01\_AE. This is consistent with the overall outcome of the VAX003 trial where immunization with AIDSVAX B/E did not significantly affect the rate of infection, the VL, the CD4^+^ count, or the clinical outcome of vaccine recipients compared to placebo recipients [@pone.0016902-Pitisuttithum2]. Lower VL and higher CD4^+^ counts, however, were significantly associated with lower mutation rates in CRF01\_AE ([Table 2](#pone-0016902-t002){ref-type="table"}). Since genetic diversity may be positively correlated with N~e~, one could expect that greater VL (census size) would also cause an increase on the number of effective virions [@pone.0016902-PrezLosada1]. Population dynamics of Thai HIV-1 subtypes {#s3e} ------------------------------------------ Previous full-genome phylogenetic analyses of HIV-1 CRF01\_AE in Southeast Asia [@pone.0016902-Liao1] indicate that CRF01\_AE was introduced from Africa to Thailand and then spread elsewhere. Our coalescent estimate of the time of emergence of CRF01\_AE in Thailand was 1984.5 (1983--1986). An slightly earlier estimate (1981±2 years) was previously reported by Liao et al. [@pone.0016902-Liao1] using the same method but including 64 near full-length CRF01\_AE nucleotide sequences from Africa, China, and Vietnam. Hence, both studies suggest that CRF01\_AE was circulating cryptically in Thailand for 3--10 years before it was first detected in 1989 [@pone.0016902-McCutchan2]. Similar time lags between evolutionary estimates and the recognition of symptomatic patients have been observed before in other countries such as United States [@pone.0016902-Gilbert1] and Vietnam [@pone.0016902-Liao1]. In Western countries, the estimated median incubation period before AIDS development in the absence of antiretroviral therapy is 10--12 years [@pone.0016902-NIAID1], although in Thailand a shorter incubation period (7 years) has been suggested [@pone.0016902-Rangsin1]. HIV testing in Thailand started in 1985 and only 3 cases were detected [@pone.0016902-Phanuphak1]. There were no cases reported in 1986, but many thousands were detected over the next 3 years, particularly among IDU from Bangkok [@pone.0016902-McCutchan2]. By February 1990, almost 15,000 cases of HIV-1 infection have been already documented across the country [@pone.0016902-Smith1]. Similarly to what happened in other regions, CRF01\_AE could have been introduced in Thailand years before its detection in 1989. Phylogenetic analyses of HIV-1 subtype B *env* data collected worldwide [@pone.0016902-Gilbert1] indicate that this subtype was introduced from Africa to Haiti and then spread elsewhere (pandemic clade). In that study, the Thai subtype B isolates did not seem to form a separate cluster (independent HIV-1 expansion), hence our coalescent estimate of the time of emergence of subtype B (1965±15) approximates that of the emergence of the subtype worldwide (1968--1969±3 years) [@pone.0016902-PrezLosada1], [@pone.0016902-Gilbert1]. Discrepancies between these and our current estimate are probably due to differences in sample size: the subtype B dataset analyzed here is geographically more restricted and includes fewer sampling points. A short interval of sampling years, for example, provides less information about the average rate during that interval than does a long interval [@pone.0016902-Abecasis1], [@pone.0016902-Seo1]. The larger HPD intervals of the Thai estimate supports that idea. Our BEAST analysis of CRF01\_AE past dynamics ([Fig. 2](#pone-0016902-g002){ref-type="fig"}) agrees well with the history of HIV/AIDS spread in Thailand and the prevalence and incidence rates reported [@pone.0016902-UNAIDSWHO1] and predicted using backcalculation models [@pone.0016902-Punyacharoensin1]. Prior to 1987 the prevalence of HIV in Thailand was low, but once HIV entered the MSM, IDU and CSW networks (1988--1993) prevalence rates exploded rising from virtually zero to up to 50% [@pone.0016902-Weniger1]--[@pone.0016902-UNAIDS1] and so did the relative genetic diversity (N~e~τ). In 1991, AIDS prevention became a national priority at the highest level and several campaigns were launched to control AIDS spread (Ministry of Public Health, Thailand; <http://eng.moph.go.th/>). Consequently, prevalence rates began to decline soon after ([Fig. 2](#pone-0016902-g002){ref-type="fig"}; see also [@pone.0016902-Hanenberg1]) and HIV incidence was reduced by a third [@pone.0016902-Punyacharoensin1]. Concomitantly, N~e~τ leveled off and slowly began to decline in 1996. In 1998, due to the Asian Financial Crisis, HIV/AIDS funding was severely reduced [@pone.0016902-Saengdidtha1] and many programs like the HIV prevention schemes were downscaled or suspended [@pone.0016902-UNAIDS1], [@pone.0016902-Saengdidtha1], [@pone.0016902-Marais1]. This led to a decline in awareness and possibly an increase in unsafe sexual behavior [@pone.0016902-UNAIDS1]. Consequently, the incidence rate spiked for two years and so did N~e~τ. In 2002 Thailand launched the third National Plan for the Prevention and Alleviation of HIV/AIDS (Ministry of Public Health, Thailand; eng.moph.go.th). Consequently, both incidence and N~e~τ decreased again until 2004, but since then the former has remained constant and the latter has increased slightly and remained relatively high. Under circumstances of low surveillance and high HIV diversity, new or existing infective strains could expand exponentially and provoke a resurgence of AIDS across the country. There are already signs that the epidemic could grow in coming years [@pone.0016902-UNAIDS1], [@pone.0016902-UNAIDS2]. More importantly, the epidemic has never eased off among certain groups like IDU, where infection rates are still very high (∼30%; [@pone.0016902-UNAIDS1]) and continue to be a reservoir for HIV fueling old and causing new epidemics [@pone.0016902-Ruxrungtham1], [@pone.0016902-Saidel1]. Thailand must then increase prevention efforts, especially among high-risk groups such as IDU and MSM, but also among the general population since the AIDS epidemic seems to be more heterogeneous now [@pone.0016902-Over1]. In light of these concerns, the current government has increased HIV/AIDS prevention efforts. In 2007, a three-year strategic plan was announced that would focus on those most at risk of HIV infection and difficult-to-reach groups [@pone.0016902-USAID1]. How these new policies are going to affect HIV-1 diversity and dynamics is for further studies to see. Materials and Methods {#s4} ===================== VAX003 vaccine trial participants {#s4a} --------------------------------- The 2,546 volunteers participating in the VAX003 trial (NCT00006327) were recruited from 17 clinics in and around Bangkok [@pone.0016902-Pitisuttithum2]--[@pone.0016902-vanGriensvan1]. They all were considered at high risk for HIV-1 infection through injection drug use. The vaccine trial protocol did not specify racial categories and no effort was made to distinguish linguistic and geographic groups. Volunteers were randomly assigned to vaccine or placebo groups according to a 1∶1 ratio. All subjects were immunized with AIDSVAX B/E, a bivalent vaccine prepared by combining purified recombinant gp120s from two different strains of the HIV-1 virus incorporated in alum (aluminum hydroxide) adjuvant: the subtype B strain (MN) and the subtype CRF01\_AE isolate (A244). All subjects were immunized according to a 0, 1, 6, 12, 18, 24, and 36-month schedule. Serum samples were collected immediately prior to each injection and two weeks after each injection, with a final blood sample taken at 6 months following the final injection. The specimen taken prior to each injection was used to calculate pre-boost anti-gp120 titer values and submitted for HIV testing (ELISA). The immunoassays selected for HIV diagnosis were unaffected by antibodies to the AIDSVAX B/E antigens. If evidence of HIV infection was obtained, confirmatory testing was carried out by immunoblot. Once HIV-1 infections were confirmed, HIV-1+ subjects were enrolled in a separate protocol (Step B) where plasma and cells were collected at regular intervals for up to two years post infection. Plasma samples were used for measurement of viral loads and envelope glycoprotein sequencing. Frozen lymphocytes were cryopreserved for immunologic and genetic testing. The date of infection was defined as the midpoint between the last seronegative specimen and the first seropositive specimen. The estimated time of infection ranged from 0 to 13 months with a mean time of infection of 3--4 months. Viral load (VL) and CD4^+^ measurements were taken and patients were subdivided into 4 or 5 categories for genetic analyses (see [Table 2](#pone-0016902-t002){ref-type="table"}). Molecular datasets {#s4b} ------------------ Of the 2,546 volunteers enrolled in the trial 230 became infected with HIV-1 [@pone.0016902-Pitisuttithum2] and we obtained clinical samples for 215 of them. Three to six clones per individual were collected from the same earliest post-infection plasma sample and sequenced for the viral gp120 gene (665 sequences total). A listing of the sequence data used for this analysis has recently been released online and can be accessed at [www.gsid.org](http://www.gsid.org). All gp120 sequences were determined using an ABI 3100 sequencer and assembled using Sequencher ([www.genecodes.com](http://www.genecodes.com)). HIV-1 subtype was determined using the REGA HIV Subtyping Tool 2.0 [@pone.0016902-deOliveira1] and the Recombinant Identification Program: RIP 3.0 [@pone.0016902-Siepel1] at Los Alamos (<http://hiv-web.lanl.gov/content/index>). Discordant (intersubtype recombinants) isolates were visually inspected and confirmed in RDP 3.0 [@pone.0016902-Martin1], [@pone.0016902-Martin2]. Two main HIV-1 subtypes were identified, CRF01\_AE (182 isolates) and subtype B (29 isolates) (see the Molecular Surveillance and Subtype Diversity section). Because of their genetic and epidemiological differences [@pone.0016902-Crandall1], these subtypes were analyzed separately. For population dynamic analyses full VAX003 gp120 sequences (only one clone per patient) were combined with other full length, dated Thai gp120 sequences from the Los Alamos database as of January 2010 to generate final datasets of 343 CRF01\_AE (from 1990 to 2006) and 47 subtype B (from 1990 to 2003) isolates. These combined datasets included 217/34 IDU, 36/4 HT, 3/0 CSW and 87/9 unknown risk-group CRF01\_AE/subtype B isolates. Sequence alignment {#s4c} ------------------ Nucleotide sequences were translated into amino acids and aligned in MAFFT 5.7 [@pone.0016902-Katoh1] using the global algorithm (G-INS-i). Ambiguous regions in the resulting alignment were identified and removed using GBlocks 0.91b [@pone.0016902-Castresana1]. Conserved amino acid regions were translated back to nucleotides generating alignments of 1,317--1,329 sites for CRF01\_AE and 1,398--1,413 sites for subtype B. Full gp120 sequences (1,497--1,629 bp) were analyzed for each patient, in which case the alignments were trivial. Phylogenetic analysis {#s4d} --------------------- The best-fit model of DNA substitution was selected with the Akaike Information Criterion [@pone.0016902-Akaike1] as implemented in jModelTest 1.0 [@pone.0016902-Posada1]. Maximum likelihood phylogenetic trees were inferred in RAxML 7.0.3 using 3 codon-position partitions [@pone.0016902-Stamatakis1]. Nodal support was assessed using the bootstrap procedure [@pone.0016902-Felsenstein1] with 1,000 replicates. Heuristic searches were performed under the best-fit model. In addition, Bayesian trees were inferred in MrBayes 3.1.1 [@pone.0016902-Ronquist1] using also 3 codon-position partitions. We ran four chains (one cold and three heated) for 2×10^7^ generations, sampling every 1,000 steps. Each run was repeated twice. Convergence and mixing of the Markov chains were assessed in Tracer 1.5 [@pone.0016902-Rambaut1]. Phylogenetic transmission (infection) clusters [@pone.0016902-Pybus1] were defined as those clades with bootstrap proportions ≥70% and posterior probabilities ≥0.95. Attendance at a particular clinic (which served as proxy for location of residence) and estimated date of seroconversion were screened for all the isolates contributing to clusters. Genetic divergence estimated as the mean pairwise genetic distances under the K2P model [@pone.0016902-Kimura1] was also calculated for comparison with previously published estimates. Genetic estimates and patient factors {#s4e} ------------------------------------- The VAX003 trial included vaccinated and non-vaccinated individuals with different VL and CD4^+^ counts. These individuals do not constitute natural populations, therefore, all genetic estimators described in this section were either applied to intra-patient datasets (3 to 6 clones) or full-subtype datasets (29 subtype B and 182 CRF01\_AE isolates). Genetic diversity (θ) and population recombination rate (ρ) was estimated for each patient using LDhat 2.1 [@pone.0016902-McVean1]. Here, each analysis was repeated 10 times and the ρ mean estimate was used for subsequent analyses. Molecular adaptation was assessed using the ratio of nonsynonymous (*d* ~N~) to synonymous (*d* ~S~) substitution rates (ω) and estimated using the model M0 (one-ratio) in PAML 3.14 [@pone.0016902-Yang1] and Fixed Effects Likelihood (FEL) in HYPHY 1.0 [@pone.0016902-KosakovskyPond1]. In the latter case, recombination was taken into account by first detecting recombination breakpoints with GARD [@pone.0016902-KosakovskyPond2] and then estimating the *d* ~N~/*d* ~S~ ratios independently for each fragment. Simultaneous estimation of ω and ρ was also performed in omegaMap [@pone.0016902-Wilson1] for the full-subtype datasets. Average estimates of ρ, θ, and ω were compared across factors (e.g., vaccinated and placebo; see [Table 2](#pone-0016902-t002){ref-type="table"}) using the Kruskal-Wallis test in Aabel 3 ([www.gigawiz.com](http://www.gigawiz.com)). Tests based on linear models (e.g., ANOVA) were not applied because their underlying assumptions were not met by some of the data sets. Population dynamics {#s4f} ------------------- Past population dynamics of CRF01\_AE in Thailand was inferred in BEAST 1.5.3 [@pone.0016902-Drummond1] using the Bayesian Skyline Plot (BSP) model [@pone.0016902-Drummond2] and a relaxed clock (lognormal) model of rate of substitution [@pone.0016902-Drummond3]. BSP searches showed overdispersed 95% High Posterior Density (HPD) intervals for subtype B, hence the exponential growth model was used instead. Relative genetic diversity through time (N~e~τ) was estimated directly from dated isolates under the best-fit model of nucleotide substitution. The hyperparameter *m* (number of grouped intervals) was set up 1/4 of the sequences in each case. Two runs 10^8^ and 2×10^7^ generations long were completed for each CRF01\_AE and subtype B, respectively. All output generated by BEAST was analyzed in Tracer 1.5 to test for convergence and mixing and implement the BSP model. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### **HIV-1 BEAST Bayesian trees.** BEAST maximum clade credibility trees of Thai HIV-1 CRF01\_AE (large tree) and subtype B (small tree) isolates. Injecting drug users (red), heterosexuals (blue), commercial sex workers (green), and unknown risk group (black) infections are indicated. Branch lengths are shown proportional to the amount of change along the branches. Only one clone per patient is represented for simplicity. (EPS) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S1 ::: {.caption} ###### **Number of isolates/percentage per year and subtype.** (DOCX) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S2 ::: {.caption} ###### **Phylogenetic transmission clusters.** Estimated date of infection and clinical site for subtype B and CRF01\_AE. (DOCX) ::: ::: {.caption} ###### Click here for additional data file. ::: **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**The data collection, clinical and genetic, for this study was funded by VaxGen as part of their Vax003 Phase III vaccine trial. The analysis of the DNA sequence data presented in this paper was supported by funding from the Bill and Melinda Gates Foundation to the non-profit organization Global Solutions for Infectious Diseases (GSID). The funders had no role in study design, data analysis, decision to publish, or preparation of this manuscript. [^1]: Conceived and designed the experiments: DVJ FS PWB. Performed the experiments: DVJ FS PWB. Analyzed the data: MPL DP MA KAC. Contributed reagents/materials/analysis tools: DVJ FS PWB MPL DP MA KAC. Wrote the paper: MPL KAC DP MA DVJ FS PWB.

PubMed Central

2024-06-05T04:04:19.758895

2011-3-10

PMC3053364

Introduction {#s1} ============ The chytrid fungus, *Batrachochytrium dendrobatidis* (Bd) [@pone.0016708-Longcore1], has been devastating amphibian populations globally [@pone.0016708-Daszak1], [@pone.0016708-Garner1], [@pone.0016708-DiRosa1], [@pone.0016708-Skerratt1], [@pone.0016708-Jones1], [@pone.0016708-Murray1], [@pone.0016708-Wake1], [@pone.0016708-Kilpatrick1], but not all species [@pone.0016708-Blaustein1], [@pone.0016708-Tennessen1] or animals in all regions [@pone.0016708-Kriger1], [@pone.0016708-Zellmer1], [@pone.0016708-Goldberg1], [@pone.0016708-Hossak1] appear equally susceptible. Two scenarios for the occurrence and spread of chytridiomycosis, likely reflecting different phases of the disease course, have been proposed [@pone.0016708-Briggs1], [@pone.0016708-Rachowicz1]. The first is that Bd is an epidemic, spreading as a wave and wiping out individuals, populations, and species in its path. This has occurred, or is occurring in Central America, in eastern Australia, and in parts of California [@pone.0016708-Berger1], [@pone.0016708-Lips1], [@pone.0016708-Lips2], [@pone.0016708-Lips3], [@pone.0016708-Lips4], [@pone.0016708-James1], [@pone.0016708-Vredenburg1]. The second scenario suggests that in certain regions of the world such as North America, much of the spread of Bd occurred decades ago and that in these places it is now endemic [@pone.0016708-Vredenburg1], [@pone.0016708-Ouellet1]. This situation may currently be the most relevant. Bd is now widespread throughout many geographic regions and is known to occur on every continent except Antarctica (where there are no amphibians); therefore, this disease may be considered global [@pone.0016708-Waldman1], [@pone.0016708-Retallick1], [@pone.0016708-Carnival1], [@pone.0016708-Adams1], [@pone.0016708-Adams2], [@pone.0016708-Longcore2], [@pone.0016708-Pearl1], [@pone.0016708-FrasAlvarez1], [@pone.0016708-Lampo1], [@pone.0016708-Rothermel1], [@pone.0016708-Scalera1], [@pone.0016708-Chatfield1], [@pone.0016708-Deguise1], [@pone.0016708-Briggs2]. A third scenario, the Bd thermal optimum hypothesis, in a sense combines the first two and has been more controversial. This hypothesis suggests widespread benign Bd distribution has been triggered to lethality by increased temperatures due to global warming [@pone.0016708-Pounds1], but there has been resistance to this idea [@pone.0016708-Lips5]. Amphibians are the only known host for Bd [@pone.0016708-Longcore1], [@pone.0016708-Vredenburg1], [@pone.0016708-Pessier1]. The life history of this fungus is composed of two stages: a free-living zoospore, which is flagellated and motile in aquatic environments, and a thallus (body), which is present in amphibian skin. Thallia form zoosporangia (vesicles), which in turn produce zoospores through asexual, and perhaps sexual, reproduction [@pone.0016708-Morgan1]. Zoospores can swim about 2 cm [@pone.0016708-Piotrowski1] and infect keratinizing squamous epithelial cells [@pone.0016708-Pessier2]. Favorable environments, where the infection can spread, are cool and wet. Hot and dry environments are considered hostile, and temperatures \>25°C may assist infected amphibians in clearing the infection [@pone.0016708-Piotrowski1], [@pone.0016708-Woodhams1], [@pone.0016708-RichardsZawacki1]. Resistance to Bd could include one of three mechanisms, which may work singly or in combination: antimicrobial properties of skin glandular secretions [@pone.0016708-Tennessen1], [@pone.0016708-RollinsSmith1], [@pone.0016708-RollinsSmith2], [@pone.0016708-RollinsSmith3], [@pone.0016708-Woodhams2]; antimicrobial properties of skin microflora [@pone.0016708-Harris1], [@pone.0016708-Harris2], [@pone.0016708-Harris3], [@pone.0016708-Lauer1], [@pone.0016708-Woodhams3], [@pone.0016708-Woodhams4]; and/or immune system function [@pone.0016708-RollinsSmith4], but this idea has been controversial [@pone.0016708-Ribas1]. Several factors complicate our attempts to understand this disease: different strains of Bd are known [@pone.0016708-Retallick2], [@pone.0016708-Rosenblum1], individuals in populations can gain and lose the infection seasonally [@pone.0016708-Briggs2], [@pone.0016708-Berger2], [@pone.0016708-Kriger2], [@pone.0016708-Gaertner1], and Bd-positive animals can show clinical signs of the disease (chytridiomycosis) or be completely asymptomatic [@pone.0016708-Pessier1], [@pone.0016708-Retallick2], [@pone.0016708-Davidson1], [@pone.0016708-Daszak2], [@pone.0016708-Hanselmann1], [@pone.0016708-Garner2], [@pone.0016708-Peterson1], [@pone.0016708-Woodhams5]. Bd infection is reported to be exacerbated by amphibian density [@pone.0016708-Briggs2], [@pone.0016708-Rachowicz2], tadpole longevity [@pone.0016708-Briggs2], Bd density (load dynamics) [@pone.0016708-Briggs2], Bd reservoirs [@pone.0016708-Mitchell1], the presence of pesticides [@pone.0016708-Davidson2], the presence of heavy metals (in tadpoles) [@pone.0016708-Parris1], drought [@pone.0016708-Lampo2], climate change [@pone.0016708-Pounds1], [@pone.0016708-Lips5], and normal climatic oscillations [@pone.0016708-Anchukaitis1]. Some amphibians, especially aquatic salamander species, African Clawed Frogs (*Xenopus laevis*), and ranids such as Bullfrogs (*Lithobates catesbeiana*), Northern Leopard Frogs (*L. pipiens*), and Rio Grande Leopard Frogs (*L. berlandieri*) are suspected to be carriers of this disease [@pone.0016708-Davidson1], [@pone.0016708-Daszak2], . Across amphibian species, behavioral, natural history, and life history features are known to affect the course of Bd infection [@pone.0016708-Lips6], [@pone.0016708-Rowley1], [@pone.0016708-Rdder1]. Chytridiomycosis may be most fulminant in cool, high-humidity habitats such as cloud forests and splash zones around streams [@pone.0016708-Kriger1], [@pone.0016708-Woodhams6], [@pone.0016708-Grant1]. Because Bd infects keratin-producing cells [@pone.0016708-Longcore1], [@pone.0016708-Berger3], [@pone.0016708-Voyles1], it affects the skin of adult frogs by disrupting physiological functions such as electrolyte balance, and can be fatal [@pone.0016708-Berger1], [@pone.0016708-Pessier2], [@pone.0016708-Voyles1], [@pone.0016708-Marcum1]. In tadpoles, however, where the skin has not yet developed keratin, Bd attacks only mouthparts [@pone.0016708-Blaustein1], [@pone.0016708-Fellers1], [@pone.0016708-Parris2], [@pone.0016708-Rachowicz3], [@pone.0016708-Smith1], [@pone.0016708-Symons1] and tadpoles can act as reservoirs for the disease [@pone.0016708-Briggs2]. During metamorphosis, Bd can spread from tadpole mouthparts to the skin and kill juveniles [@pone.0016708-Berger1], [@pone.0016708-Rachowicz4]. Variations in natural history and life history features of amphibians should produce different patterns, course, and effect of Bd infection [@pone.0016708-Briggs2], [@pone.0016708-Woodhams7]. To fully comprehend this disease---that is, to observe and understand variations in its life history---it will be essential to understand how Bd affects amphibians across their remarkably diverse natural history and life history patterns. Crawfish Frogs (*L. areolatus*) are members of the *Nenirana* subgenus [@pone.0016708-Hillis1]. The other members of this group are Gopher Frogs (*L. capito*), Dusky Gopher Frogs (*L. sevosus*), and Pickerel Frogs (*L. palustris*). Both Gopher Frog species use Gopher Tortoise (*Gopherus polyphemus*) burrows, stump holes, small mammal burrows, and other retreats as refuges; Crawfish Frogs obligately utilize crayfish burrows, therefore both Gopher Frogs and Crawfish Frogs rely on other animals to create upland retreats. Given this dependence, it is no surprise that all three species are imperiled: Dusky Gopher Frogs are listed as Federally Endangered, Gopher Frogs and Crawfish Frogs have experienced sharp declines in population numbers. In Indiana, where this study was conducted, Crawfish Frogs are State Endangered. Crawfish Frogs exhibit a notable life history/natural history pattern from the perspective of disease transmission and the broader issue of epidemiology. While Crawfish Frogs resemble most North American frogs in forming spring breeding aggregations in fishless seasonal and semipermanent wetlands, they are unique in that when not breeding they usually live singly, isolated from other Crawfish Frogs, in burrows dug by crayfish. Crawfish Frogs may occupy single burrows for long periods of time [@pone.0016708-Hoffman1]. Crayfish burrows are narrow bore but deep, extending to the water table perhaps a meter or more below the soil surface [@pone.0016708-Thompson1]. During warm seasons, Crawfish Frogs occupy the upper portion of their crayfish burrow (a plow depth of 7 or 8 cm will excavate frogs) [@pone.0016708-Hossak1], either in their burrow, at the burrow entrance with their heads out, or out of their burrows on their "feeding platforms" [@pone.0016708-Hoffman1]. Time-lapse photography reveals that Crawfish Frogs will spend long periods---days at a time---outside their burrows on their feeding platforms [@pone.0016708-Hoffman1]. At these times frogs can be active around the clock, including the hottest portions of the hottest days of the year (\>37°C). When out of their burrows, Crawfish Frogs generally do not leave the feeding platform unless to lunge at prey, and then immediately return to their feeding platform [@pone.0016708-Hoffman1]. During the winter, and perhaps during the summer when rehydrating, Crawfish Frogs will sit in the water at the bottom of the burrow (JLH, unpubl.). This water is about what you would imagine it to be after a season\'s (or more) accumulation of frog excrement. Thompson [@pone.0016708-Thompson1] writes: "At the bottom of the frog burrows, which usually terminated at a distance of about three feet, was a mass of foul smelling clayey material containing quantities of beetle remains and considerable dead grass, the latter probably having been washed in or accidentally carried down by the frog." When burrows are flooded following heavy rains, Crawfish Frogs will also be submerged, but in presumably cleaner water (more dilute with a reduction in solids) near the burrow entrance (JLH, unpubl.). Here we report the first case of chytridiomycosis in Crawfish Frogs. More importantly, given the unusual natural history features of Crawfish Frogs, we describe the nature and the course of this disease in this species. We ask whether there is a life history pattern or a seasonal pattern to infection by this fungus, and whether we can determine when and where the infection is being acquired and shed. Given the tenuous conservation status of this species, we were also concerned whether chytridiomycosis is fatal to Crawfish Frogs or whether, as with other large North American ranids, they are carriers. Of course, given the idea of Vredenburg and colleagues\' that an infection intensity of 10,000 zoospore equivalents leads to amphibian declines, both situations could be true [@pone.0016708-Vredenburg1], [@pone.0016708-Briggs2]. Materials and Methods {#s2} ===================== Ethics Statement {#s2a} ---------------- This research was conducted under IACUC number 3-24-2008 issued by Indiana State University, and Scientific Purposes License Permit number 09-0084 issued by the Indiana Department of Natural Resources. No animals were harmed while collecting Bd samples. Field Samples {#s2b} ------------- Crawfish Frogs were handled with nitrile gloves and swabbed using cotton, wood-handled swabs. Swabs were rubbed by rolling the cotton over the body surface [@pone.0016708-Pessier1]; five rubs each on the back, sides, belly, head, between the thighs, and the bottom of each foot for a total of 50 rubs. The head of the swab was then broken off in an individually labeled 0.6 ml microcentrifuge tube (Fisherbrand 05-407-01), stored cold and shipped on ice packs prior to analysis. ### Breeding Adults {#s2b1} Breeding adults were captured along drift fences [@pone.0016708-Heemeyer1] or in pitfall traps (buckets) adjacent to drift fences as they attempted to enter or exit two wetlands. Nate\'s Pond is a seasonal/semipermanent wetland with a surface area of 1,355 m^2^ and a perimeter of 208 m that usually dries by late summer; Cattail Pond is a semipermanent/permanent wetland with a surface area of 3,287 m^2^ and a perimeter of 255 m. Adults in a third wetland (Big Pond; surface area 10,146 m^2^; perimeter 573 m), too large for us to encircle with a drift fence and monitor, were captured in mesh traps. In 2009, 66 breeding Crawfish Frogs were sampled as follows: 41 breeding Crawfish Frogs were sampled at drift fences at Nate\'s Pond, 21 were sampled at drift fences at Cattail Pond, and four were sampled from mesh traps deployed at Big Pond ([Table 1](#pone-0016708-t001){ref-type="table"}). Breeding frogs were sampled between 31 March and 15 May. In 2010, 65 breeding Crawfish Frogs were sampled, as follows: 44 animals were sampled entering or exiting Nate\'s Pond, 20 frogs were sampled from Cattail, one frog was sampled from Big Pond. Sampling dates of breeding animals ranged from 12 March to 20 May. Of these animals, 14 from Nate\'s Pond and seven from Cattail Pond had been sampled in 2009. ::: {#pone-0016708-t001 .table-wrap} 10.1371/journal.pone.0016708.t001 Table 1 ::: {.caption} ###### Number of adult Crawfish Frogs sampled per pond per year. ::: {#pone-0016708-t001-1} Nate\'s Cattail Big Burrow Total breeding Overall Total ------- --------- --------- ----- ------------------------------------- ---------------- --------------- 2009 41 21 4 12 66 78 2010 44 20 1 13[\*](#nt101){ref-type="table-fn"} 65 65 Total 85 41 5 25 \*Not including four frogs sampled as they emerged from their burrows after overwintering. ::: After 30 March, 2009, all entering and exiting Crawfish Frog adults were sampled, except if they were handled in a way that might have contaminated the sample. An additional 12 samples were excluded due to cross-contamination during the laboratory analysis. In Nate\'s Pond, 154 Bd samples were analyzed from 84 frogs as follows ([Fig. 1](#pone-0016708-g001){ref-type="fig"}): 27 frogs were sampled once, 28 were sampled twice, six were sampled three times, eight animals were sampled four times, two animals were sampled six times, and one animal was sampled nine times. At Cattail Pond, 68 samples were analyzed from 41 frogs as follows: 12 frogs were sampled once, 13 were sampled twice, six were sampled three times, and three were sampled four times. Five animals were sampled (once) from Big Pond. ::: {#pone-0016708-g001 .fig} 10.1371/journal.pone.0016708.g001 Figure 1 ::: {.caption} ###### The number of times individual Crawfish Frogs were sampled for Bd across our dataset. Most frogs were sampled once or twice, one frog was sampled eight times. :::  ::: ### Upland Adults {#s2b2} During the summer and fall of 2009 (21 July to 21 October) and the summer of 2010 (16 April to 23 July), 25 upland Crawfish Frogs (12 in 2009, 13 in 2010; one animal each year was sampled twice) were swabbed after extricating them from their crayfish burrows [@pone.0016708-Heemeyer2]. Crawfish Frogs were extracted for reasons other than disease sampling, because we wished to either replace radiotransmitters, determine the status of surgical incisions following internal transmitter implantation, or determine if external belt-attached transmitters were abrading the skin. ### Juveniles {#s2b3} Newly metamorphosed juveniles were captured along drift fences while exiting wetlands. In 2009, 52 juvenile Crawfish Frog samples (40/286 from Nate\'s, 10/11 from Cattail, two found associated with other wetlands), collected between 19 June and 17 August, were selected for analysis. In 2010, postmetamorphic juveniles were sampled randomly from 5 June to 17 July as follows. All animals sampled were from Nate\'s Pond; there was no Crawfish Frog metamorphosis at Cattail (VCK, unpubl.). We swabbed the first animal processed from each bucket to avoid pseudoreplication due to cross-contamination. A total of 264 swabs were taken; from these, a subsample of 99 swabs (representing 4.7% of juveniles, and 38% of swabs) were analyzed. ### Adults Emerging from Overwintering Burrows {#s2b4} In 2010, four Crawfish Frog adults were sampled immediately after emerging from overwintering burrows (between 3 March and 24 March), prior to beginning their breeding migrations. These animals were either captured on the night they first emerged within mesh fences placed around burrows, or hand captured within 2 m of their burrow. ### Post-mortem {#s2b5} Four adult Crawfish Frog carcasses were swabbed for Bd. One animal was found freshly killed (blood had not yet coagulated and the body was not in rigor); three had died some time (from days to weeks) prior to being sampled. ### Other Bd samples {#s2b6} In 2009, samples from 15 sponges (sponges were placed, one each, in pitfall trap buckets to provide a floating substrate during bucket flooding and a source of water during dry conditions) were analyzed for the presence of chytrid. Eighteen newly-metamorphosed Marbled Salamanders (*Ambystoma opacum*), the most abundant amphibian species at our wetlands in 2009 [@pone.0016708-Lannoo1], and two Smallmouth Salamanders (*A. texanum*) were also swabbed. Laboratory Analyses {#s2c} ------------------- ### PCR techniques {#s2c1} In 2009, Bd swabs were analyzed using conventional PCR (polymerase chain reaction) techniques [@pone.0016708-Annis1] in the laboratory of Dr. Irene Macallister. Briefly, to extract Bd DNA from field samples, one ml of 70% ethanol was added to microcentrifuge tubes containing sample swabs and stored overnight at −20°C. Swabs were removed and the supernatant was centrifuged (16,000×g for 10 min). Then, ATL-PK (Qiagen) tissue lysis buffer (200 ml) was added to the pelleted fraction and incubated overnight (55°C). To detect Bd spores, a nested PCR approach was used [@pone.0016708-Gaertner2]. Amplification products were visualized on a 3% agarose gel (Ameresco agarose 3∶1 HRB). Presence or absence of a 300-bp band was compared against the EZ Load 100-bp molecular ruler (Bio-Rad) and a positive control. Negative controls were also run with each sample; samples were analyzed twice. Following the seasonal pattern of Bd uptake and loss detected in 2009 (see below), we decided to sample a second year (2010) using real-time Taqman PCR, a more sensitive analytical technique. In particular we were concerned about the presence of false negatives (Bd present but not detected for reasons of analytical or diagnostic sensitivity) [@pone.0016708-Pessier1]. Because we usually sampled the same individuals more than once (see below), a single negative result within a run of positive samples could either indicate acquisition, shedding, and re-acquisition of the infection, or could be the result of Bd present but not detected for analytical or diagnostic reasons. To facilitate the correct interpretation of these data, we wished to reduce the possibility of false negatives. For Taqman PCR, we followed the method of Boyle [@pone.0016708-Boyle1], [@pone.0016708-Hyatt1]. Briefly, template DNA was prepared by treatment of air-dried rayon-tipped swabs (Dryswab™ Fine Tip MW113; United States: [www.mwe-usa.com](http://www.mwe-usa.com)) with Prepman Ultra (Applied Biosystems/Life Technologies, Carlsbad, CA). PCR assays were run on a ABI/Applied Biosystems 7900HT thermocycler using 384 well plates with Applied Biosystems exogenous internal positive control labeled with Vic in separate wells to test for the presence of PCR inhibitors. For each sample, 5 uL of 1∶10 dilution (10 uL Prepman Ultra DNA extract and 90 uL water) swab DNA was added to each well with final total volume of 20 uL. Standard curves were generated with 10-fold serial dilutions (range: 10,000 to 0.001 zoospores) of laboratory cultivated *B. dendrobatidis* zoospores. With Taqman PCR, fluorescent reporter probes are used to detect Bd spores. Internal controls were run to detect the presence of PCR inhibitors. Samples were run in triplicate. Intensity of infection from Taqman PCR results was expressed as zoospore equivalents/swab. ### Histology {#s2c2} Following our first suspected deaths from Bd in 2009, fresh carcasses were analyzed histologically (using conventional paraffin section and staining techniques) [@pone.0016708-Kiernan1], [@pone.0016708-Berger4], [@pone.0016708-Berger5] for the presence of Bd [@pone.0016708-Pessier1]. Results {#s3} ======= Breeding Crawfish Frogs {#s3a} ----------------------- Over the course of 2009 and 2010, 110 individual breeding Crawfish Frogs were sampled for Bd ([Table 1](#pone-0016708-t001){ref-type="table"}; several frogs were sampled across years---see below); swabs from 58 animals (53%) tested positive, as follows. In 2009, 44% (11/25) of Crawfish Frogs entering Nate\'s Pond were Bd positive; 37% (11/30) of animals exiting were positive ([Table 2](#pone-0016708-t002){ref-type="table"}; both here and below, the numbers of animals entering and exiting wetlands are not equal due to deaths, trespassing, lost samples, ambiguous sample results, or animals simply staying in wetlands through the summer). Fifty-five percent (6/11) of animals entering Cattail Pond were Bd positive, 59% (10 out of 17) of animals exiting were positive ([Table 2](#pone-0016708-t002){ref-type="table"}). In total in 2009, 47% (17/36) of Crawfish Frogs sampled upon entering wetlands tested positive for Bd; 45% (21/47) of frogs sampled upon exiting wetlands tested positive ([Table 2](#pone-0016708-t002){ref-type="table"}). All four breeding adults caught in mesh traps within Big Pond were Bd positive. ::: {#pone-0016708-t002 .table-wrap} 10.1371/journal.pone.0016708.t002 Table 2 ::: {.caption} ###### Rates of Bd-positive adults entering and exiting Nate\'s Pond and Cattail Pond in 2009 and 2010. ::: {#pone-0016708-t002-2} 2009 2010 -------------- ------------- ------------- ------------ ------------- Nate\'s Pond 44% (11/25) 37% (11/30) 13% (5/38) 43% (15/35) Cattail Pond 55% (6/11) 59% (10/17) 18% (3/17) 67% (6/9) Total 47% (17/36) 45% (21/47) 15% (8/55) 58% (21/44) ::: In 2010 at Nate\'s Pond, 13% (5/38) of Crawfish Frogs entered Bd positive; 43% (15/35) exited Bd positive. In Cattail, 18% (3/17) of Crawfish Frogs entered Bd positive; 67% (6/9) exited positive. In total, in 2010, 15% (8/55) of Crawfish Frogs entered breeding wetlands Bd positive, 58% (21/44) exited breeding wetlands Bd positive. One animal caught in a mesh trap within Big Pond was Bd negative. Combining 2009 and 2010 data (ignoring for the moment that a subset of animals were sampled both years), 25% (16/63) of animals entering Nate\'s Pond were Bd positive, 40% (26/65) of animals exiting Nate\'s Pond were Bd positive. In Cattail Pond, 32% (9/28) of animals entering Cattail Pond were Bd positive, 62% (16/26) of animals exiting were Bd positive. Of the five breeding Crawfish Frogs captured in mesh traps at Big Pond, four (80%) were Bd positive. Overall, across both years and both drift-fenced wetlands, 27% (25/91) of Crawfish Frog adults entered breeding wetlands Bd positive; 46% (42/91) of animals exited Bd positive. From among these animals, 13 Crawfish Frogs from Nate\'s Pond and six from Cattail Pond were sampled at some point in both 2009 and 2010. Of the 21 breeding Crawfish Frogs repeatedly sampled in 2009 (entering and exiting breeding wetlands), 71% (15) did not change their infection status during breeding (seven entered and exited Bd negative, eight entered and exited Bd positive); 29% (six) animals changed their status (three lost the infection, three became infected; [Table 3](#pone-0016708-t003){ref-type="table"}). Among 44 breeding Crawfish Frogs sampled repeatedly in 2010, 68% (30) did not change their infection status during breeding (24 entered and exited Bd negative, six entered and exited Bd positive); 32% (14 animals) changed their status, all acquired the infection while in breeding wetlands. ::: {#pone-0016708-t003 .table-wrap} 10.1371/journal.pone.0016708.t003 Table 3 ::: {.caption} ###### Summary of Crawfish Frogs arranged according to Bd infection status (Positive or Negative) as they entered and exited wetlands, by wetland and by year. ::: {#pone-0016708-t003-3} Cattail Nate\'s ------------------- --------- --------- ---- ---- Positive→Negative 0 0 3 0 Positive→Positive 5 2 3 4 Negative→Positive 1 4 2 10 Negative→Negative 2 5 5 19 Total 8 11 13 33 ::: Five animals were sampled entering and exiting breeding wetlands in both 2009 and 2010 ([Table 4](#pone-0016708-t004){ref-type="table"}). Of these: two animals were completely negative both years; one animal was negative except when exiting in 2009; one animal was positive entering and exiting in 2009, but negative in 2010; and one animal lost the infection while breeding in 2009 then re-acquired it during breeding in 2010. ::: {#pone-0016708-t004 .table-wrap} 10.1371/journal.pone.0016708.t004 Table 4 ::: {.caption} ###### The Bd infection histories of five Crawfish Frogs, two from Cattail Pond, three from Nate\'s Pond, sampled in both 2009 and 2010. ::: {#pone-0016708-t004-4} 2009 2010 ------------- ------ ------ --- ---- Cattail \#1 − − − − Cattail \#2 − \+ − − Nate\'s \#1 \+ \+ − − Nate\'s \#2 \+ − − \+ Nate\'s \#3 − − − − "In" indicates entering breeding wetland, "Out" indicates exiting breeding wetland. ::: Over the two years of this study, 12% (5/42) of Bd infected frogs that exited wetlands developed chytridiomycosis and died. Histological examination [@pone.0016708-Berger1], [@pone.0016708-Berger3] of the first animal we suspected to have died from chytridiomycosis showed severe epidermal hyperplasia and hyperkeratosis with myriad chytrid thalli consistent with lethal chytridiomycosis (diagnosis confirmed by APP). Swabs from the five animals that died from chytridiomycosis showed consistently high infection intensity ([Fig. 2](#pone-0016708-g002){ref-type="fig"}), ranging from a mean of 2,104 to 24,436 zoospore equivalents. These zoospore equivalents were among the eight highest values recorded in this study ([Fig. 2](#pone-0016708-g002){ref-type="fig"}). We do not know the fate of the other three animals; when last seen they were exiting wetlands. ::: {#pone-0016708-g002 .fig} 10.1371/journal.pone.0016708.g002 Figure 2 ::: {.caption} ###### Zoospore equivalents for the 16 frogs with the highest rates of Bd infection (\>100 zoospore equivalents). Values are averages of three analyses from the same swab; where multiple swabs were performed at different times on the same animal, the swab with the highest zoospore equivalents was used. Black bars are animals from Nate\'s Pond, gray bars are from Cattail Pond. Numbers above bars are maximum zoospore equivalents of animals that died from Bd infection. Animals with zoospore equivalents near or \>10,000 that we did not find dead (numbers 12 and 16), were last observed leaving breeding wetlands. Animals with zoospore equivalents \<100 did not show clinical signs of the disease. :::  ::: Upland Adult Crawfish Frog Samples {#s3b} ---------------------------------- All 12 adults in upland crayfish burrows sampled opportunistically during the summer of 2009, and all 13 individuals sampled during the summer of 2010, were Bd negative ([Table 1](#pone-0016708-t001){ref-type="table"}). Of these animals, seven (33%) were Bd positive upon exiting breeding wetlands, and must have shed the infection or exhibited a low-level infection not detected by available PCR assays. Juvenile Crawfish Frog Samples {#s3c} ------------------------------ In 2009, all 52 postmetamorphic juvenile Crawfish Frogs sampled when exiting wetlands from June through August were Bd negative. In 2010, two of the 99 animals sampled tested positive. In total, 1% (2/151 juveniles) tested positive for Bd. This finding corroborates our anecdotal observations that tadpoles in our study wetlands generally have fully keratinized mouthparts (i.e., without signs of de-keratinization characteristic of Bd infection) [@pone.0016708-Fellers1], [@pone.0016708-Skerratt2]. Adults Emerging from Overwintering Burrows {#s3d} ------------------------------------------ Following overwintering, 50% (2/4) of Crawfish Frogs tested positive for Bd (exhibiting an infection intensity of 4 and 56 zoospore equivalents) as they emerged from their burrows. Post-mortem {#s3e} ----------- Of the four adult Crawfish Frog carcasses sampled, one tested positive. This animal was freshly killed. The older carcasses, discovered days to weeks after death occurred, were Bd negative. Other Bd samples {#s3f} ---------------- All pitfall trap sponge samples tested Bd negative. From among the 20 ambystomatid salamanders---18 Marbled Salamanders, two Smallmouth Salamanders---sampled, 8/18 (44%) tested positive. All positive samples were from Marbled Salamanders: four from Nate\'s Pond, four from Cattail. Discussion {#s4} ========== Our results and conclusions are summarized in the empirical model presented in [Figure 3](#pone-0016708-g003){ref-type="fig"}, and detailed here. Crawfish Frogs inhabit two distinct aquatic ecosystems that are potential sources for Bd infection: breeding wetlands, where they congregate with conspecifics as well as with other amphibian species during brief periods (days to weeks; [Fig. 3](#pone-0016708-g003){ref-type="fig"}, top); and crayfish burrows, where they generally live alone for most of the remainder of the year (10--11 mo; [Fig. 3](#pone-0016708-g003){ref-type="fig"}, bottom, right and left). ::: {#pone-0016708-g003 .fig} 10.1371/journal.pone.0016708.g003 Figure 3 ::: {.caption} ###### A simple model showing the patterns of Bd gain and loss in Crawfish Frogs based on our data. Red box indicates highest rate of Bd infection, blue box indicates cleared or low-level infection, purple box indicates intermediate level infection. Note that following breeding, infected frogs lose the disease during the summer when their activity is centered at the burrow entrance. During the winter, frogs inhabit the water at the base of the burrow, where a subset of animals re-acquire the disease. These infected animals then transmit Bd to their breeding wetland during relatively short (from a few hours to several days) migrations. In breeding wetlands, a subset of animals acquire Bd and a subset shed the disease, but most Crawfish Frogs maintain their status (Bd positive or negative). Some animals exiting wetlands develop chytridiomycosis and die. Exiting juvenile Crawfish Frogs were generally Bd negative (1% infection rate). Juveniles may be exposed to Bd while overwintering during the ≥two years (males) or ≥three years (females) prior to their first breeding attempts, or they may become exposed during their first breeding attempts. Once young Crawfish Frogs begin breeding, they follow the model outlined for breeding adults. :::  ::: Our data suggest that Crawfish Frogs acquire Bd while overwintering in upland burrows or have low-level infections not detected by available PCR assays. While it is recommended that three tests be conducted over a 2-week period to detect all animals with low-level infections [@pone.0016708-Hyatt1], because of the conservation status of Crawfish Frogs and the necessity for us to allow them to perform natural behaviors, we could not do this. Of four frogs sampled immediately upon emerging from overwintering burrows, two (50%) were Bd positive, with a low infection intensity (4 and 56 zoospore equivalents). Overall, 27% (25/91) of samples from Crawfish Frogs entering breeding wetlands on our study site were Bd positive. Our data also suggest that Crawfish Frogs acquire Bd during breeding activities. For example, a Bd-positive female entered Nate\'s Pond on 8 April, 2010 with a low infection intensity (20 zoospore equivalents) and exited 15 days later with a high infection intensity (8,607 zoospore equivalents). A similar situation occurred on 19 April 2010, when a Bd-positive subadult Crawfish Frog entered Nate\'s Pond with 119 zoospore equivalents and exited 5 days later with 23,006 zoospore equivalents. Overall, 46% (42/91) of samples from Crawfish Frogs exiting breeding wetlands on our study site were Bd positive. When Crawfish Frogs acquired Bd in wetlands, we do not know whether zoospores originated from the wetland substrate, directly from syntopic species of amphibians (e.g., Marbled Salamanders) [@pone.0016708-Lannoo1], or whether the fungus was transmitted through zoospores spread between Crawfish Frogs during breeding-associated activities (through male-male combat or amplexus). If the latter is true, Bd-positive Crawfish Frogs entering wetlands are acting as carriers. Twelve percent (5/42) of Crawfish Frogs sampled exiting breeding wetlands are known to have died as a result of chytridiomycosis. Survivors migrate away from wetlands and eventually into crayfish burrows. As summer proceeds, Bd-positive frogs reduce, and may lose, their infection, perhaps through behavioral thermoregulation by basking on their feeding platforms [@pone.0016708-Piotrowski1], [@pone.0016708-Woodhams1], [@pone.0016708-RichardsZawacki1]. Sample sizes for summer, upland-dwelling Crawfish Frogs were small compared to the number of breeding adults and postmetamorphic juveniles. This could not be helped. Crawfish Frogs are among the most secretive frogs in North America [@pone.0016708-Smith2]---it is extraordinarily unlikely that a field biologist would stumble onto and be able to sample a healthy Crawfish Frog in the open during the summer. We feel it is important that out of the 25 summer burrow-dwelling adults that were sampled (all had radiotransmitters implanted so they could be located, and were swabbed after first being extracted from burrows [@pone.0016708-Heemeyer2] for reasons other than disease monitoring), all were Bd negative. At a Bd infection rate of 25% (approximating the infection rate of animals entering breeding wetlands and assuming no false negatives) the probability of 21 negative samples without a positive sample is \>0.001%, at an infection rate of 50% (approximating the infection rate of animals leaving breeding wetlands), the probability is much lower (5×10^−7^). During winters, a subset of Crawfish Frogs re-acquire Bd; 27% (25/91) of frogs entering our study wetlands tested positive. While it is possible that Crawfish Frogs remain Bd free throughout the winter and instead acquire Bd during breeding migrations, we suspect they do not. Two of four animals swabbed immediately upon emerging from their overwintering burrows were Bd positive. Further, 40% (4 of 10) of telemetered Crawfish Frogs migrated from overwintering burrows to breeding wetlands using a single movement lasting one night (JLH, unpubl.). It is unlikely that a Crawfish Frog could acquire zoospores during an overnight upland migration and test positive for Bd the following morning. The remaining Crawfish Frogs used two movements to migrate from burrows to breeding wetlands; these movements were usually several days apart. When stopped during these migrations, Crawfish Frogs generally use retreats located in upland sites, often under cover of Big Bluestem (*Andropogon gerardii*) or Indian Grass (*Sorghastrum nutans*). We also feel it is unlikely that most Crawfish Frogs acquired the infection in drift fence pitfall traps (buckets). Because of the reluctance of Crawfish Frogs to move laterally along drift fences [@pone.0016708-Heemeyer1], we are present most nights when breeding migrations occur, and capture a large proportion of Crawfish Frogs along the fence or as they approach the fence, before they can enter buckets. Further, all samples of sponges within buckets were Bd negative. One percent of postmetamorphic juveniles ([Fig. 3](#pone-0016708-g003){ref-type="fig"}, center) exit wetlands Bd positive (see [@pone.0016708-Sadinski1] for data on other Midwestern species), and may not be exposed to Bd again until they take up residence in crayfish burrows, or until their first breeding (predicted to be two years later for males, three years for females) [@pone.0016708-Redmer1]. Vredenburg and colleagues [@pone.0016708-Vredenburg1] have suggested that a zoospore equivalent of approximately 10,000 triggers amphibian declines. Our data support this assertion. Four of our five Crawfish Frog deaths occurred in animals that exhibited zoospore equivalents near or \>10,000 (frog numbers 11, 13, 14, 15; [Fig. 2](#pone-0016708-g002){ref-type="fig"}); the remaining frog was last sampled 15 days prior to being found dead, near the drift fence, presumably on its way back into Cattail Pond. Cattail Pond had consistently higher rates of Bd positive animals. Cattail Pond is deeper, cooler, more permanent, and supports Green Frog (*L. clamitans*) and Bullfrog adults and larvae---potential carriers to sustain infection---throughout most years. Nate\'s Pond, in contrast, is shallower, warmer and semipermanent; in 2009 it dried by early September then rehydrated following heavy mid-October rains. Differences in temperature and hydroperiod may account for the differences in infection rate between animals exiting the two wetlands (a total of 40% \[26/65\] for Nate\'s, 62% \[16/26\] for Cattail), although these two factors would not account (at least directly) for the differences in infection rate among animals entering wetlands (25% \[16/63\] for Nate\'s, 32% \[9/28\] for Cattail). The overall trend both years was for Cattail Pond to have fewer breeding adult and juvenile Crawfish Frogs present, but to have a higher percentage of these animals Bd positive. In contrast, among Bd-positive animals, infection intensity, as judged by zoospore equivalents, was over four times higher at Nate\'s Pond (x = 4,685±8,999) than at Cattail Pond (x = 1,367±2,797), a significant difference (p = 0.01). Following Crawfish Frog breeding and juvenile metamorphosis, Bd may be sustained in Nate\'s Pond (at least temporarily) and Cattail Pond (throughout most years) through the presence of the 12 other syntopic amphibian species [@pone.0016708-Lannoo1]. It is more difficult to understand the persistence of Bd in the water at the base of upland crayfish burrows during the summer. Bd zoospores persist in sterilized pond water containing organic materials for as long as seven weeks, and survive at least 12 weeks in sterilized sand [@pone.0016708-Johnson1], [@pone.0016708-Johnson2]. But for Bd to be able to re-infect adult Crawfish Frogs in the bottom of their burrows when Crawfish Frogs at the entrance are Bd negative, zoospores would need to remain viable for up to six months (26 weeks). We have considered the possibility that crayfish, which can share burrows with Crawfish Frogs (JLH, unpubl.), may be transmitting the infection from wetlands to burrows, but a study demonstrating that other crustaceans (freshwater shrimp) host Bd [@pone.0016708-Rowley2] was almost immediately retracted [@pone.0016708-Rowley3]. Unlike other species of chytrid fungus, a Bd resting spore has not been identified [@pone.0016708-Longcore1]. A dormant life history stage, perhaps after sexual reproduction [@pone.0016708-Morgan1], could account for the persistence of Bd in crayfish burrows. Conversely, confinement in a small burrow might increase auto-re-infection by Bd. We can imagine a scenario where a Crawfish Frog enters hibernation with a low-level infection (perhaps undetectable by testing) or acquires the infection while overwintering. Over time, this low-level infection releases zoospores that infect adjacent skin cells on the same frog and intensity builds over time (maybe into the range detectable by testing). We plan on coupling methods to non-destructively sample water in crayfish burrows with techniques for detecting Bd in environmental samples [@pone.0016708-Kirshtein1], [@pone.0016708-Walker1] to determine whether Bd is present in crayfish burrows and if so, the nature, if any, of seasonal variations in density. Auto-re-infection may explain the differences in Bd infection rate in animals entering breeding wetlands between years. In 2010, infection rates of animals were lower at both wetlands (at Nate\'s Pond 44% of Crawfish Frogs were Bd positive in 2009, 13% in 2010; at Cattail Pond 55% were positive in 2009, 18% in 2010). At face value, these numbers suggest Bd was less fulminant in 2010, and this may be true. However, numbers of breeding Crawfish Frogs were substantially reduced in 2010 compared with 2009: Nate\'s Pond exhibited a 39% drop (69 in 2009, 42 in 2010); Cattail Pond exhibited a 25% drop (28 in 2009, 21 in 2010). It is possible that Bd prevalence was less in 2010 because Bd mortality was higher during the winter of 2009/2010--- that is, animals that might usually be infected subclinically instead developed chytridiomycosis due to auto-re-infection and died. While this remains speculation, this interpretation is consistent with the observations that wetter conditions promote Bd infection, and that the fall of 2009 was unusually wet. In October, 114 mm of rain fell---28.7 mm above the 10-yr monthly average---with heavy rains coming on the 8th, 9th and 14th. These rains raised the water table to the soil surface and inundated Crawfish Frog burrows, and for much of the winter the water table remained near the soil surface [@pone.0016708-Heemeyer3]. We do not have enough yearly data to tie differences in Bd infection rates to environmental (temperature and moisture) conditions [@pone.0016708-Gaertner3], but the data from 2009 and 2010 suggest that we might expect more annual variation in Bd infection rates in upland-dwelling frogs such as Crawfish Frogs than in aquatic frogs such as Mountain Yellow-legged Frogs [@pone.0016708-Vredenburg1], where water is always present and temperature extremes are moderated. Finally, the behavior of some Bd-positive frogs at drift fences differed from the behavior of non-infected animals. Normally, Crawfish Frogs crossed our drift fences twice: once to enter wetlands prior to breeding, and once to exit after breeding. But several Crawfish Frogs repeatedly crossed our drift fences, and these animals tended to be Bd positive. In 2009, 73% (8/11) of Crawfish Frogs that crossed the fence more than twice (one entry, one exit) were Bd positive; one Bd-positive frog crossed the fence eight times (in 36 d). In 2010, 100% (6/6) of Crawfish Frogs that crossed the fence more than twice were Bd positive. We suspect the innate drive to leave wetlands following breeding was countered by the inability to osmoregulate due to chytridiomycosis [@pone.0016708-Voyles1], [@pone.0016708-Marcum1], and animals moved towards or away from wetlands depending on which urge was strongest. A subset of these animals (five) later died. One male from Big Pond was found Bd positive entering Nate\'s Pond on 5 May 2010; it never exited. Crawfish Frogs in Indiana were once described as "locally plentiful" until around 1970, when many populations began to experience unexplained declines---extirpations in the absence of habitat loss [@pone.0016708-Minton1]. We do not know what our observed annual mortality rate of 12% of breeding adults due to chytridiomyosis means to the survival of Crawfish Frog populations, but given the hypothesis of Ouellet and colleagues [@pone.0016708-Ouellet1], we offer that at least a portion of these declines were due to the arrival of Bd in southwestern Indiana 40 years ago. We thank Katie Smith for supporting this project, Ron Ronk for allowing us to work at Hillenbrand, and Tenia Wheat, John Ryan, Alex Robinson, David Bakken, Austin McClain, and Shane Stephens for field assistance. We thank Irene Macallister and Neil Wesslund of the U.S. Army Engineer Research and Development Center, Construction Engineering Research Laboratory, Champaign, IL for analyzing our 2009 chytrid samples. Thanks are also extended to Mark Schrenzel, Isa Navarrete, and Kristin Benson of the Amphibian Disease Laboratory of the San Diego Institute for Conservation Research for performing Taqman PCR, and Yvonne Cates for histologic processing. Finally, thanks to Pete Lannoo for creating [Figure 3](#pone-0016708-g003){ref-type="fig"}, and for Susie Lannoo and Nate Engbrecht for proof reading earlier drafts of this manuscript. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**Support for this project came from a State Wildlife Grant (E2-08-WDS13) awarded through the Indiana Department of Natural Resources (<http://www.in.gov/dnr>), start-up funds provided by Indiana University School of Medicine (<http://www.medicine.iu.edu>), and a United States Department of Defense Legacy grant (\# 09-426; <https://www.dodlegacy.org/legacy/index.aspx>). The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: MJL VK JH AP. Performed the experiments: MJL VK JH AP. Analyzed the data: MJL VK JH AP. Contributed reagents/materials/analysis tools: MJL AP. Wrote the paper: MJL VK JH AP.

PubMed Central

2024-06-05T04:04:19.762206

2011-3-10

PMC3053365

Introduction {#s1} ============ Gene duplication followed by the fixation of a mutation providing a different function, is one of the major sources to create genetic novelty [@pone.0017512-Osada1]. The rate at which eukaryotic duplicated genes are retained, i.e., go to fixation, has been originally estimated to be 0.01 per gene per million years [@pone.0017512-Lynch1]. This value was obtained under the assumption that the age of the duplication events can be estimated by looking at within species synonymous divergence rates between pairs of paralogous genes. Nevertheless, this estimate is inflated due to concerted evolution [@pone.0017512-Osada1], [@pone.0017512-Gao1]. Concerted evolution arises due to frequent gene conversion between paralogous genes. This process leads to a severe reduction in the divergence rate between paralogous genes from the same species but not when comparing different species (see for instance [@pone.0017512-Cornman1]). Using species of the *D. melanogaster* subgroup, and taking into account the effect of concerted evolution, i.e, using a phylogenetic approach, Osada and Innan [@pone.0017512-Osada1], estimated the rate of duplication to be 0.001 per gene per million years, an order of magnitude below the original estimate. Not all gene duplicates are predicted to be equally retained. For instance, duplication of genes, that encode for proteins that are part of a complex, are likely deleterious [@pone.0017512-Papp1], [@pone.0017512-Deutschbauer1], [@pone.0017512-Soyer1]. Moreover, theory suggests that, duplication of genes that encode for proteins involved in regulatory networks are rarely retained, since they likely disrupt network dynamics and consequently the expression pattern of many genes [@pone.0017512-Wagner1]. Duplications of genes encoding for proteins involved in signaling networks are also expected to be rarely retained [@pone.0017512-Wagner1]. Gene duplicates that encode for proteins that participate in many reactions are, as well, less likely to be retained than genes that encode proteins that participate in a single reaction [@pone.0017512-Sopko1]. Duplicates of genes that encode for activators are also expected to be more frequently retained than genes that encode for receptors [@pone.0017512-Soyer1]. In *Drosophila*, developmental constraint, for instance, does appear to reduce gene duplicability, but the effect is moderate [@pone.0017512-Yang1]. How the gene duplicates came to be also influences gene duplicate retention. For instance, in *Arabidopsis*, when large-scale duplication events are involved, genes that encode transcription factors, proteins with kinase activity, proteins that are involved in protein binding and modification, or in signal transduction pathways are retained at high rates, but the same categories are retained at low rates when small-scale duplications are involved [@pone.0017512-Maere1]. As discussed by Maere *et al.* [@pone.0017512-Maere1] large scale duplication events may not disrupt stoichiometric balances, while small-scale duplication events likely do. In *Drosophila*, however, most duplication events seem to involve less than four genes, and for the vast majority of blocks, the length of the region between the original and the duplicated block is less than 5 Kb [@pone.0017512-Osada1]. No large scale duplications have ever been described in *Drosophila*. Many meiotic pathways are highly conserved across distantly related sexually reproducing eukaryotes (for a review, see [@pone.0017512-Gerton1]). Such conservation could mean that meiotic pathways tolerate little change. Moreover, in *Arabidopsis*, duplicates of genes involved in DNA repair, DNA replication, DNA recombination, and cell-cycle genes are generally little retained [@pone.0017512-Maere1]. This is not surprising since meiosis-related genes are known to participate in multiple pathways, be involved in protein complexes, and, when disrupted, affect multiple aspects of meiosis (see [Table S1](#pone.0017512.s001){ref-type="supplementary-material"}). Nevertheless, the time it takes to complete meiosis is known to be very variable, even among species without developmental holds. Environmental factors (temperature for instance), nuclear DNA content and genotype are among the most important factors affecting meiosis duration [@pone.0017512-Bennett1]. In *Drosophila*, nuclear DNA content is known to vary significantly (the C-values vary between 0.12 and 0.39; <http://www.genomesize.com>), but meiosis duration has been recorded so far only in *D. melanogaster* [@pone.0017512-Bennett1]. Unusual prophase structures such as fibrillar structures apparently coupled to the nucleolus, and multiple nucleoli are also observed in species of the *virilis* group [@pone.0017512-Klasterska1]. These observations suggest that, even within a single genus, such as *Drosophila,* meiosis features are after all variable. Whether *Drosophila* meiosis-related neomorphs (meiosis-related genes with new functions) have evolved is also unknown. This is important, in order to infer the tolerated degree of change of an ancient machine such as the meiotic one. Recently, Anderson *et al.* [@pone.0017512-Anderson1] studied 33 genes involved in meiosis or meiosis-related tasks, such as, chromosome segregation, achiasmate segregation, crossover regulation, double-strand-break formation, heterochromatin binding, recombination and/or repair, sister-chromatid cohesion, spindle assembly, and telomere maintenance. That study revealed that, in *Drosophila*, variability patterns compatible with adaptive protein divergence and polymorphism can be found at four meiosis (*Klp3A*, *Ku80*, *mtrm*, and *ord*) and two telomere maintenance genes (*mre11* and *rad50*). Nevertheless, as argued by Anderson *et al.* [@pone.0017512-Anderson1], the observed patterns can also be explained as a consequence of the fixation/persistence in *Drosophila* populations, of meiotic drive elements (elements that in females influence the preferential sorting of a chromosome to the pronucleus, and thus to the ovule; [@pone.0017512-Anderson1]). If meiotic drive elements are common (about 18% of the meiotic genes surveyed by Anderson *et al.* [@pone.0017512-Anderson1] could show evidence for meiotic drive elements), then such elements could conceivably also increase the probability of fixation, and thus the retention of meiosis gene duplicates. It should be noted, however, that the extent to which the observed within and between species amino acid variation at meiosis genes is adaptive is unknown. In this work, in order to avoid the confounding effect of concerted evolution (see above), a phylogenetic approach is used for the estimation of the rate at which meiosis-related genes are duplicated and retained. A segmental duplication may lead to the simultaneous duplication of many neighboring genes. When segmental duplications are not taken into account, the gene duplication rate is overestimated. Therefore, in this work, the time of origin, as well as the lineage where the gene duplication occurred, is also taken into account, when inferring the number of independent gene duplication events. Due to the methodological approach used, only gene duplications that occurred after the separation of the *Drosophila* and *Sophophora* sub-genera are counted. Recent gene duplicates are expected to be found in tandem, unless they are the result of a segmental duplication, or retrotransposition is involved. Nevertheless, the separation of the two *Drosophila* subgenera occurred about 40 million years ago [@pone.0017512-Russo1]. Therefore, a fraction of the inferred gene duplications may be old. Because gene order can be shuffled due to inversions and translocations, those duplications are no longer expected to be in tandem. Moreover, we infer whether the gene duplicates are functional, since such genes are potential meiosis-related neomorphs. We speculate on whether variation in meiosis gene copy numbers, as well as the appearance of putative neomorphs, can account for the variability in *Drosophila* meiosis features, although these findings must be corroborated by detailed functional studies. Materials and Methods {#s2} ===================== Strains {#s2a} ------- *D. virilis* 1051.49 (Chaco, Argentina); *D. persimilis* 14011-0111.48 (California, USA); *D. willistoni* 14030-0811.16 (Rocha, Uruguay) and *D. mojavensis* 15081-1352.00 (California, USA) were used to address the expression profile of the different genes found to be duplicated and their respective duplicates. Furthermore, in order to determine the age of the *mtrm* gene duplication (*mtrm-dup*) the following species from the *virilis* group of *Drosophila* were used: *D. novamexicana* 15010-1031.00 (Colorado, USA), *D. lummei* 200 (Russia), *D. littoralis* BP41 (Bragança, Portugal), *D. kanekoi* 15010-1061.00 (Sapporo, Japan), *D. ezoana* E20 (Kemi, Finland), *D. montana* Mo1 (Kemi, Finland), *D. flavomontana* 15010-0981-00 (Idaho, USA), *D. lacicola* 15010-0991-00 (New York, USA), *D. borealis* 15010-0961-00 (Minnesota, USA) and *D. borealis* 15010-0961-03 (Idaho, USA). To test the hypothesis of preferential transmission of chromosomes having one of the variants at *mtrm-dup* gene, the following strains were used: *D. a. americana* NN97.4, NN97.8 (Nebraska, USA), W11, W23 (Lake Wappapelo, USA) and *D. a. texana* W29 (Lake Wappapelo, USA), LP97.7 (Louisiana, USA), ML97.5; ML97.4.2 (Louisiana, USA). Genomic DNA extraction {#s2b} ---------------------- Genomic DNA from single males was extracted using the QIAamp DNA Mini Kit from QIAGEN (Izasa Portugal, Lda.) according to the manufacturer\'s instructions. PCR amplification {#s2c} ----------------- Specific primers were developed for each of the genes found to be duplicated and their respective duplicates ([Table S2](#pone.0017512.s002){ref-type="supplementary-material"}). To test the hypothesis of preferential transmission of chromosomes having one of the variants at *mtrm-dup* this gene was amplified in the species from the *virilis* group of *Drosophila* using primers 543F690 and 543R43 as described in Vieira *et al.* [@pone.0017512-Vieira1]. Standard amplification conditions were 35 cycles of denaturation at 94°C for 30s, primer annealing according to [Table S2](#pone.0017512.s002){ref-type="supplementary-material"}, for 45s, and primer extension at 72°C for 3min. RT-PCR {#s2d} ------ Ovaries and testes were dissected from *D. virilis* (1051.49), *D. willistoni* (14030-0811.16), *D. mojavensis* (15081-1350.00) and *D. persimilis* (14011-0111.48). Total RNA was isolated from the dissected tissues using TRIzol Reagent (Invitrogen) according to the manufacturer\'s instructions and treated with *DNase I* (*RNase*-Free) (Ambion). cDNA was synthesized by reverse transcription with SuperScript III First-Strand Synthesis SuperMix for qRT-PCR (Invitrogen). cDNAs were amplified using the PCR conditions described above and the specific primers shown on [Table S2](#pone.0017512.s002){ref-type="supplementary-material"}. Specific primers were also used for the endogenous *ribosomal protein L32* (*RpL32*) as a control for cDNA quality. No-template controls and reactions with RNA that was not reverse transcribed were performed in order to confirm the absence of genomic DNA contamination. Moreover, when possible, primers were designed in order to encompass a region of the gene with one intron. Therefore, the cDNA amplification product is expected to have a shorter size than the amplification product from genomic DNA. The results were analyzed by agarose gel electrophoresis. It should be noted that expression levels of different genes should not be compared since, for instance, amplification product sizes are different, and primer features (such as GC content, or melting temperatures) are different. Direct sequencing was performed using all the PCR products obtained from cDNA amplification as template to confirm the specificity of the primers developed for all the genes found to be duplicated and their respective duplicates. Moreover, for a given gene and its duplicates, when using cDNA, most PCR amplification products have different sizes. Sequencing {#s2e} ---------- The amplification products obtained for the species from the *virilis* group of *Drosophila* using primers 543F690 and 543R43 [@pone.0017512-Vieira1] were cloned using the TOPO-TA Cloning Kit for Sequencing from Invitrogen (Barcelona, Spain). Positive colonies were picked randomly, grown in 5mL of LB with Ampicillin, and plasmids were extracted using the QIAprep Spin Miniprep Kit from QIAGEN (Izasa, Portugal, Lda.). Four colonies were sequenced in order to correct for possible nucleotide missincorporations that may have occurred during the PCR reaction. Sequencing was performed using ABI PRISM Big Dye cycle-sequencing kit version 1.1 (Perkin Elmer, CA, USA) and the primers for the M13 forward and reverse priming sites of the pCR2.1 vector. Sequencing runs were performed by STABVIDA (Lisbon, Portugal). Restriction enzyme typing of a common polymorphism on the *mtrm-dup* gene {#s2f} ------------------------------------------------------------------------- To test the hypothesis of preferential transmission of chromosomes having one of the amino acid variants at *mtrm-dup* gene [@pone.0017512-Vieira1] a total of 32 crosses were established corresponding to all possible combinations between *D. a. americana* and *D. a. texana* strains (F0) in both directions. After emergence of new born individuals brother-sister mating was performed (F1). All the females were heterozygous for the *mtrm-dup* amino acid variants. In the next generation (F2) 10 males from each of the F1 crosses established were selected in a total of 320 individuals. The genomic DNA from these individuals was extracted and they were genotyped for the presence of the amino acid variant on *mtrm-dup* associated with the *X*/*4* fusion, using the restriction enzyme *Bst*BI and the PCR amplification products obtained with primers 543F69 and 543R43 (see [@pone.0017512-Vieira1]). Datasets, sequence alignment and phylogenetic analyses {#s2g} ------------------------------------------------------ The *D. melanogaster* coding sequences of the 33 meiosis-related genes listed in [@pone.0017512-Anderson1], was retrieved from FlyBase (<http://flybase.org/>). In order to retrieve sequences from non-*melanogaster Drosophila* species, the tblastn option with standard parameters, as implemented in FlyBase, was used. The *D. melanogaster* coding sequences were used as a query. Coding sequences with an associated expected value less than 0.05 were retrieved. When gene sequences were non-annotated, a tentative manual annotation of the putative coding region was performed. For every gene dataset, translated amino acid sequences were aligned using CLUSTALW, as implemented in DAMBE [@pone.0017512-Xia1]. The resulting amino acid alignment was used as a guide to obtain the corresponding nucleotide alignment. Bayesian trees were obtained using MrBayes [@pone.0017512-Huelsenbeck1], and nucleotide sequences, under the GTR model of sequence evolution, thus allowing for among-site rate variation and a proportion of invariable sites. Third codon positions are allowed to have a gamma distribution shape parameter that is different from that of first and second codon positions. Two simultaneous and completely independent analyzes, starting from random trees, were run for 500,000 generations (each with one cold and three heated chains). Samples were taken every 100th generation. The first 1250 samples were discarded (burn-in). The final datasets (accession numbers for the nucleotide sequences used can be found in [Table S3](#pone.0017512.s003){ref-type="supplementary-material"}) were obtained after inspecting the results of the phylogenetic analyses. Because of the methodology used, only gene duplications that occurred after the separation of the *Drosophila* and *Sophophora* sub-genera are counted. Since our goal was to estimate the rate of duplication of meiosis-related genes, whether the duplicated genes were created as a result of the duplication of a segment of the genome (segmental duplications) or as the result of the duplication of a single gene was not assessed. Nevertheless, given the inferred time of origin and the lineage where gene duplications are inferred to have occurred, the detected gene duplications must be the result of independent duplication events (see [Results](#s3){ref-type="sec"} section). Divergence estimates {#s2h} -------------------- Per site non-synonymous (*K~a~*) and synonymous (*K~s~*) rates were estimated using DNasp [@pone.0017512-Rozas1]. Values are Jukes-Cantor corrected for multiple hits. Tajima\'s relative rate tests {#s2i} ----------------------------- In order to infer whether duplicated genes evolve at the same rate as the gene that was duplicated, Tajima\'s relative rate tests were performed, as implemented in the MEGA software [@pone.0017512-Kumar1], using all codon positions, or third codon positions (those most likely to be neutral) only. For this test, two ingroup and one outgroup sequences are used. Under the molecular clock hypothesis, irrespective of the substitution model and whether or not the substitution rate varies with the site, the number of mutations inferred for the two ingroup branches should be similar. If this hypothesis is rejected, then the molecular clock hypothesis can be rejected for this set of sequences. When the two ingroup sequences have different amino acid constraints but are subject to a similar mutation rate, statistically significant differences are expected when using all codon positions but not when using third codon positions [@pone.0017512-Tajima1]. Results {#s3} ======= The vast majority (85%) of the genes involved in meiosis related tasks are not duplicated {#s3a} ----------------------------------------------------------------------------------------- Of the 33 meiosis-related genes studied (those listed in [@pone.0017512-Anderson1]), 31 could be found in the 12 publicly available *Drosophila* genomes (<http://flybase.org/>) although a non-negligible fraction is non-annotated or likely miss-annotated ([Table S4](#pone.0017512.s004){ref-type="supplementary-material"}). The *c(3)G* gene could not be found in *D. ananassae*. Nevertheless, it is found in all other species examined and thus, it is likely that the *D. ananassae* genomic region encompassing gene *c(3)G* has not been sequenced. Gene *CG7676* (also known as *cona*; <http://flybase.org/>) could not be found in *D.ananassae*, *D. willistoni*, *D. mojavensis*, *D. virilis* and *D. grimshawi*. Therefore, the latter gene is never found in species of the *Drosophila* subgenus. In [Fig. 1](#pone-0017512-g001){ref-type="fig"}, the per site non-synonymous rate of evolution between *D. melanogaster* and *D. virilis* is shown for the 33 meiosis-related genes. For *CG7676* gene this value has been extrapolated under the assumption of a molecular clock and that *D. melanogaster* and *D. virilis* have been diverging for about 40 million years while *D. melanogaster* and *D. yakuba* have been diverging for about 10 million years (see [Fig. 2](#pone-0017512-g002){ref-type="fig"}). *CG7676* is not evolving faster than other meiosis-related genes that have a clearly recognizable orthologous copy in *D. virilis* ([Fig. 1](#pone-0017512-g001){ref-type="fig"}). Therefore, we should have been able to detect the *CG7676* orthologous copy in species of the subgenus *Drosophila*. Given these observations it seems likely that gene *CG7676* does not have an orthologous copy in the subgenus *Drosophila*, an unexpected observation for a gene involved in a tightly regulated process. This gene has been described as being required for the stable 'zippering' of transverse filaments to form the central region of the *Drosophila* synaptonemal complex [@pone.0017512-Page1]. ::: {#pone-0017512-g001 .fig} 10.1371/journal.pone.0017512.g001 Figure 1 ::: {.caption} ###### Jukes-Cantor corrected per site rate of non-synonymous substitutions between *D. melanogaster* and *D. virilis* for 33 meiosis genes. For *CG7676* gene this value has been extrapolated under the assumption of a molecular clock and that *D. melanogaster* and *D. virilis* have been diverging for about 40 million years while *D. melanogaster* and *D. yakuba* have been diverging for about 10 million years. :::  ::: ::: {#pone-0017512-g002 .fig} 10.1371/journal.pone.0017512.g002 Figure 2 ::: {.caption} ###### Relationship of the *Drosophila* species studied. Adapted from <http://flybase.org>. Numbers are estimated divergence times in million years. :::  ::: For 26 (*ald, asp, Axs, c(2)M, c(3)G, Su(var)205, Klp3A, Ku70, Ku80, mei-218, mei-41, mei-P22, mei-P26, mei-9, mus304, ncd, okr, ord, rad50, smc1, spn-A, spn-B, spn-D, subito, teflon* and *tefu*) out of the 33 genes analyzed, there is a single copy in the 12 *Drosophila* genomes and thus there is no evidence for gene duplications. For two genes (*MeiW68* and *CG7676*) two copies could be found in *D. sechellia* and *D. yakuba*, respectively. It is, however, likely that these are artifacts of the genome assembly process. Indeed, the two *MeiW68* gene copies are identical at the nucleotide level and the duplicated copy is located on a small scaffold that has not been anchored to any chromosome. The two proteins encoded by gene *CG7676* are 194 and 190 amino acids long. Besides the indel, there is a single nucleotide difference between the two coding sequences. It should be noted that the shorter putatively duplicated gene is located on a small scaffold that has not been anchored to any chromosome. Therefore, we conclude that there is no solid evidence for *MeiW68* and *CG7676* gene duplications. Genus-wide, 85% of the meiosis-related genes do not have duplicates. However, nine independent gene duplications involving the genes *cav*, *mre11*, *meiS332*, *polo* and *mtrm* were found. The 12 *Drosophila* species here analyzed imply about 230 million years of independent evolution ([Fig. 2](#pone-0017512-g002){ref-type="fig"}). Therefore, *Drosophila* meiosis-related genes are duplicated at a rate of 0.0012 per gene per million years. This rate is similar to that estimated for the whole *Drosophila* genome [@pone.0017512-Osada1]. In what follows, for each gene showing duplicates, their evolutionary history, as well as evidence that the gene duplicate(s) are functional is presented. Three independent *cav* gene duplications {#s3b} ----------------------------------------- cav is a DNA-binding protein that is a component of the multiprotein *Drosophila* origin recognition complex [@pone.0017512-Badugu1]. Phylogenetic analyses revealed three independent *cav* gene duplications ([Fig. 3](#pone-0017512-g003){ref-type="fig"}). There is always a *cav* gene on Muller\'s element E, thus it seems reasonable to assume that this is the location of the ancestral *cav* gene. In the four species showing two *cav* copies, the duplicated gene is on three different Muller\'s elements, namely Muller\'s element A (*D. virilis*), element B (*D. willistoni*) or element E (*D. persimilis, D. pseudoobscura*). This finding is compatible with a scenario invoking three independent duplications, as suggested by the phylogenetic analyses. All *cav* gene duplicates have introns ([Table S4](#pone.0017512.s004){ref-type="supplementary-material"}), thus retrotransposition seems an unlikely explanation for the observed change in gene location. It should be noted that the phylogenetic tree presented in [Fig. 3](#pone-0017512-g003){ref-type="fig"} implies that the *cav* gene duplication on Muller\'s element A predates the separation of the *D. grimshawi* and *D. virilis*/*D. mojavensis* lineages, but a duplicated copy cannot be found in either *D. grimshawi* or *D. mojavensis*. Indeed, this *cav* gene duplication is estimated to be as old as the split between the *Sophophora* and *Drosophila* subgenera, and thus about 40 million years old, under the assumption of a molecular clock for synonymous mutations (data not shown). It should be noted that these two *cav* genes are subjected to similar mutation rates but different amino acid constraints ([Table 1](#pone-0017512-t001){ref-type="table"}). The accelerated rate of non-synonymous evolution of the *D. virilis cav-dup* gene (*GJ17001*) could suggest that it is a pseudogene. Nevertheless, this gene is expressed in both males and females ([Fig. 4](#pone-0017512-g004){ref-type="fig"}). ::: {#pone-0017512-g003 .fig} 10.1371/journal.pone.0017512.g003 Figure 3 ::: {.caption} ###### Bayesian phylogram of *Drosophila cav*-like genes. Numbers are posterior credibility values. :::  ::: ::: {#pone-0017512-g004 .fig} 10.1371/journal.pone.0017512.g004 Figure 4 ::: {.caption} ###### Expression patterns of genes found to be duplicated. G -- genomic DNA; F -- female gonads; M -- male gonads. The cDNA of duplicated genes were sequenced in order to assure amplification specificity. In males, for genes *D. virilis cav-dup*, *D. persimilis cav-dup*, *D. mojavensis mre-11-dup* and *D. persimilis polo-dup1* a band with the size expected for an amplification from genomic DNA is observed. In order to rule out the possibility of contamination with genomic DNA, the PCR reactions were performed three times independently starting from different cDNA aliquots and identical results were obtained. The observation that when using the same aliquot, the duplicated gene shows two bands but the genes *D. virilis cav*, *D. persimilis cav*, *D. mojavensis mre-11* and *D. persimilis polo* show only one band of the expected size also shows that there is no contamination with genomic DNA. :::  ::: ::: {#pone-0017512-t001 .table-wrap} 10.1371/journal.pone.0017512.t001 Table 1 ::: {.caption} ###### Tajima\'s relative rate tests using all coding positions or third codon positions only. ::: {#pone-0017512-t001-1} Gene Ingroup species Outgroup All positions Third positions only --------- -------------------------------- ------------------- --------------- ---------------------- *cav* *D. virilis* *D. melanogaster* P\<0.001 P\>0.05 *cav* *D. willistoni* *D. melanogaster* P\>0.05 P\>0.05 *cav* *D. persimilis* *D. melanogaster* P\>0.05 P\>0.05 *cav* *D. pseudoobscura* *D. melanogaster* P\>0.05 P\>0.05 *mre11* *D. mojavensis* *D. grimshawi* P\<0.001 P\>0.05 *polo* *D. persimilis (polo-dup1)* *D. melanogaster* P\<0.001 P\>0.05 *polo* *D. pseudoobscura (polo-dup1)* *D. melanogaster* P\<0.001 P\>0.05 *polo* *D. persimilis (polo-dup2)* *D. melanogaster* P\<0.001 P\>0.05 *polo* *D. pseudoobscura (polo-dup2)* *D. melanogaster* P\<0.001 P\>0.05 *mtrm* *D. willistoni* *D. melanogaster* P\>0.05 P\>0.05 *mtrm* *D. virilis* *D. melanogaster* P\>0.05 P\>0.05 ::: There are two *cav* genes in *D. willistoni* that are under similar amino acid constraint, and thus evolving at the same rate ([Table 1](#pone-0017512-t001){ref-type="table"}). This *cav* gene duplication is estimated to be 10 million years old, under the assumption of a molecular clock for synonymous mutations (data not shown). There is no evidence that *cav-dup* is evolving faster than *cav* ([Table 1](#pone-0017512-t001){ref-type="table"}). The duplicated gene seems to be weakly expressed and in males only ([Fig. 4](#pone-0017512-g004){ref-type="fig"}). There is thus no evidence that it is a pseudogene. Two *cav* genes were also found in the two closely related species *D. persimilis* and *D. pseudoobscura*. This *cav* gene duplication is estimated to be 14 million years old, under the assumption of a molecular clock for synonymous mutations (data not shown). There is no evidence that *cav-dup* is evolving faster than *cav* ([Table 1](#pone-0017512-t001){ref-type="table"}). The duplicated gene is expressed in both males and females ([Fig. 4](#pone-0017512-g004){ref-type="fig"}). A recent *mre11* gene duplication in the *D. mojavensis* lineage {#s3c} ---------------------------------------------------------------- Two *mre11* copies (both on Muller\'s element B) are found in *D. mojavensis* ([Fig. 5](#pone-0017512-g005){ref-type="fig"}). The protein encoded by this gene is involved in telomere maintenance [@pone.0017512-Bi1]. The *mre11* gene duplication is estimated to be about 15 million years old, under the assumption of a molecular clock for synonymous mutations (data not shown). This duplication occurred in the *D. mojavensis* lineage after the separation from the sister group *D. virilis* lineage. It should be noted that the two *D. mojavensis mre11* genes are subjected to similar mutation rates but different amino acid constraints ([Table 1](#pone-0017512-t001){ref-type="table"}). The accelerated rate of amino acid evolution of the *D. mojavensis mre11-dup* gene (*GI20694*) could suggest that it is a pseudogene. Nevertheless, the *mre11-dup* gene is expressed. *mre11-dup* expression levels are higher in males than in females, a pattern also observed for the *mre11* gene ([Fig. 4](#pone-0017512-g004){ref-type="fig"}). ::: {#pone-0017512-g005 .fig} 10.1371/journal.pone.0017512.g005 Figure 5 ::: {.caption} ###### Bayesian phylogram of *Drosophila mre11*-like genes. Numbers are posterior credibility values. :::  ::: Two *polo* gene duplications in the *obscura* group {#s3d} --------------------------------------------------- Polo is a protein kinase that, in *Drosophila* female meiosis promotes nuclear envelope breakdown [@pone.0017512-Smith1]. Three *polo* genes are found in the two closely related species *D. pseudoobscura* and *D. persimilis* ([Fig. 6](#pone-0017512-g006){ref-type="fig"}). The *D. persimilis GL25129* and the *D. pseudoobscura GA11545* genes that are on Muller\'s element D (where the *D. melanogaster polo* gene is also located) are orthologous. The *D. persimilis GL25881* and the *D. pseudoobscura GA25172* genes that are on Muller\'s element B are also orthologous, and are here named *polo-dup1*. The *D. persimilis GL19429* and the *D. pseudoobscura GA25958* genes that are on Muller\'s element B are also orthologous and are here named *polo-dup2*. ::: {#pone-0017512-g006 .fig} 10.1371/journal.pone.0017512.g006 Figure 6 ::: {.caption} ###### Bayesian phylogram of *Drosophila polo*-like genes. Numbers are posterior credibility values. :::  ::: There are three predicted introns in *polo-dup1*. Therefore, retrotransposition seems an unlikely explanation for the observed change in gene location (from Muller\'s element D to element B). This *polo* gene duplication is about 6.5 million years old (under the assumption of a molecular clock for synonymous mutations; data not shown), and is thus expected to be found in species of the *obscura* group only. It should be noted that the two *polo* genes are subjected to similar mutation rates but different amino acid constraints ([Table 1](#pone-0017512-t001){ref-type="table"}). This observation could suggest that *polo-dup1* is a pseudogene. Nevertheless, *polo-dup1* is expressed in males ([Fig. 4](#pone-0017512-g004){ref-type="fig"}). *polo-dup2* is about 12 million years old (under the assumption of a molecular clock for synonymous mutations; data not shown), and thus is also expected to be found in species of the *obscura* group only. There are no introns in *polo-dup2*. Therefore, in this case, retrotransposition could be an explanation for the origin of this duplication. It should be noted that such an hypothesis relies on the quality of the annotation of the *D. pseudoobscura* and *D. persimilis* genomes. *polo* and *polo-dup2* are subjected to similar mutation rates but different amino acid constraints ([Table 1](#pone-0017512-t001){ref-type="table"}). Nevertheless, *polo-dup2* is expressed in males ([Fig. 4](#pone-0017512-g004){ref-type="fig"}), and thus, is unlikely to be a pseudogene. There is one, three, and one fixed amino acid changes between the two *polo-dup1* gene sequences and the other *polo* sequences here analyzed, at the first, second and third Polo boxes, respectively. In general, it is difficult to infer how important these changes might be. It should be noted, however, that the amino acid change observed in Polo box 1 (a change of a V to a I) changes an amino acid that is conserved in *polo* sequences from fungi to humans (see [Fig. 1](#pone-0017512-g001){ref-type="fig"} in [@pone.0017512-Reynolds1]). The *polo-dup2* gene is a truncated version of *polo* where the last one third of the coding region of the gene is missing. Therefore the protein encoded by *polo-dup2* does not show any POLO boxes. Two independent *mtrm* gene duplications {#s3e} ---------------------------------------- The *D. melanogaster* Mtrm protein is a meiosis-specific 1∶1 stoichiometric inhibitor of the Polo kinase protein [@pone.0017512-Xiang1]. This gene is not annotated in most *Drosophila* genomes ([Table S3](#pone.0017512.s003){ref-type="supplementary-material"}) but can be always found within one intron of the *exo70* gene. In *D. willistoni* there are two *mtrm*-like genes, ([Fig. 7](#pone-0017512-g007){ref-type="fig"}), one on Muller\'s element B (that codes for a 186 amino acids long protein) and another one on Muller\'s element D (that codes for a 196 amino acids long protein). Since the *D. melanogaster mtrm* gene is located on Muller\'s element D, it seems likely that the duplicated gene copy is that on Muller\'s element B, and thus this *D. willistoni* copy is here named *mtrm-dup*. Although the two copies are on different Muller\'s elements, at the nucleotide level, the two sequences are 94% identical. Since *mtrm* gene does not have introns the possible involvement of retrotransposition in the translocation of the gene cannot be assessed. This is a recent gene duplication event, estimated to be 1.6 million years old, under the assumption of a molecular clock for synonymous sites. The two genes seem to be under similar amino acid constraint ([Table 1](#pone-0017512-t001){ref-type="table"}). In *D. melanogaster,* phosphorylation sites (including one Polo-box domain binding motif and one Plk-phosphorylation motif, that differs at one amino acid site from the canonical sequence D/E-X-S/T-Ø-X-D/E where Ø is an hydrophobic amino acid), have been reported [@pone.0017512-Xiang1]. Both *D. willistoni mtrm* genes show a Polo-box domain binding motif and a typical Plk-phosphorylation motif in the same protein region as in *D. melanogaster* ([Table 2](#pone-0017512-t002){ref-type="table"}). Moreover, most of the other phosphorylation sites reported for the *D. melanogaster* Mtrm protein are also present in the *D. willistoni* Mtrm and Mtrm-dup proteins. The only phosphorylation site that is not present is also not conserved in Mtrm proteins from other *Drosophila* species. Nevertheless, we could not find any evidence that the *D. willistoni mtrm-dup* is expressed ([Fig. 4](#pone-0017512-g004){ref-type="fig"}). Therefore, the hypothesis that this gene is a recent pseudogene that did not have yet time to degenerate cannot be ruled out. ::: {#pone-0017512-g007 .fig} 10.1371/journal.pone.0017512.g007 Figure 7 ::: {.caption} ###### Bayesian phylogram of *Drosophila mtrm*-like genes. Numbers are posterior credibility values. :::  ::: ::: {#pone-0017512-t002 .table-wrap} 10.1371/journal.pone.0017512.t002 Table 2 ::: {.caption} ###### Mtrm phosphorylation sites [@pone.0017512-Xiang1]. ::: {#pone-0017512-t002-2} Species T40 (STP) S48 S52 S121 S123 S124 S132 S137 ----------------------------------- ----------- ----- ----- ------ ------ ------ ------------------------------------ ------ *D. melanogaster mtrm* \+ \+ \+ \+ \+ \+ \+ \+ *D. simulans mtrm* \+ \+ \+ \+ \+ \+ \+ *D. sechellia mtrm* \+ \+ \+ \+ \+ \+ \+ *D. yakuba mtrm* \+ \+ \+ \+ \+ \+ \+ *D. erecta mtrm* \+ \+ \+ \+ \+ \+ \+ *D. ananassae mtrm* \+ \+ \+ \+ \+ \+ \+ *D. pseudoobscura mtrm* \+ \+ \+ \+ \+ \+ \+ *D. persimilis mtrm* \+ \+ \+ \+ \+ \+ \+ *D. willistoni mtrm* \+ \+ \+ \+ \+ +[\*](#nt102){ref-type="table-fn"} \+ *D. willistoni mtrm-dup* \+ \+ \+ \+ \+ +[\*](#nt102){ref-type="table-fn"} \+ *D. grimshawi mtrm* \+ \+ \+ \+ \+ *D. mojavensis mtrm* \+ \+ \+ \+ \+ \+ \+ *D. virilis mtrm* \+ \+ \+ \+ \+ \+ *D. virilis mtrm-dup* \+ \+ \+ \+ \+ *D. lummei mtrm-dup* \+ \+ \+ \+ \+ *D. novamexicana mtrm-dup* \+ \+ \+ \+ \+ *D. americana texana mtrm-dup* \+ \+ \+ \+ \+ *D. americana americana mtrm-dup* \+ \+ \+ \+ \+ *D. littoralis mtrm-dup* \+ \+ \+ \+ \+ *D. kanekoi mtrm-dup* \+ \+ \+ \+ \+ *D. ezoana mtrm-dup* \+ \+ \+ \+ \+ *D. borealis Western mtrm-dup* \+ \+ \+ \+ \+ *D. flavomontana mtrm-dup* \+ \+ \+ \+ \+ *D. lacicola mtrm-dup* \+ \+ \+ \+ \+ *D. montana mtrm-dup* \+ \+ \+ \+ \+ *D. borealis Eastern mtrm-dup* \+ \+ \+ \+ \+ The referred amino acid positions are those of the *D. melanogaster* Mtrm sequence. \*in agreement with the D/E-X-S/T-Ø-X-D/E pattern where Ø is a hydrophobic amino acid. ::: In *D. virilis* there are also two *mtrm*-like genes, namely, one on Muller\'s element A and another one on Muller\'s element D, being the latter the orthologous of the *D. melanogaster mtrm* gene. *mtrm* and *mtrm-dup* are intronless genes. Therefore, it is not possible to infer the role of retrotransposition in the transposition of this gene from Muller\'s element D to A. Bayesian phylogenetic analyses suggest that this *mtrm* gene duplication predates the separation of the *D. grimshawi*/(*D. mojavensis*/*D. virilis*) lineages ([Fig. 7](#pone-0017512-g007){ref-type="fig"}), and this conclusion is independent of the alignment algorithm used (data not shown). Moreover, *mtrm-dup* is not evolving faster than the *mtrm* gene ([Table 1](#pone-0017512-t001){ref-type="table"}). The pair-wise synonymous divergence values suggest that, under the assumption of a molecular clock, *mtrm-dup* is about 35 million years old, and as such, would indeed predate the separation of the *D. grimshawi*/(*D. mojavensis*/*D. virilis*) lineages. Nevertheless, there is no evidence for a *mtrm* gene duplicate in either *D. grimshawi* or *D. mojavensis*. Therefore, taken at face value, these results imply two independent losses of the *mtrm-dup* gene. The two neighbors of the *D. virilis mtrm-dup* gene (the *D. melanogaster CG7326* and *CG34401* orthologous genes) are gene neighbors in *D. grimshawi* and *D. mojavensis*. Each independent gene loss should be a unique event and thus leave a different genomic signature. Therefore, the comparison of the *CG7326* - *CG34401* region in *D. virilis, D. grimshawi* and *D. mojavensis* could, in principle, shed light on this issue. The intergenic regions can be confidently aligned, as revealed by the per site rates of change, namely 0.36 and 0.54 for the *D. virilis* -- *D. mojavensis* and the *D. virilis* -- *D. grimshawi* comparisons, respectively. The largest insertion, besides the *mtrm* coding region, in *D. virilis* relative to the other two species is only 27 bp long, and in total, there are 61 fixed gapped positions between the *D. virilis* sequence and the *D. grimshawi*/*D. mojavensis* sequences. Therefore, it seems that the only main difference in *D. virilis* relative to the other species is the insertion of the *mtrm* coding region. *mtrm-dup* is not, however, a pseudogene, since this gene is expressed in females ([Fig. 4](#pone-0017512-g004){ref-type="fig"}). The *mtrm-dup* gene could also be amplified from 12 species of the *virilis* group from all major group phylads. Therefore, the *mtrm-dup* gene must be older than the age of the *virilis* group that is estimated to be 10 million years old [@pone.0017512-Reis1]. Although 90% of the coding region of this gene was analyzed in the 12 species of the *virilis* group, no evidence for in-frame stop codons has been found. All *mtrm-dup* sequences show conservation of the T40 (a putative Cdk5 phosphorylation site), S48, S52 (putative Cdk or MAPK phosphorylation sites), S137, and S124 phosphorylation orthologous sites identified in *D. melanogaster* by Xiang *et al.* [@pone.0017512-Xiang1]. The S121 and S123 phosphorylation sites are not conserved in the *mtrm-dup* gene. Nevertheless, not all *mtrm* sequences show conservation of these sites either ([Table 2](#pone-0017512-t002){ref-type="table"}). The *mtrm-dup* gene does not show a Plk phosphorylation-like amino acid motif, due to a four amino acid insertion that is present in all *mtrm-dup* copies. It should be noted, however, that the *D. virilis*, *D. mojavensis* and *D. grimshawi mtrm* amino acid sequences do not have such a feature either, due to a three amino acid insertion. Therefore, in species of the *Drosophila* subgenus, the presence of a Plk phosphorylation-like amino acid motif is not an essential feature. Although *mtrm-dup* is a functional gene, there are no data to support the assumption that this gene plays an essential role in meiosis in species of the *virilis* group of *Drosophila*. Indeed, it is conceivable that this gene represents a non-essential meiotic drive element that went to fixation in the common ancestor of species of the *virilis* group. Once fixed, it may be difficult to lose such an element since chromosomes carrying it are more represented in the next generation than chromosomes carrying alternative deleted copies of this element. Thus, such a gene could show most of the features expected for an essential gene. For *D. melanogaster*/*D. simulans* Anderson *et al.* [@pone.0017512-Anderson1] showed patterns of evolution at the *mtrm* gene that are compatible with both adaptive protein evolution and the sequential fixation of meiotic drive elements. Therefore, this hypothesis is here addressed in *D. americana*, a species of the *virilis* group of *Drosophila*. Vieira *et al.* [@pone.0017512-Vieira1] reported an amino acid polymorphism for *D. americana*, at the gene *CG18543* (*mtrm-dup*) that is a marker for the common polymorphic *X*/*4* fusion. We have followed the transmission of the two types of chromosomes by looking at the male progeny of females heterozygous for the *mtrm-dup* amino acid variant under different genomic backgrounds ([Table 3](#pone-0017512-t003){ref-type="table"}). There is no evidence that the reported amino acid polymorphism represents meiotic drivers of different strength (Chi-square test with one degree of freedom; P\>0.05). ::: {#pone-0017512-t003 .table-wrap} 10.1371/journal.pone.0017512.t003 Table 3 ::: {.caption} ###### Segregation of the common *D. americana mtrm-dup* amino acid polymorphism that is a marker for *X*/*4* fusion chromosomes. ::: {#pone-0017512-t003-3} Crosses ♀x♂ ♂x♀ ------------------- ----- ----- ---- ---- NN97.4 × W29 8 2 5 5 NN97.4 × LP97.7 5 5 6 4 NN97.4 × ML97.4.2 6 4 6 2 NN97.4 × ML97.5 4 6 6 4 NN97.8 × W29 1 9 7 3 NN97.8 × LP97.7 10 0 8 2 NN97.8 × ML97.4.2 4 6 5 5 NN97.8 × ML97.5 6 4 8 2 W11 × W29 7 3 6 4 W11 × LP97.7 6 4 7 3 W11 × ML97.4.2 5 5 4 6 W11 × ML97.5 7 3 6 4 W23 × W29 3 6 2 8 W23 × LP97.7 3 7 6 4 W23 × ML97.4.2 3 7 6 4 W23 × ML97.5 5 5 2 8 Total 83 76 90 68 ::: Concerted evolution at the *Drosophila* subgenus *meiS332*-like genes {#s3f} --------------------------------------------------------------------- *meiS332* gene duplications have been found as well. The phylogeny presented in [Fig. 8](#pone-0017512-g008){ref-type="fig"} suggests that this gene has been independently duplicated three times. Nevertheless, the two copies of the gene are located on Muller\'s element C always with opposite transcription orientations, and at about the same distance. The finding of a similar gene arrangement in *D. virilis*, *D. mojavensis* and *D. grimshawi* thus suggests a unique duplication event, rather than three independent recent duplications. The little divergence observed between the two copies in each species suggests that this is a case of concerted evolution. Concerted evolution has been reported at *Drosophila* genes other than rRNA gene loci (see for instance [@pone.0017512-Cornman1], [@pone.0017512-Somogyi1], [@pone.0017512-Carmon1], [@pone.0017512-Beisswanger1], [@pone.0017512-Hickey1], [@pone.0017512-Wang1]. The *meiS332* gene duplication is an example of long-term (more than 30 million years) concerted evolution in the *Drosophila* subgenus. Similar long-term concerted evolution (also lasting for more than 30 million years) has been reported at the *polyhomeotic* (*ph*) gene duplication in the *Sophophora* subgenus [@pone.0017512-Beisswanger1]. ::: {#pone-0017512-g008 .fig} 10.1371/journal.pone.0017512.g008 Figure 8 ::: {.caption} ###### Bayesian phylogram of *Drosophila meiS332*-like genes. Numbers are posterior credibility values. :::  ::: In *D. melanogaster,* there are two Polo binding sites in MEI-S332, namely SSP from residue 233 to 235, and STP from residue 330 to 332 [@pone.0017512-Clarke1]. As shown in [Table 4](#pone-0017512-t004){ref-type="table"}, the SSP motif is conserved in species of the *melanogaster* subgroup, in the two *D. grimshawi* sequences and in one of the two *D. virilis* sequences. The STP motif is conserved in all sequences with the exception of the *D. mojavensis* duplicated copy. It should be noted that, in *D. melanogaster,* phosphorylation was unaffected by the S234A mutation but was abolished with the T331A mutation [@pone.0017512-Clarke1]. This finding fits our observation of a better conservation of the STP motif than that of the SSP motif. Most MEI-S332 sequences from species of the *Drosophila* subgenus show three S(S/T)P motifs. These findings suggest that, with the exception of the *D. mojavensis* duplicate, all other duplicated genes are functional. Nevertheless, expression was detected in *D. virilis* and *D. mojavensis meiS332-dup* gene. We could not obtain non-mutant *D. grimshawi* strains, and thus expression was not tested in this species. ::: {#pone-0017512-t004 .table-wrap} 10.1371/journal.pone.0017512.t004 Table 4 ::: {.caption} ###### MeiS332 Polo binding sites (SSP and STP). ::: {#pone-0017512-t004-4} Species Motif and amino acid site reference(position in the *D. melanogaster* sequence) ----------------------------- --------------------------------------------------------------------------------- ---- ---- ---- ---- *D. melanogaster meiS332* \+ \+ *D. simulans meiS332* \+ \+ *D. sechellia meiS332* \+ \+ *D. yakuba meiS332* \+ \+ *D. erecta meiS332* \+ \+ *D. ananassae meiS332* \+ *D. pseudoobscura meiS332* \+ *D. persimilis meiS332* \+ *D. willistoni meiS332* \+ *D. grimshawi meiS332* \+ \+ \+ *D. grimshawi meiS332-dup* \+ \+ \+ *D. mojavensis meiS332* \+ \+ \+ *D. mojavensis meiS332-dup* *D. virilis meiS332* \+ \+ \+ *D. virilis meiS332-dup* \+ \+ ::: Discussion {#s4} ========== Nine independent gene duplications involving the genes *cav*, *mre11*, *meiS332*, *polo* and *mtrm* were found. The 12 *Drosophila* species here analyzed imply about 230 million years of independent evolution. Therefore, *Drosophila* meiosis-related genes are duplicated and retained at a rate of 0.0012 per gene per million years. This value is similar to that estimated for the whole *Drosophila* genome using species of the *melanogaster* subgroup [@pone.0017512-Osada1]. The rate at which gene duplicates are created and go to fixation, i.e, are retained, depends on population genetics variables such as birth rate, mutation rate, and effective population size (see for instance, [@pone.0017512-Lynch2]). While it is unlikely that those population genetics variables have remained constant over 230 million years of independent evolution, there is no reason to believe that using all 12 *Drosophila* genomes and all genes would produce an estimate that is substantially different from that provided by Osada and Innan [@pone.0017512-Osada1]. For instance, when the dataset of 33 meiosis related genes is used, the rate of gene duplication and fixation is estimated to be 0.0013 and 0.0011 for species of the *Drosophila* and *Sophophora* subgenera, respectively (the estimate becomes 0.0009 for the *Sophophora* subgenus if the only likely non-functional *D. willistoni mtrm-dup* gene is not included in the calculations; see [Table 5](#pone-0017512-t005){ref-type="table"}). It should be noted that, a detailed analysis of the 33 meiosis genes, revealed that a substantial fraction is non-annotated or likely miss-annotated. Although we do not provide a human-curated annotation for the studied genes in the 12 *Drosophila* genomes, we did analyze in detail the gene annotation for those cases where the non-annotation or miss-annotation could lead to erroneous conclusions (see [Results](#s3){ref-type="sec"}). ::: {#pone-0017512-t005 .table-wrap} 10.1371/journal.pone.0017512.t005 Table 5 ::: {.caption} ###### Summary of the inferences made for the meiosis genes found to be duplicated. ::: {#pone-0017512-t005-5} Duplicated gene copy Location (Muller\'s element) Estimated age in million years Comments -------------------------------------------------------------- ------------------------------ -------------------------------- ----------------------- *D. persimilis* and *D. pseudoobscura cav-dup* A ∼40 Likely functional *D. willistoni cav-dup* B 10 Likely functional *D. virilis cav-dup* E 14 Likely functional *D. mojavensis mre11-dup* B 15 Likely functional *D. persimilis* and *D. pseudoobscura polo-dup1* B 6.5 Likely functional *D. persimilis* and *D. pseudoobscura polo-dup2* B 12 Likely functional *D. willistoni mtrm-dup* B 1.6 Likely non-functional *D. virilis mtrm-dup* A ∼35 Likely functional *D. grimshawi*, *D. mojavensis* and *D. virilis meiS332-dup* C \> 30 Likely functional ::: The finding that functional meiosis-related gene duplications go to fixation at the same rate as the average for all genes is surprising, especially in the light of the complex roles that the genes under study perform (see [Table S1](#pone.0017512.s001){ref-type="supplementary-material"}). Indeed, meiosis-related genes are known to participate in multiple pathways, be involved in protein complexes, and, when disrupted, affect multiple aspects of meiosis (see [Table S1](#pone.0017512.s001){ref-type="supplementary-material"}). It remains to be shown whether the gene duplicates play an essential role in meiosis-related features in the species where they are found. Therefore, it could be argued that they are non-essential meiotic drive gene duplicates that went to fixation. Nevertheless, the segregation experiments performed with the *D. americana mtrm-dup* gene did not reveal evidence for meiotic drive elements. The possibility of subfunctionalization [@pone.0017512-Lynch3] cannot be, however, ruled out. In *Arabidopsis*, gene duplicates involved in DNA repair, replication and recombination, as well as in cell-cycle are little retained [@pone.0017512-Maere1]. The possibility that the duplicated meiosis related genes represent cases of neofunctionalization should thus be addressed by performing additional detailed cellular and biochemical experiments that are beyond the scope of this work. Indeed, about 50% of the gene duplicates are evolving faster than the original gene, a pattern that is compatible with a short period of relaxed selection or/and acquisition of a new function. Moreover, three out of the five genes that have been found to be duplicated are known to physically interact (*meiS332*, *polo* and *mtrm*). There are no reasons to believe that these genes are more prone to accumulate meiotic drive elements or more prone to subfunctionalization. Indeed, given the known function of these genes, they were, *a priori*, unlikely to be found duplicated. The *D. melanogaster* Mtrm protein is a meiosis-specific 1∶1 stoichiometric inhibitor of the Polo kinase protein. In this species activation of Cdc25 by an excess of Polo protein at stage 13 triggers nuclear envelope breakdown and entry into prometaphase [@pone.0017512-Xiang1]. Therefore, any changes in protein levels in either Polo or Mtrm could result in precocious entry into prometaphase or meiotic arrest. On the other hand, Polo antagonizes MeiS332 and removes this protein from centromeres, a step required for proper chromosome segregation at the metaphase II/anaphase II transition [@pone.0017512-Clarke1]. If meiosis is not completed, no gametes will be produced. On the other hand, significant defects in achiasmate segregation (the segregation of chromosomes that did not experience recombination) are observed when there is a precocious entry into prometaphase [@pone.0017512-Xiang1]. Therefore, in what follows we speculate on the conceivable adaptive value of each gene duplicate(s). cav is a DNA-binding protein that is a component of the multiprotein *Drosophila* origin recognition complex [@pone.0017512-Badugu1]. In *Drosophila*, the *cav* gene has been duplicated three times independently. All three independent duplications are old and all *cav* duplicates are expressed. The functional significance of having two *cav* gene copies in *D. virilis* with similar expression patterns is unclear, but it could be related to the high *D. virilis* heterochromatin content. Although *D. melanogaster* and *D. virilis* have similar euchromatin sizes, the C- value for these species is about 0.17 and 0.37, respectively (<http://www.genomesize.com>). It has been proposed that heterochromatin protein 1 (HP1), in association with the origin recognition complex, recruits underphosphorylated isoforms of HP1 to sites of heterochromatin nucleation [@pone.0017512-Shareef1]. High cav-related protein levels could be advantageous in species with high heterochromatin content such as *D. virilis*. The functional significance of having in *D. persimilis*/*D. pseudoobscura* a *cav* gene duplicate is also unclear. Even more puzzling is the functional significance of having one *cav* gene duplicate in *D. willistoni* with an apparent male-specific expression, since the original *cav* gene is expressed in both females and males. It should be noted that the C-value of these species is similar to the one reported for *D. melanogaster* (<http://www.genomesize.com>). Detailed expression studies are needed in order to address this issue. The Mre11 protein is involved in telomere maintenance by preventing telomere fusion [@pone.0017512-Bi1], [@pone.0017512-Gao2]. In *D. mojavensis* there are two *mre11*-like genes. *mre11* is expressed both in males and females being, however, more highly expressed in males. The *mre11-dup* gene seems to be expressed in males only. Therefore, in principle, the effect of the gene duplication is to exacerbate even more the difference in Mre11 expression levels in males and females. It can be speculated that *D. mojavensis* telomeres are for some reason stickier than those of other species. This is a possibility because the telomeric and half-telomeric retrotransposons of *D. mojavensis* display a number of unique features when compared to other *Drosophila* species [@pone.0017512-Villasante1]. In *D. melanogaster*, as in any eukaryote, recombination-based mechanisms also help maintain chromosome termini [@pone.0017512-Kahn1]. Nevertheless, in *Drosophila* males, there is no recombination, and thus the higher *mre11* expression levels in males than in females might have been anticipated. Two functional *polo* gene duplicates are observed in *D. persimilis*/*D. pseudoobscura*. *polo-dup1* and *polo-dup2* are apparently exclusively expressed in males. It can thus be predicted that in the *obscura* group of *Drosophila* nuclear envelope breakdown and entry into prometaphase occurs earlier in males from these species when compared with what happens in *D. melanogaster*. The *D. melanogaster* Mtrm protein is a meiosis-specific 1∶1 stoichiometric inhibitor of the Polo kinase protein. Two independent duplications of this gene were found, one in *D. willistoni* and the other in *D. virilis*. The *D. willistoni mtrm-dup* gene seems to be a recent pseudogene, whereas strong evidence is here presented supporting the fact that the *D. virilis mtrm-dup* is an old functional gene duplication. It is unlikely that *mtrm-dup* is a meiotic drive element that was duplicated just by chance. It can thus be predicted that in *D. virilis* nuclear envelope breakdown and entry into prometaphase occurs later than in *D. melanogaster*. It should be noted that the *D. virilis mtrm-dup* is expressed in females only. There are functional gene duplicates of *meiS332* in *D. mojavensis* and *D. virilis*. If there is more MeiS332 protein to be removed from centromeres by Polo, then meiosis would be delayed, since removal of MeiS332 from centromeres is a step required for proper chromosome segregation at the metaphase II/anaphase II transition. Interestingly, in *D. virilis* females the *mtrm* gene is also duplicated. As noted above, an increase in Mtrm protein levels is also predicted to result in a delay in meiosis. A delayed meiosis could result in more time available to deal with large genomes such as that of *D. virilis*. It is, however, unclear whether the high heterochromatin content found in *D. virilis* is the consequence of an historically advantageous long meiosis duration that allowed the accumulation of high amounts of heterochromatin without deleterious consequences, or whether the long meiosis duration is an adaptive response aiming at handling the large amount of heterochromatin found in this species, that may have accumulated due to other reasons. In conclusion, in this work we find that, contrary to theoretical expectations, meiosis-related genes are duplicated and retained at the same rate as the average for all genes. The duplicated genes were, *a priori*, unlikely to be found duplicated, and may represent examples of neofunctionalization. Detailed cellular and biochemical experiments must be performed in order to address this issue. Nevertheless, given the nature of the genes that were found duplicated, it is here speculated that the duplicated genes may affect meiosis duration. *D. melanogaster* is the only *Drosophila* species where meiosis duration has been recorded (it takes about 1-2 days; [@pone.0017512-Bennett1]). The results here presented suggest that in the *obscura* group of species, male meiosis duration may be shorter than in *D. melanogaster*, while in *D. virilis,* where three meiosis genes are duplicated, meiosis duration may be much longer than in *D. melanogaster*. Interestingly, *D. virilis* is among the *Drosophila* species the one with highest nuclear DNA content, and Bennett [@pone.0017512-Bennett1] has shown a linear correlation in insects between nuclear DNA content and the duration of meiosis. If the correlation derived by Bennett holds true, then, at the same temperature, meiosis should take about twice as long in *D. virilis* than in *D. melanogaster*. Environmental factors should be taken into consideration as well, when making such predictions. Indeed, Bennett [@pone.0017512-Bennett1] shows that in insects, a decrease of 10°C in environmental temperature means a doubling in meiosis duration. Therefore, under their natural environments, *Drosophila* temperate species (such as species of the *virilis* group) should show, anyway, longer meiosis duration times than tropical species (such as *D. melanogaster* African populations). Supporting Information {#s5} ====================== Table S1 ::: {.caption} ###### **Overview of the meiosis-related genes studied.** (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S2 ::: {.caption} ###### **List of primers used.** (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S3 ::: {.caption} ###### **Accession numbers for the 33 meiosis genes studied from 12** ***Drosophila*** **species.** (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S4 ::: {.caption} ###### **Coding sequence size and intron number (in brackets) of 33 meiosis genes from 12** ***Drosophila*** **species.** (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: The authors want to thank R. Scott Hawley and S. Kendall Smith for critically reading the manuscript, as well as David Liberles, Fyodor Kondrashov, and two anonymous reviewers for providing useful insights into this work. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was funded by the Fundação para a Ciência e Tecnologia (FCT) through research projects PTDC/BIA-BDE/66765/2006 and PTDC/BIA-BEC/099933/2008 and through PhD grants attributed to MR and SG, funded by Programa Operacional para a Ciência e Inovação (POCI-2010), co-funded by Fundo Europeu para o Desenvolvimento Regional (FEDER) and Programa Operacional para a Promoção da competividade (COMPETE). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: MR SG CPV CS JV. Performed the experiments: MR SG CPV CS JV. Analyzed the data: MR SG CPV CS JV. Contributed reagents/materials/analysis tools: MR SG CPV CS JV. Wrote the paper: MR SG CPV CS JV.

PubMed Central

2024-06-05T04:04:19.766175

2011-3-10

PMC3053366

Introduction {#s1} ============ Ecosystems are facing ever increasing levels of human pressures which imperil the goods and services they provide to humanity. It is now recognized that both changes in environmental conditions (e.g., global warming) and modifications to biological communities (e.g., biodiversity erosion) affect ecosystem processes [@pone.0017476-Chapin1], [@pone.0017476-Hooper1], [@pone.0017476-Smith1], the latter issue having stimulated convincing advances but also controversy [@pone.0017476-Loreau1]. During the last two decades the positive relationship between biodiversity and ecosystem functioning (BEF hereafter) has been demonstrated through experiments manipulating species composition in model assemblages [@pone.0017476-Hooper1], [@pone.0017476-Loreau1], [@pone.0017476-Cardinale1]. These studies helped to place the problems of environmental change and biodiversity loss into the mainstream political agenda [@pone.0017476-Chapin2]. However, there is an urgent need to move beyond the heuristic objective of early biodiversity experiments, and then to disentangle the contributions of the various components of biodiversity on ecosystem processes and, ultimately, to build a predictive framework for BEF research that can forecast the potential effects of biodiversity changes that all ecosystems on earth are experiencing. To reach this objective, at least two limitations remain. First, biodiversity has been recognized as a multidimensional concept [@pone.0017476-Purvis1], [@pone.0017476-Devictor1] but many BEF studies rely solely on species richness for practical reasons and remain silent on the functional structure of communities. Yet, the functional trait composition of biological communities is a key component that most often explains ecosystem functioning better than species richness *per se* whatever the biota [@pone.0017476-Hooper1], [@pone.0017476-Danovaro1], a functional trait being any morphological, physiological or phenological feature, measurable at the individual level, that determines species effects on ecosystem properties [@pone.0017476-Violle1]. The limited predictive power of BEF research, even if biodiversity effects were demonstrated to be positive and significant [@pone.0017476-Balvanera1], [@pone.0017476-Tilman1], [@pone.0017476-Hector1], is certainly due to the clear initial focus on testing diversity effects (mostly on the species richness level) irrespective of other compositional factors, such as species or functional identity, and the resultant lack of an integrative framework where different components of biodiversity were considered altogether as predictor variables. Second, the vast majority of BEF studies have focused on a single ecosystem process (e.g. productivity) while overall ecosystem functioning is sustained by several processes [@pone.0017476-Reiss1]. Recent results suggest that the effect of biodiversity in natural ecosystems may be much larger than currently thought if we consider a multiple-processes framework [@pone.0017476-Hector2], . Taxonomic diversity, functional identity and functional diversity of ecological communities are each known to influence ecosystem processes but their relative effects remain largely untested, particularly in predicting rates of multiple ecosystem processes [@pone.0017476-Hooper2], [@pone.0017476-Diaz1], [@pone.0017476-Mokany1]. Species richness was the first biodiversity component to be related to ecosystem functioning [@pone.0017476-Tilman2] supporting the hypothesis that species complementarity and sampling effects both enhance resource use and productivity. Then, species evenness, or how equitably abundance is distributed among species within a community, was demonstrated to positively influence productivity [@pone.0017476-Wilsey1]. The functionally orientated BEF research began as early as 1941 [@pone.0017476-Jenny1] with the study of the effect of particular species functional traits (functional identity) on ecosystem processes (soil formation). Then, other authors have pointed out that ecosystem properties should primarily depend on the identity of dominant species and their functional traits following the 'mass ratio hypothesis' [@pone.0017476-Grime1], [@pone.0017476-Hillebrand1]. Indeed, functional identity, usually expressed as the biomass-weighted mean trait value for a community, has been demonstrated to be a key driver of ecosystem functioning from local [@pone.0017476-Mokany1] to regional [@pone.0017476-Garnier1] and global [@pone.0017476-Cornwell1] scales. Beyond functional identity, functional diversity, defined as the diversity and abundance distribution of traits within a community [@pone.0017476-Mason1], has been shown to be an accurate predictor of ecosystem functioning [@pone.0017476-Tilman3], [@pone.0017476-Hooper3], [@pone.0017476-SchererLorenzen1], reinforcing the importance of niche complementarity for enhancing ecosystem processes [@pone.0017476-Fargione1]. All these biodiversity components are not mutually exclusive but are unlikely to exert equal influence on ecosystem processes and on the multifunctionality of ecosystems. Thus the question is no longer whether each of the three components of biodiversity (taxonomic diversity, functional identity and functional diversity) matters but whether it still matters after removing the effect of two other components? In other words, we examined the additional effect of each biodiversity component on the prediction of ecosystem processes to determine whether each component has an essential and complementary contribution to the explanation of ecosystem multifunctionality. Further, by including the eight biodiversity indices, embracing all aspects of taxonomic and functional structure of communities, we built a minimum adequate model that reached an unprecedented level of explanatory power with functional identity and functional diversity together as predictor variables of multiple ecosystem processes. Finally, we implemented structural equation models to explore both causal and spurious associations between predictors of ecosystem processes. We used data on several ecosystem processes including biomass production and decomposition trials within the German BIODEPTH experiment (BIODiversity and Ecological Processes in Terrestrial Herbaceous Ecosystems) to predict the effects of biodiversity change on ecosystem functioning. This experiment allows testing all components of biodiversity given that, for each species richness level, different species combinations were constructed. Materials and Methods {#s2} ===================== Experiment {#s2a} ---------- Data have been collected at the German site of the pan-European BIODEPTH project [@pone.0017476-SchererLorenzen2]. A gradient of plant species richness and number of functional groups (grasses, legumes, non-leguminous herbs) was created by sowing mixtures containing 1, 2, 4, 8 or 16 species, typically found in mid-European hay meadows (total species pool was 31 species). Total seed density was 2,000 viable seeds per m^2^, equally divided among all species following a substitutive replacement series design. Each diversity level was replicated with several mixtures differing in species composition. The whole experiment was replicated in a second block with new randomization of plots, yielding a total of 60 plots of 2×2 m in size. Unsown seedlings were continuously weeded, and the plots were not fertilized. Among the 60 plots, we retained only 26 since we had to select only those with species richness ranging from 4 to 16 species to be able to estimate indices of functional community structure. Indeed, there is no functional volume (functional richness index) with 1 or 2 species. As an alternative, we should use the Functional Diversity (FD) index [@pone.0017476-Petchey1], based on a dendrogram, to cope with species poor communities (less than 3 species) and thus use the entire range of species richness available in BIODEPTH. However, the building of functional dendrograms is contentious [@pone.0017476-Mouchet1] and we cannot estimate the other functional diversity components (those including species abundances) with this approach. Ecosystem processes {#s2b} ------------------- Among the ecological processes that were measured at the German BIODEPTH site we selected those that were relatively independent (mean correlation over all selected processes was 0.5) since two highly correlated processes would be trivially ruled by the same biodiversity components. The response variables were cotton decomposition in 1997 and 1998, litter decomposition in 1998, productivity in 1997 and 1998, and nitrogen pool size in aboveground biomass 1998. Cotton decomposition trials are a standard method to test for effects of microenvironmental conditions on decomposition processes. It was measured as dry weight loss (g.g^−1^.d^−1^) of a standard cotton fabric using strips of 5×12 cm (Shirley Soil Burial Test Fabric, c. 95% cellulose; initial nitrogen concentration of 0.09%) during 10 weeks of field exposure in all experimental communities, with three strips per plot [@pone.0017476-SchererLorenzen1], [@pone.0017476-Spehn1]. Litter decomposition was the dry weight loss (g.g^−1^.d^−1^) of plot-specific senescent leaf and stem material, sealed in litter bags of 5×5 cm made of a 0.5 mm nylon mesh, during 10 weeks in autumn 1998. Macrofauna was excluded with this mesh size. Assuming an equal effect of a small mesh size in all treatments, excluding one decomposer group should not have an effect on our results. This assumption might not be true in case of a diversity effect on macrofauna occurrence. In another experiment, carried out on the same plots and half a year later, we could show, however, that several indices of soil fauna, including different groups of earthworms (litter feeding epigeics and anecics) and nematodes, were not influenced by our plant diversity treatments [@pone.0017476-Gastine1]. The litter bags were placed in a homogeneous patch of an adjacent meadow, thus quantifying the effect of community-specific litter composition and quality on decomposition processes, independent of differences in microenvironmental conditions induced by the experimental communities. Thus, both decomposition trials independently quantified different pathways of potential diversity effects on decomposition processes [@pone.0017476-SchererLorenzen1]. Productivity was the sum of two harvests per year (June and September) in 1997 and 1998, as a proxy for annual biomass production (dry weight, g.m^−2^). Standing biomass was cut at a height of 5 cm in two areas of 0.5 m×0.2 m each within a permanent quadrat placed in the center of each plot [@pone.0017476-SchererLorenzen2], [@pone.0017476-Spehn1]. Nitrogen pool size in aboveground biomass 1998 was the nitrogen content in dried aboveground biomass of the year 1998, calculated as the product of N concentration and biomass (gN.m^−2^). Nitrogen was measured by dry combustion with an automated C/N analyzer (Carlo Erba NA 1500, Carlo Erba, Mailand, Italy) [@pone.0017476-SchererLorenzen2], [@pone.0017476-Spehn1]. We also calculated a multifunctionality variable as the mean performance of communities over the four processes after standardizing each community performance (mean of 0 and standard deviation of 1) in order to give them the same weight. When the same process was measured for two years we first calculated a mean value for this process over the two years. Functional traits {#s2c} ----------------- The selected traits were growth form: caespitosa, reptantia, scandentia, semirosulata and rosulata; leaf size: nanophyllous (20--200 mm^2^), microphyllous (2--6 cm^2^), submicrophyllous (6--20 cm^2^) and mesophyllous (20--100 cm^2^); leaf seasonality: summergreen, partly evergreen and evergreen, CN ratio of plant litter [@pone.0017476-SchererLorenzen1]; SLA based on measurements in another biodiversity experiment [@pone.0017476-Heisse1]; and leaf angle: predominantly vertical leaf orientation, predominantly inclined leaf orientation and predominantly horizontal leaf orientation [@pone.0017476-Heisse1]. See [Table S1](#pone.0017476.s001){ref-type="supplementary-material"} for details by species. Indices for community structure {#s2d} ------------------------------- We considered two independent variables related to taxonomic composition: species richness and the evenness of abundance distribution among species using the Pielou index [@pone.0017476-Legendre1]. Since we have both quantitative and qualitative traits, we performed a Principal Coordinate Analysis (PCoA) on a Gower distance matrix to provide three independent axes that summarize species distribution within a trait functional space [@pone.0017476-Legendre1]. The functional structure of each community was assessed within this 3-dimensional PCoA space which represents more than 90% of the total inertia. These three independent functional axes from PCoA were used to measure functional identity through biomass-weighted mean trait values for each community. Three independent variables were related to functional diversity [@pone.0017476-Villeger1] ([Figure 1](#pone-0017476-g001){ref-type="fig"}). Functional richness was measured as the amount of functional space filled by the community which is the volume inside the convex hull that contains all trait combinations represented in the community, which basically corresponds to a multivariate functional range [@pone.0017476-Villeger1], [@pone.0017476-Cornwell2]. Functional evenness was estimated as the regularity of abundance distribution in the multidimensional functional space, i.e. the regularity with which species abundances fill the functional space. Finally, functional divergence quantified whether higher abundances are close to the volume borders, i.e. whether specialist species *sensu* Elton [@pone.0017476-Devictor2] have the highest abundances. See [Text S1](#pone.0017476.s003){ref-type="supplementary-material"} for details on functional diversity indices. ::: {#pone-0017476-g001 .fig} 10.1371/journal.pone.0017476.g001 Figure 1 ::: {.caption} ###### Geometrical presentation of functional diversity indices. For simplicity, only two traits are considered to define a two-dimensional functional space. For the 6 panels, a local community of 10 species (dark disks) is considered among a regional pool of 25 species (grey crosses). Species are plotted in this space according to their respective trait values while the circle areas are proportional to their abundances. Functional diversity of a community is thus the distribution of species and of their abundances in this functional space. Functional richness is the functional space occupied by the community, functional evenness is the regularity in the distribution of species abundances in the functional space and functional divergence quantifies how species abundances diverge from the centre of the functional space. For each component of functional diversity, two contrasting communities are represented, the right column showing an increase of the index value. More details on indices can be found in [Text S1](#pone.0017476.s003){ref-type="supplementary-material"}. :::  ::: Statistical analyses {#s2e} -------------------- In order to disentangle the relative effect of each biodiversity component on ecosystem processes, several alternative nested models were tested. We used the generalized likelihood ratio test [@pone.0017476-Burnham1] to determine whether each biodiversity component has a significant additional contribution to the explanation of ecosystem processes. Then the parsimony of each model was assessed using the AICc criteria given the ratio between the number of observations (26) and the number of variables (8) [@pone.0017476-Burnham1]. In order to prioritize the biodiversity indices related to ecosystem processes, and to investigate their effects (coefficients), we followed a multiple regression approach. Starting with a full model including all 8 indices, the relative importance of indices was assessed using a backward selection procedure. The significance of the increase in deviance resulting from the deletion of a variable in the model was estimated using the chi-squared deletion test [@pone.0017476-Chatterjee1]. The minimal adequate model was selected as the one containing nothing but significant variables. For each response variable (ecosystem process), we performed multiple regressions and we then selected the minimal adequate model. We did not rely on classical AIC, BIC or AICc criteria to select the most parsimonious model, i.e. the one offering an optimal trade-off between increased information (number of explicative variables) and decreased reliability (goodness-of-fit), since the number of potential models with 8 predictors vastly exceeds the number of observations [@pone.0017476-Chatterjee1]. This may lead to spurious model selection results [@pone.0017476-Burnham1]. To correctly estimate the influence of each biodiversity index on ecosystem processes we need to rely on independent biodiversity predictors, since the inherent collinearity among explanatory variables has blurred many statistical and inferential interpretations in ecology [@pone.0017476-Graham1]. This potential multicollinearity among predictive variables was tested using the variance inflation factor (VIF) [@pone.0017476-Fox1]. However, even if VIF values are lower than 10, we may still obtain significant biases in parameter estimates and low statistical power, potentially impairing the identification of significant effects and invalidating approaches assuming no collinearity among predictor variables [@pone.0017476-Graham1]. To examine the role of co-varying factors, we constructed and applied structural equation models (SEMs) for each ecosystem process. This allows direct and indirect effects of the variables of interest to be teased apart and has already applied in BEF research [@pone.0017476-Mokany1]. On one hand, taxonomic diversity or species composition may have significant effects on ecosystem processes but they should be driven by relationships with functional community structure [@pone.0017476-Loreau1]. On the other hand, taxonomic diversity is not expected to be perfectly correlated with functional structure [@pone.0017476-Gastine1]. SEMs allow us to test simultaneously how well functional structure accounts for any effects of taxonomic diversity on EF, and how strongly taxonomic diversity influences functional structure. This will ultimately provide a causal framework linking taxonomic diversity and EF via functional community structure. All statistical analyses were carried out using R software and packages 'qpcR', 'car' and 'lavaan'. Results {#s3} ======= Contribution of each biodiversity component {#s3a} ------------------------------------------- First we ran four linear models for each ecosystem process: the full model including taxonomic diversity (TD), functional identity (FI) and functional diversity (FD) with 8 biodiversity indices (TD+FI+FD), the model without any taxonomic component (FI+FD) where species richness and evenness were removed, the model without any functional identity component (TD+FD) where the biomass-weighted mean trait values were removed, and the model without any functional diversity component (TD+FI) where the 3 functional diversity indices were removed. The most parsimonious model, according to the AICc criteria, was the model without any taxonomic component for all processes but litter decomposition ([Table 1](#pone-0017476-t001){ref-type="table"}). This FI+FD model also provided the highest adjusted *R^2^* values whatever the process except litter decomposition and productivity in 1997 ([Table 1](#pone-0017476-t001){ref-type="table"}). ::: {#pone-0017476-t001 .table-wrap} 10.1371/journal.pone.0017476.t001 Table 1 ::: {.caption} ###### Summary of model comparisons for each ecosystem process as well as multifunctionality. ::: {#pone-0017476-t001-1} Process Model *df* *AICc* *R^2^* *p* Test *L.Ratio* *p* -------------------- ---------- ------ ------------ ----------- --------- -------------------- ----------- ----------- Cottondecom 97 TD+FI+FD 16 −216.9 0.25 0.114 FI+FD 18 **−227.0** **0.313** 0.041 TD+FI+FD vs. FI+FD 0.243 0.787 TD+FD 19 −226.9 0.225 0.076 TD+FI+FD vs. TD+FD 1.205 0.340 TD+FI 19 −222.1 0.059 0.306 TD+FI+FD vs. TD+FI 2.612 **0.087** Cottondecom 98 TD+FI+FD 17 −229.8 0.29 0.073 FI+FD 19 **−239.5** **0.354** 0.022 TD+FI+FD vs. FI+FD 0.149 0.863 TD+FD 20 −238.8 0.256 0.049 TD+FI+FD vs. TD+FD 1.320 0.301 TD+FI 20 −228.2 0.119 0.794 TD+FI+FD vs. TD+FI 4.838 **0.013** Litterdecom TD+FI+FD 17 −291.9 **0.648** \<0.001 FI+FD 19 −295.8 0.599 \<0.001 TD+FI+FD vs. FI+FD 2.317 0.129 TD+FD 20 −297.0 0.572 \<0.001 TD+FI+FD vs. TD+FD 2.436 0.100 TD+FI 20 **−298.1** 0.589 \<0.001 TD+FI+FD vs. TD+FI 2.124 0.135 Productivity 97 TD+FI+FD 17 368.1 **0.794** \<0.001 FI+FD 19 **361.2** 0.791 \<0.001 TD+FI+FD vs. FI+FD 1.139 0.344 TD+FD 20 367.6 0.701 \<0.001 TD+FI+FD vs. TD+FD 4.030 **0.025** TD+FI 20 362.9 0.751 \<0.001 TD+FI+FD vs. TD+FI 2.418 0.102 Productivity 98 TD+FI+FD 17 367.0 0.713 \<0.001 FI+FD 19 **358.4** **0.725** \<0.001 TD+FI+FD vs. FI+FD 0.594 0.563 TD+FD 20 358.6 0.695 \<0.001 TD+FI+FD vs. TD+FD 1.413 0.273 TD+FI 20 367.5 0.567 \<0.001 TD+FI+FD vs. TD+FI 4.381 **0.019** Npool bm 98 TD+FI+FD 17 131.3 0.823 \<0.001 FI+FD 19 **122.2** **0.834** \<0.001 TD+FI+FD vs. FI+FD 0.363 0.701 TD+FD 20 127.8 0.77 \<0.001 TD+FI+FD vs. TD+FD 2.980 **0.061** TD+FI 20 134.9 0.698 \<0.001 TD+FI+FD vs. TD+FI 5.689 **0.007** Multifunctionality TD+FI+FD 17 51.3 0.751 \<0.001 FI+FD 19 **42.8** **0.762** \<0.001 TD+FI+FD vs. FI+FD 0.552 0.586 TD+FD 20 47.7 0.679 \<0.001 TD+FI+FD vs. TD+FD 2.916 **0.064** TD+FI 20 52.1 0.62 \<0.001 TD+FI+FD vs. TD+FI 4.495 **0.017** The weight of support for the alternative models (TD: taxonomic diversity, FI: functional identity, FD: functional diversity) and estimates of model parameters for each ecosystem process (Cottondecomp: cotton decomposition, Litterdecom 98: litter decomposition in 1998, Productivity: productivity as annual biomass production, Npool bm: nitrogen pool size in aboveground biomass, Multifunctionality: mean performance over all processes). Results of likelihood ratio tests comparing nested models (*L.Ratio*) and associated p-values. Adjusted *R^2^s* for the ordinary least squares regression models and p-value associated to the multiple regressions are presented. The lowest AICc value for each process, the highest adjusted *R^2^* and the significant differences between models (*p*\<0.1) are in bold. ::: Then, we examined whether each of the three biodiversity components added a significant contribution to the explanation of ecosystem processes using generalized likelihood-ratio tests comparing nested models. The taxonomic component (richness and evenness) never made an additional contribution to the explanation of ecosystem processes since the FI+FD model was not significantly outperformed by any full model (TD+FI+FD) with all 8 indices ([Table 1](#pone-0017476-t001){ref-type="table"}). Conversely, functional identity and functional diversity added a significant contribution for, respectively, 3 and 5 processes. We found that all variance inflation factors were lower than the critical heuristic value of 10 suggesting that collinearity among explanatory variables did not strongly affect our results (see [Table S2](#pone.0017476.s002){ref-type="supplementary-material"} for values by predictor). Selection of the minimal adequate model and its explanatory power {#s3b} ----------------------------------------------------------------- For each ecosystem process, we performed a multiple regression including the 8 indices as predictive variables with a backward procedure to select the minimal adequate model ([Table 2](#pone-0017476-t002){ref-type="table"}). Biodiversity indices explained significantly, albeit weakly, cotton decomposition (*R* ^2^ = 0.34--0.42) but only functional aspects of community structure were retained in the minimal adequate model, with functional divergence having the main effect (positive) over the two years. For litter decomposition, 69% of the variation was explained by community structure with a combination of three indices: species evenness, functional identity on the second axis and functional divergence, the latter having a positive influence. More interestingly, up to 82% of the variation in productivity was explained by community structure with consistent effects of functional divergence and functional identity (first and second axis) over the two years. Similarly, the functional structure of communities explained nitrogen pool size at 84%, with a predominant positive effect of functional divergence, while species richness was not retained in the final model. Finally, 80% of the level of multifunctionality was explained by only three variables: functional identity (first and third PCoA axes) and functional divergence, with functional divergence having the greatest influence (positive). In other words, the aggregated mean position of the community within functional trait space in combination with functional divergence accurately predicts the level of ecosystem multifunctionality. [Figure 2a](#pone-0017476-g002){ref-type="fig"} shows the influence of position in functional space for multifunctionality, with communities having higher values than −0.1 on the first PCoA axis also have higher levels of multifunctionality than the others while communities with low values on both the first and third PCoA axes have a low average multifunctionality values. In addition, all communities with high functional divergence values (\>0.85) show high multifunctionality levels ([Figure 2b](#pone-0017476-g002){ref-type="fig"}). ::: {#pone-0017476-g002 .fig} 10.1371/journal.pone.0017476.g002 Figure 2 ::: {.caption} ###### Relationships between community structure and ecosystem multifunctionality. \(A) Multifunctionality performance of each community in the functional trait space (first and third axes of the PCoA -- PCoA 1 and PCoA 3 respectively). (A) Multifunctionality performance against functional divergence (FDiv). Circle sizes are proportional to performance of communities. See [Table 1](#pone-0017476-t001){ref-type="table"} for associated statistics. :::  ::: ::: {#pone-0017476-t002 .table-wrap} 10.1371/journal.pone.0017476.t002 Table 2 ::: {.caption} ###### Summary of the minimal adequate models. ::: {#pone-0017476-t002-2} S E PC1 PC2 PC3 FRic FEve FDiv *R* ^2^ *p* -------------------- ---------------------------------------- ------------------------------------------- ------------------------------------------ ---------------------------------------- ------------------------------------------- ------------------------------------------ ------------------------------------------- ------------------------------------------ --------- ---------- Cottondecom 97 −2.0\* −2.1[\*\*](#nt104){ref-type="table-fn"} −1.9\* 3.6[\*\*\*](#nt105){ref-type="table-fn"} 0.34 0.0140 Cottondecom 98 −2.3[\*\*](#nt104){ref-type="table-fn"} −4.1[\*\*\*](#nt105){ref-type="table-fn"} 2.5[\*\*](#nt104){ref-type="table-fn"} 0.42 0.0016 Litterdecom homo −3.1[\*\*\*](#nt105){ref-type="table-fn"} 2.4[\*\*](#nt104){ref-type="table-fn"} −1.7\* 2.9[\*\*\*](#nt105){ref-type="table-fn"} 0.69 \<0.0001 Productivity 97 2.3[\*\*](#nt104){ref-type="table-fn"} 3.8[\*\*\*](#nt105){ref-type="table-fn"} −2.8[\*\*\*](#nt105){ref-type="table-fn"} 2.7[\*\*](#nt104){ref-type="table-fn"} 0.82 \<0.0001 Productivity 98 1.8\* −2.2[\*\*](#nt104){ref-type="table-fn"} 3.0[\*\*\*](#nt105){ref-type="table-fn"} 2.4[\*\*](#nt104){ref-type="table-fn"} 0.75 \<0.0001 Npool bm 98 3.1[\*\*\*](#nt105){ref-type="table-fn"} −2.8[\*\*](#nt104){ref-type="table-fn"} 2.4[\*\*](#nt104){ref-type="table-fn"} 3.4[\*\*\*](#nt105){ref-type="table-fn"} 0.84 \<0.0001 Multifunctionality 3.0[\*\*\*](#nt105){ref-type="table-fn"} −3.5[\*\*\*](#nt105){ref-type="table-fn"} 4.2[\*\*\*](#nt105){ref-type="table-fn"} 0.80 \<0.0001 Results of regressions of ecosystem processes (Cottondecomp: cotton decomposition, Litterdecom 98: litter decomposition in 1998, Productivity: productivity as annual biomass production, Npool bm: nitrogen pool size in aboveground biomass, Multifunctionality: mean performance over all processes) against 8 biodiversity indices (S: species richness, E: species evenness, PC1 PC2 and PC3: aggregated mean trait values along three PCoA axes, FRic: functional richness, FEve: functional evenness, FDiv: functional divergence). *t*-value for each selected variable, adjusted *R* ^2^s for the ordinary least squares regression models and *p*-value associated to the multiple regressions are presented. Explanatory variables (biodiversity indices) were selected using a backward selection procedure starting with a maximal model towards the one containing nothing but significant terms (*p*\<0.1). *p*\<0.1, \*\**p*\<0.05, \*\*\**p*\<0.01. ::: [Figure 3](#pone-0017476-g003){ref-type="fig"} shows two communities containing the same number of species (8) with extreme values along the gradient of multifunctionality level (community *a*\>community *b*). In the high functioning community *a*, all the dominant species are specialists (*i.e.* with extreme combinations of traits), which contributes to a high functional divergence value. Community *a* also has a higher mean value on the first PCoA axis of all communities (indicated by the black triangle in [Figure 3a](#pone-0017476-g003){ref-type="fig"}). Conversely, the low functioning community *b* has a lower functional divergence value with some dominant species being generalists (i.e. close to the center of the functional space occupied by the community) that are functionally redundant ([Figure 3b](#pone-0017476-g003){ref-type="fig"}). This community has also a lower mean value on the first PCoA axis. ::: {#pone-0017476-g003 .fig} 10.1371/journal.pone.0017476.g003 Figure 3 ::: {.caption} ###### Two species communities represented in functional space with contrasting multifunctionality levels. Two 8-species communities of our experiment with the highest multifunctionality level (a) and the lowest (b). Positions of species are presented in the functional space (first and third PCoA axes). The black triangle labeled "Agg" represents the biomass-weighted mean trait values (aggregated trait) along the two PCoA axes while the lines represent the functional volume occupied by each community. The sizes of grey circles are proportional to species relative abundances. Full species names and trait values can be found in [Table S1](#pone.0017476.s001){ref-type="supplementary-material"}. :::  ::: Structural Equation Model {#s3c} ------------------------- Using a structural equation model (SEM) for ecosystem multifunctionality (models for other processes are provided in [Text S2](#pone.0017476.s004){ref-type="supplementary-material"}), we confirm that taxonomic composition of communities had no direct significant influence on ecosystem multifunctionality ([Figure 4](#pone-0017476-g004){ref-type="fig"}); only functional identity (through first and third PCoA axes) and functional divergence had a significant direct effect with functional divergence having the greatest influence (positive). Taxonomic diversity did have a significant influence on the functional structure of communities, but the greatest effect was the positive influence of species richness on functional richness, which had no significant effect on multifunctionality. Functional indices were weakly related between each other and only two correlations were significant and positive (functional divergence and first PCoA axis, functional richness and second PCoA axis). The SEM illustrates that despite the co-linearity between the first PCoA axis and functional divergence, both indices had significant independent effects on multifunctionality. ::: {#pone-0017476-g004 .fig} 10.1371/journal.pone.0017476.g004 Figure 4 ::: {.caption} ###### Results of the structural equation model (SEM) linking the multifunctionality of ecosystems to biodiversity indices. (S: species richness, E: evenness in species abundances, PC1 PC2 and PC3: aggregated mean trait values along three PCoA axes, FRic: functional richness, FEve: functional evenness, FDiv: functional divergence.) Numbers next to unidirectional arrows are standardized slopes and those next to bidirectional arrows are correlations. Only significant effects or correlations are shown (*\* p*\<0.1, *\*\* p*\<0.05, *\*\*\* p*\<0.01). For detailed statistics and for each process, see [Text S2](#pone.0017476.s004){ref-type="supplementary-material"}. :::  ::: Discussion {#s4} ========== Our results demonstrate that biodiversity components differ greatly in their influence on ecosystem processes. The taxonomic component, after removing the effects of functional identity and diversity, has no additional effect on processes with consistently low and non significant likelihood-ratio values ([Table 1](#pone-0017476-t001){ref-type="table"}). In addition, species richness and evenness were rarely retained in the minimal adequate model ([Table 2](#pone-0017476-t002){ref-type="table"}) or by the SEM ([Figure 4](#pone-0017476-g004){ref-type="fig"}) for their direct influence on ecosystem processes ([Text S2](#pone.0017476.s004){ref-type="supplementary-material"}). This result can be partly explained by the positive relationship between functional richness and species richness ([Figure 4](#pone-0017476-g004){ref-type="fig"}) [@pone.0017476-Villeger1] since communities with more species are more likely to hold a higher diversity of traits and thus perform more functions [@pone.0017476-Halpern1]. Therefore, the additional effect of species richness is likely to be weak after removing the effect of functional richness. Similarly, species evenness has no significant influence on ecosystem processes (except litter decomposition) but it influences the functional structure of communities as revealed by the SEM analysis ([Figure 4](#pone-0017476-g004){ref-type="fig"}). We conclude that while the influence of taxonomic structure on ecosystem processes is less important than that of functional identity and diversity, taxonomic composition mediates functional structure. This implies that the taxonomic composition of communities may have indirect effects on ecosystem processes since they are not their proximate, but partly their ultimate, drivers. While it remains difficult to provide a definite mechanistic explanation for the relationship between functional structure and multifunctionality, existing literature and our own observations may provide some clues. Two of the key functional traits for explaining multifunctionality were leaf phenology (evergreen vs. partly evergreen vs. summergreen) and leaf inclination. There is only very little evidence that increased phenological complementarity can have a positive effect on annual productivity in early successional forb communities, although such an effect might be stronger at low levels of species richness [@pone.0017476-Stevens1]. Evergreen species at our site might have some photosynthetic activity during mild winter days, but biomass production is very low until the onset of spring. Some of the evergreen or partly evergreen species, however, shown an early onset of growth in spring with an early peak in the season (e.g. *Alopecurus pratensis*, *Plantago lanceolata*), while the summergreen species have a tendency to peak later in the year (e.g. *Centaurea jacea*, *Geranium pratense*). Thus, this temporal complementarity of growth might have induced higher productivity with higher functional divergence in leaf phenology. Variability in leaf inclination is known to enhance the photosynthetic light capture of individual tree crowns (e.g. [@pone.0017476-Posada1]), while in a study of mixed red clover (*Trifolium pratense*) and tall fescue (*Festuca arundinacea*) canopy differences in leaf inclination between the two species increase equality in light partitioning between the taller fescue and shorter clover [@pone.0017476-Sonohat1]. In our model plant community, the influence of functional divergence on productivity may be due to temporal and spatial partitioning in light capture via complementarity in phenology and leaf inclination, respectively. In a previous study on the same site, it has been shown that increasing functional diversity positively influences decomposition rates of plant litter, while species richness had no such effect [@pone.0017476-SchererLorenzen1]. These results suggest that this positive effect of functional diversity was due to improved microenvironmental conditions for decomposer fauna, and due to higher litter quality. With reference to functional identity, increasing dominance of species with more horizontal leaf inclination might enhance productivity by increasing total light capture relative to communities dominated by species with vertical inclination, which might partially explain the influence of PC1 on multifunctionality. Supporting this mechanism, communities dominated by forb species with horizontal leaf inclination also had higher leaf area index than those dominated by grasses with a more vertical inclination. In addition, all nitrogen-fixing legumes planted show a horizontal leaf inclination, partly confounding leaf inclination with N-fixation, the latter being known to positively influence productivity at our site [@pone.0017476-Spehn2]. However, it is unclear how aggregate mean phenology would affect multifunctionality. Perhaps summergreen species are able to grow faster since decreased leaf longevity is associated with increased photosynthetic rates [@pone.0017476-Reich1]. Short lived leaves also have traits associated with more rapid decomposition rates (e.g. high nutrient content, [@pone.0017476-Cornwell1]), which would explain the influence of PC3 on litter decomposition in the minimal adequate regression. The higher nutrient content of summergreen leaves is supported by the negative relationship between PC3 and the amount of nitrogen in biomass in the minimal adequate regression. The predominance of variables linked to the functional structure of communities over taxonomic variables in predicting ecosystem processes is in accordance with the most recent findings obtained in experiments [@pone.0017476-Mokany1] or with empirical data [@pone.0017476-Danovaro1]. Except for decomposition, we show that functional identity and diversity bring independent and additional explanatory power to ecosystem processes with consistently high likelihood-ratio values. Overall, the results suggest that neither functional identity nor functional divergence was more important than the other in explaining ecosystem processes and particularly the multifunctionality. So, contrary to other studies, demonstrating the higher contribution of one component over the others [@pone.0017476-Mokany1] [@pone.0017476-Arenas1], we demonstrate that this differential contribution may depend on the process involved, and when considering multiple processes the magnitude of the two component effects is similar. Thus, to reach high levels of predictability in modelling multiple ecosystem processes, functional identity and diversity components have to be taken into account in a common framework [@pone.0017476-Schumacher1]. It has been suggested that, since different species often influence different functions, the level of biodiversity needed to sustain multifunctionality in ecosystems is higher than previously thought [@pone.0017476-Hector2] [@pone.0017476-Gamfeldt1]. By integrating across four ecosystem processes in assessing the level of community multifunctionality, we show that both functional divergence and functional identity have a predominant role, while species richness has no direct effects ([Table 2](#pone-0017476-t002){ref-type="table"}) and few indirect effects ([Figure 4](#pone-0017476-g004){ref-type="fig"}). We suggest that this absence of a species richness effect is partly explained by the relatively high richness values considered in our study (4 to 16 species) while past evidence for positive effects of species richness on ecosystem processes have often been due to the weak performances of monocultures or very species poor communities [@pone.0017476-Hooper1]. Indeed our results are not in contradiction with previous studies demonstrating positive species diversity effects on ecosystem functioning. Rather, they suggest that, except at the extreme low end of species richness gradients, the taxonomic structure of ecological communities is no longer the main driver of ecosystem processes, with the functional structure being the primary determinant. Our study reconciles two hypotheses that have been alternatively suggested to primarily underpin ecosystem processes: the complementarity and the mass ratio hypotheses. We suggest that a combined effect of functional identity and functional divergence is the most parsimonious explanation for key ecosystem processes. Taken separately, each biodiversity component has weak explanatory power for ecosystem functioning [@pone.0017476-Mokany1] [@pone.0017476-SchererLorenzen1] [@pone.0017476-Spehn1]. However, the combined effect of biodiversity components related to the functional structure of communities used in our study consistently reached unprecedented levels of predictive accuracy (up to 84%) whatever the process and for all processes together. Our finding is crucial since recent work has demonstrated that global gradients in decomposition rates, for example, are primarily driven by plant functional traits rather than climate [@pone.0017476-Cornwell1], emphasizing the need for better understanding of the interplay between functional structure of communities and ecosystem functioning. The predominance of functional divergence effects on most of ecosystem processes sheds light on the need to preserve specialist species *sensu* Elton (i.e. those that have a particular combination of traits and perform particular functions in the system). However, since under the combined influence of habitat degradation or global change, we are increasingly losing local specialist species [@pone.0017476-Devictor3] [@pone.0017476-Villeger2], the level of functional diversity held by communities is declining worldwide [@pone.0017476-Flynn1]. Our results show that modifying the functional structure of communities has a strong impact on ecosystem processes and should receive more attention in assessing and countering the global decline of biodiversity. Supporting Information {#s5} ====================== Table S1 ::: {.caption} ###### Species used in the German BIODEPTH experiment with their traits. (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S2 ::: {.caption} ###### Values of the variance inflation factor (VIF) for each biodiversity index (S: species richness, E: species evenness, PC1 PC2 and PC3: aggregated mean trait values along three PCoA axes, FRic: functional richness, FEve: functional evenness, FDiv: functional divergence). (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Text S1 ::: {.caption} ###### Calculation of functional diversity indices. (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Text S2 ::: {.caption} ###### Results from Structural Equation Models (SEM) for each process. (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: We are grateful to those colleagues that commented on earlier versions of this manuscript, including Michel Loreau, Andy Hector and Eric Garnier. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**The BIODEPTH project ran from 1996 to 1998 and was funded by the European Commission within the Framework IV Environment and Climate programme (ENV-CT95-0008). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: MSL. Performed the experiments: MSL. Analyzed the data: DM SV NWHM MSL. Contributed reagents/materials/analysis tools: DM SV NWHM MSL. Wrote the manuscript: DM SV NWHM MSL.

PubMed Central

2024-06-05T04:04:19.771419

2011-3-10

PMC3053367

Introduction {#s1} ============ Patients without a spleen or with diminished splenic function are at risk of severe infections such as post-splenectomy sepsis (PSS), which carries a mortality of up to 70% [@pone.0017302-Holdsworth1], [@pone.0017302-Styrt1]. Patients can be protected from these infections if preventive measures such as immunizations and use of antibiotics are taken [@pone.0017302-Gaston1], [@pone.0017302-Halasa1]. As PSS is preventable, several relevant organizations have developed guidelines with recommendations for the management of patients without a spleen [@pone.0017302-Mourtzoukou1]. The British Committee for Standards in Haematology (BCSH) has developed a guideline for the management of patients after splenectomy in 1996 and updated it in 2002 [@pone.0017302-Working1], [@pone.0017302-Davies1]. The recommendations such as made by the BCSH (see [table 1](#pone-0017302-t001){ref-type="table"}) are currently thought to reflect 'best-practice' and physicians should adhere at least to similar practice. In the Netherlands, hospital specialists as well as general practitioners (GPs) are involved in the care for asplenic patients. During hospital admission, before and after splenectomy, specialists of Internal medicine and Surgery are responsible for immunizing patients and provide prophylactic antibiotics. After discharge from the hospital, patients are transferred to their GPs for further post-splenectomy management to prevent infections. ::: {#pone-0017302-t001 .table-wrap} 10.1371/journal.pone.0017302.t001 Table 1 ::: {.caption} ###### Key recommendations for the management of asplenic patients by the British Committee for Standards in Haematology. ::: {#pone-0017302-t001-1} Key recommendations -------------- ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Immunization Splenectomised patients should receive pneumococcal immunization (23-valent polysaccharide vaccine, PPV-23) and lifelong revaccination. They should also receive *Haemophilus influenzae* type B (Hib) and meningococcal C vaccine. Yearly influenza immunization is recommended. Antibiotics Continuous prophylactic antibiotics are recommended for the first two years after splenectomy. In case of suspected or proven infection during or after these 2 years, patients should immediately use antibiotics and be admitted to a hospital. Education All patients should be educated about the risks of infection associated with traveling (such as infection with *Plasmodium falciparum*) and unusual infections (i.e. dog bites). Patient records should be labeled to indicate the risk of infection as well. ::: Research on the quality of care for patients without a spleen has frequently reported low rates of physician guideline adherence [@pone.0017302-Bruni1]--[@pone.0017302-Ramachandra1], for instance in the United Kingdom, Scotland, Denmark, Spain and Canada. In the Netherlands, it was recently shown that management of patients after splenectomy is unsatisfactory as well [@pone.0017302-Lammers1], [@pone.0017302-MeerveldEggink1], and patients are therefore at risk of serious infection. A better understanding of the reasons for non-compliance with guidelines is needed to improve the care for asplenic individuals. Although low guideline adherence has been reported, to our knowledge potential reasons for this nonadherence have never been investigated. It is increasingly recognized that to improve guideline adherence, implementation should be preceded by the assessment of barriers [@pone.0017302-Grol1], [@pone.0017302-Grol2]. A wide range of potential barriers to guideline adherence has been identified operating at different levels, such as the level of the physician, of the patient, the organizational context and the social and cultural context [@pone.0017302-Grol1], [@pone.0017302-Cabana1]--[@pone.0017302-Maue1]. The aim of this study was to investigate barriers to comply with the best practice recommendations for asplenic individuals as made by the BCSH. We assessed reasons for nonadherence to recommendations for asplenic patients among general practitioners as well as hospital specialists and we identified the most important barriers that need to be addressed to improve adherence to best practice recommendations. Methods {#s2} ======= We conducted a cross-sectional survey using written questionnaires, in a sample of Dutch general practitioners (GPs), and specialists of Internal medicine and Surgery. The survey was preceded by focus groups discussions to identify potential physicians\' experienced barriers to adherence with best-practice (see [Table 1](#pone-0017302-t001){ref-type="table"}) in the management of patients after splenectomy, as defined in the BCSH guideline [@pone.0017302-Working1], [@pone.0017302-Davies1]. Since no human subjects were used and questionnaires were filled out by physicians anonymously, this study did not require an ethics approval, hence we did not contact the ethics committee of our institute. Questionnaire development {#s2a} ------------------------- Conform state-of-the-art questionnaire development [@pone.0017302-Barbour1], we used focus-group discussions to gain insight into potential existing barriers to guideline adherence for GPs and medical specialists. Semi-structured focus group discussions were planned separately for GPs and hospital specialists (of Internal Medicine and Surgery). GPs were contacted by telephone, hospital specialists by email. Participants of different age, sex, years in practice and practice setting were selected. We used the framework of Cabana and colleagues [@pone.0017302-Cabana1] to classify potential barriers to adherence to best practice recommendations into three main categories: barriers related to physicians\' knowledge (i.e. lack of awareness and familiarity with the guideline), physicians\' attitudes (i.e. lack of agreement, outcome expectancy or motivation) and external barriers (i.e. patient-, organization-, and guideline-related factors). These potential barriers were used to develop a topic guide. This topic guide, with open-ended questions, was then used to structure and moderate the discussion for each guideline recommendation. Each focus group discussion lasted two hours. Sessions were tape-recorded and fully transcribed. Two researchers (KL, JL) conducted an independent analysis of the transcript contents. Differences in interpretation of the transcripts were minimal and consensus was promptly achieved. Subsequently, the experienced barriers to guideline adherence that were identified during the two focus group discussions, were used in the development of two distinct questionnaires; one for GPs and one for hospital specialists. The first part of the questionnaire was developed to determine current practice among the participating cohort of physicians. Each item had a 4-point Likert-type response: 'always', 'frequently (in more than 50% of cases)', 'sometimes (in less than 50% of cases)' or 'never'. The second part of the questionnaire was designed to investigate experienced barriers to best-practice management, as outlined in the recommendations by the BCSH [@pone.0017302-Working1], [@pone.0017302-Davies1]. For these items the 5-point Likert-type response was: 1 (strongly disagree), 2 (disagree), 3 (agree nor disagree), 4 (agree), and 5 (strongly agree). Suggestions for improving guideline adherence and respondents\' demographics were requested in the third part of the questionnaire using closed questions. The questionnaires were pilot-tested before mailing them to the study sample. To obtain adequate response rates, we applied methodologies suggested in the literatures [@pone.0017302-Edwards1], [@pone.0017302-VanGeest1], including factors such as length of questionnaire, traditional mail survey (as opposed to internet-based survey), providing return envelopes and follow-up mailing with replacement questionnaires. Study sample {#s2b} ------------ The online database of the Academic Medical Center (AMC, Amsterdam) contains addresses of all Dutch GPs (ADB-ICT/Cluster Software Engineering, 2006 ADICT-AB, Amsterdam) and was used to randomly select a sample of GPs and obtain their contact details. Contact details of the sample of hospital specialists (Internists as well as Surgeons) were randomly selected from the "Geneeskundig adresboek Nederland", where all Dutch specialists are registered. A total of one hundred GPs and two hundred hospital specialists (one hundred Internists and one hundred Surgeons) were invited to participate in the study. Questionnaires were sent with an accompanying informative letter and return envelope to all selected physicians. Two weeks after the initial mailing, a reminder card was sent to non-responders with the request to complete the form. Physicians received a second copy of the questionnaire if the first was not returned within 4 weeks. Data analysis {#s2c} ------------- The results from all returned questionnaires were entered in a database. Data were analyzed with the Statistical Program for the Social Sciences (SPSS 16.0 for Windows®, SPSS Inc., Chicago, Illinois, USA). For analysis of demographic data, descriptive statistics were obtained. For the analysis of current practice, we categorized the answers 'always' and 'frequently (in more than 50% of cases)' as "positive", and the answers 'never' and 'sometimes (in less than 50% of cases)' were categorized as "negative". For analysis of statements about barriers given on the five-point scale, the answers were dichotomized to enable division between "yes" (barrier experienced) and "no" (barrier not experienced), by ranking *strongly agree* (5) and *agree* (4) as "yes" and *strongly disagree* (1) and *disagree* (2) as "no". All p-values were computed by Chi-square test for three groups of physicians or types of work-setting by GraphPad (GraphPad Prism, version 4.00 for Windows, GraphPad Software, San Diego California USA). Results {#s3} ======= Study population {#s3a} ---------------- In total, 47 GPs and 73 hospital specialists participated in the study, yielding response rates of 47% and 36,5% respectively. One hundred and twenty questionnaires were suitable for analysis. Six questionnaires were excluded from analysis because they were returned without completion due to: no asplenic patients in practice (3), not engaged in care for asplenic patients (1), list too long (1), retired (1). The demographic characteristics of participating physicians are presented in [Table 2](#pone-0017302-t002){ref-type="table"}. ::: {#pone-0017302-t002 .table-wrap} 10.1371/journal.pone.0017302.t002 Table 2 ::: {.caption} ###### Characteristics of specialists (of Internal medicine and Surgery) and general practitioners (GPs) participating in the questionnaire survey. ::: {#pone-0017302-t002-2} Internists Surgeons GPs ------------------------------------------------------------------------------------------------- ------------ ---------- ------ Number of participating physicians (N) 42 31 47 Mean age (years) 47 51 50 Gender (% male) 71 93 67 Mean years since registration as Specialist (MSRC registration)[a](#nt101){ref-type="table-fn"} 14,9 18,3 Work setting (%): University hospital 14,3 26,7 Non-university Teaching hospital 64,3 33,3 General Non-teaching hospital 21,4 40,0 Solo practice 30,4 Group practice[b](#nt102){ref-type="table-fn"} 63,0 Health center 6,5 Mean number of patients serviced by GP practice 2891 a MSRC = Medical Specialists Registration Committee. b Group practice includes 'duo-practices' and 'HOED-practices' (Huisartsen Onder Één Dak; a number of GPs working independently in the same building). ::: Current practice {#s3b} ---------------- Results of current practice are shown in [Table 3](#pone-0017302-t003){ref-type="table"}. GPs and Specialists vaccinated their patients against pneumococci in 82,6% and 94,4% of cases respectively. Immunizations against *H. influenzae* B (Hib) and meningococci were given less frequently. The recommendation to take antibiotics immediately in case of fever was given in 90,5% of cases by Internists, in 60% of cases by Surgeons and 66% of cases by GPs. Continuous antibiotics for the first two years after splenectomy were prescribed in a minority (less than 15%) of patients by all physicians. These results indicate that current practice in the prevention of infections in asplenic individuals is not optimal. ::: {#pone-0017302-t003 .table-wrap} 10.1371/journal.pone.0017302.t003 Table 3 ::: {.caption} ###### Current practice of asplenic patients\' management as reported by specialists of Internal medicine and Surgery, as well as general practitioners. ::: {#pone-0017302-t003-3} Percentage of physicians reporting to provide asplenic patients with: Internists (%) Surgeons (%) GPs (%) P value[a](#nt104){ref-type="table-fn"} ----------------------------------------------------------------------- ---------------- -------------- --------- ----------------------------------------- Pneumococcal immunization 95,2 93,3 82,6 0,1123 *H. influenzae* B immunization 88,1 50,0 45,7 \<0.0001 Meningococcal C immunization 81,0 56,7 30,4 \<0.0001 Lifelong boosters of Pneumovax[b](#nt105){ref-type="table-fn"} 83,3 36,7 66,0 0,0002 Annual flu immunization 73,8 26,7 91,3 \<0.0001 Continuous antibiotics for 2 years after splenectomy 9,5 13,3 6,4 0,5885 On-demand antibiotics 88,1 66,7 78,7 0,0887 Advice to take antibiotics immediately in case of fever 90,5 60,0 66,0 0,0123 Advice to gather information upon travelling 78,1 40,0 70,2 0,0026 Immediate antibiotic therapy after cat or dog bites 61,9 40,0 66,0 0,0648 Percentages indicate the number of physicians that answered with either 'always' or 'frequently' (in more than 50% of cases) when asked if they provided their asplenic patients with the recommended preventive measures. a P value calculated by Chi-square test, for 3 groups of physicians. b Pneumovax ®: 23-valent conjugate pneumococcal vaccination. ::: Comparison of different type of hospitals (university-, non-university teaching-, and general non-teaching hospitals) yielded no significant correlation between performance of hospital specialists and hospital teaching status (data not shown). Differences were minimal between types of GP-practices (solo practice, group practice and health centers) as well, although physicians working in solo-practices vaccinated their patients significantly less frequent against pneumococci (61,5%) and Hib (23,1%) as compared to GPs working in group practices (89,7% and 51,7% respectively) and health centers (100% for both vaccines) (data not shown). Knowledge related barriers {#s3c} -------------------------- Physicians\' knowledge of the recommendations for asplenic patients is shown in [Table 4](#pone-0017302-t004){ref-type="table"}. Although not always familiar with the recommendations, less than 25% of hospital specialists and GPs indicated that they lacked sufficient awareness of the need for preventive measures in asplenic patients. ::: {#pone-0017302-t004 .table-wrap} 10.1371/journal.pone.0017302.t004 Table 4 ::: {.caption} ###### Knowledge related barriers to best practice for asplenic patients. ::: {#pone-0017302-t004-4} Barriers Internists(% agree) Surgeons(% agree) GPs(% agree) ------------------------------------------------------------------------------------------------ --------------------- ------------------- -------------- **Familiarity** I am not familiar with the existence of recommended immunizations 4,9 33,3 37 I am not familiar with the existence of recommended 'prophylactic' and 'on-demand' antibiotics 55 70 71,7 I am not familiar with the existence of recommended precautions 29,3 58,6 55,3 **Awareness** I am not aware of the need for immunizations in asplenic patients 0 13,3 17,4 I am not aware of the need for antibiotics in asplenic patients 7,5 23,2 26,1 Percentages indicate the number of physicians that experienced the barrier, by answering either 'strongly agree' or 'agree'. ::: Attitude related barriers {#s3d} ------------------------- Physicians\' attitudes towards the recommendations for asplenic patients are shown in [Table 5](#pone-0017302-t005){ref-type="table"}. Physicians generally agreed with guideline contents (65% of specialists, 73% of GPs) and evidence (58% of specialists, 68% of GPs). ::: {#pone-0017302-t005 .table-wrap} 10.1371/journal.pone.0017302.t005 Table 5 ::: {.caption} ###### Barriers experienced by specialists of Internal medicine and Surgery, as well general practitioners. ::: {#pone-0017302-t005-5} Internists (%)[a](#nt107){ref-type="table-fn"} Surgeons (%)[a](#nt107){ref-type="table-fn"} General practitioners (%)[a](#nt107){ref-type="table-fn"} ---------------------------------------------------------------------------------- ------------------------------------------------ ---------------------------------------------- ----------------------------------------------------------- ------ ------ ------ ------ ------- ------ **Attitude-related barrier** I do not agree with the guideline contents 7,1 10,5 4,9 0 20 3,3 2,2 4,3 6,4 Recommendations are not evidence-based 9,5 15,4 12,2 0 13,3 0 2,2 4,3 8,5 Recommendation is time consuming 14,3 2,6 \- 0 3,3 \- 4,3 4,3 4,3 Patients\' comorbidity 42,9 \- \- 17,2 \- \- \- \- \- Long-term use of antibiotics is a patient burden \- \- \- \- \- \- \- 43,5 \- **External factors** Physicians\' responsibilities are not clarified 53,5 44,7 51,2 43,3 46,7 48,4 63 60,9 40,4 The specialty registrar[c](#nt109){ref-type="table-fn"} is not aware of the need 61,9 56,4 51,2 50 53,3 48,3 55,3 54,3 51,1 The patient is not informed about the need 83,3 81,6 \- 90 90 \- 80,9 89,1 \- Patient is resistant to receive the measure 16,7 20,5 \- 13,3 20 \- 23,4 28,3 \- The GP does not comply with my suggestion 33,3 59 58,5 40 41,4 41,4 \- \- \- The specialists\' instructions are incorrect \- \- \- \- \- \- 32,6 45,7 27,7 The specialists\' instructions in the discharge letter are incomplete \- \- \- \- \- \- 46,8 52, 2 51,1 Different hospitals recommend different policies \- \- \- \- \- 31,9 45,7 36,2 Lack of reimbursement for NeisVac-C vaccin \- \- \- \- \- \- 25,5 \- \- a percentages indicate the number of physicians that either "strongly agree" or "agree" with the proposed barrier. b prevention: give advice to patient when travelling and prompt treatment of unusual infections. c specialty registrar = in training for Medical or Surgical consultant, general practitioner in training. ::: External barriers {#s3e} ----------------- Several barriers related to patient- and organizational levels were experienced ([Table 5](#pone-0017302-t005){ref-type="table"}). According to respondents, adherence to the recommendations would improve if patients themselves would be better informed (over 80% of hospital specialists and GPs). Physicians reported barriers on organizational level to be important as well: clarity about which physician is responsible for the management of asplenic patients was lacking (50% of Internists, 46% of Surgeons, 55% of GPs), and physicians were not able to rely on their residents in the management of asplenic patients (over 50% of hospital specialists and GPs). Moreover, hospital specialists were uncertain if the GP would follow their advice given on patient discharge (33--59%), whereas GPs were not convinced that the specialists provided them with a discharge letter containing the correct recommendations (47--52%). Differences in these perceived barriers between specialists working in different hospital types were minimal. Medical specialists working in university hospitals were most confident in their residents (41,2%, as opposed to 14,1% of Specialists in non-university teaching hospitals, and 11,1% in non-teaching hospitals). Surgeons working in general non-teaching hospitals especially reported lack of clarity on transmural responsibilities as a barrier (48,5%, compared to 17,2% of surgeons in non-university teaching hospitals, and 8,3% in non-teaching hospitals) (data not shown). Lack of clarity on responsibilities as well as incorrect discharge letters were reported most frequently by GPs working in health centers (both 44.4%), as compared to GPs working in GPs working in group practices (19,5% and 14,9% respectively) or solo GPs (10% and 5%). Solo GPs reported the lowest confidence in residents (2,4%, as compared to 10,3% of GPs working in group practices and 33,3% of GPs working in health centers) (data not shown). Improving adherence to recommendations {#s3f} -------------------------------------- All physicians indicated that the availability of a comprehensible national guideline containing all recommendations would improve adherence to best practice (93% of Internists, 94% of Surgeons and 97% of GPs, data not shown). Respondents scored the predefined potential improvements for (better) compliance with best practice recommendations. Results are shown in [Figure 1](#pone-0017302-g001){ref-type="fig"}. Of all given suggestions, the respondents rated the online availability of a guideline as most useful. Other suggestions for improvement, identified during the focus group discussions, that were regarded as useful by most respondents were 'a patient brochure' (69,8%) and 'transmural agreements about responsibilities' (63,3%). ::: {#pone-0017302-g001 .fig} 10.1371/journal.pone.0017302.g001 Figure 1 ::: {.caption} ###### Potential improvements in the adherence to recommendations for asplenic patients. Percentage of general practitioners (GPs), Specialists of Internal medicine and Surgery that indicated the suggestion as potentially useful. :::  ::: Discussion {#s4} ========== Main findings {#s4a} ------------- Current management of asplenic patients in the Netherlands is not in compliance with best practice standards. This study identified, for the first time, the barriers which inhibit physicians\' adherence to recommendations for asplenic patients. The study revealed that in the Netherlands, problems on an organizational level and poorly informed patients most likely explain non-adherence of physicians to the best practice recommendations. Explaining the results {#s4b} ---------------------- This study used physicians\' self-reporting to measure current practice of post-splenectomy management of patients. The reported findings are in correspondence with our previous study [@pone.0017302-Lammers1], where comparable vaccination- and antibiotic prescription rates were found. Internationally, publications report comparable numbers as well [@pone.0017302-Deodhar1], [@pone.0017302-Kyaw1], [@pone.0017302-Ramachandra1], [@pone.0017302-Waghorn1]. Maybe the most alarming finding is that GPs as well as Surgeons fail to recommend well over one third of all asplenic patients to take antibiotics immediately in case of infection, which is the most important measure to prevent lethal infections in asplenic patients. For improvement purposes we identified physicians\' barriers to best-practice after splenectomy which we classified using Cabana\'s framework [@pone.0017302-Cabana1]. In this study we found that poor adherence to the British guideline recommendations is probably not due to lack of knowledge or negative attitude towards the guidelines. Physicians\' knowledge of the guideline recommendations for asplenic patients was generally adequate. Although not always familiar with the recommendations, less than a quarter of hospital specialists and GPs indicated that awareness was a barrier to adherence. This is in correspondence with reported figures from Canada, were physicians\' knowledge of management of asplenic patients was also found to be relatively satisfactory [@pone.0017302-Brigden1]. In our study, general practitioners estimated to serve a mean of 3,3 asplenic patients per 2500 patients in their practice (data not shown). Therefore, although prevalence is low, it is encouraging to see that knowledge of the infectious risks of these patients is fair. Further, physician\'s attitudes towards the recommendations for asplenic patients were generally positive. Physicians reported to have confidence in the guideline contents. In addition, despite the fact that many BCSH recommendations were formulated based on expert opinion (low level of evidence), physicians indicated they found the evidence to be sufficient. Appropriate knowledge and attitudes of physicians are necessary but not sufficient for adherence to guidelines [@pone.0017302-Cabana1], [@pone.0017302-Solberg1]. A physician may still encounter barriers that limit his/her ability to perform the recommended behavior due to so called external factors, that is patient related, guideline related or organizational factors. Indeed, we found that Dutch physicians experienced external factors to be important barriers to adherence. Organizational factors, such as lack of clarity on the responsibilities in care delivery for asplenic patients were reported to be most important. In the Netherlands, care for asplenic patients is a joint responsibility of specialists (during hospital stay) and general practitioners (after the patient has been discharged from the hospital or outpatient clinic). Both groups of physicians indicate that clarity about the division of transmural tasks most likely would improve care for asplenic patients. There is also a lack of mutual trust: hospital specialists are uncertain whether GPs will follow their recommendations after discharge, whereas at the same time the GP is not confident that the referring consultant has provided a discharge letter containing complete and correct recommendations. Clearly, here is room for improvement, when both groups of physicians reach consensus on the responsibilities of implementation of the recommendations for asplenic patients. In addition, both GPs and hospital specialists indicated that better informed patients may contribute to improved quality of care. Although patient education alone is not sufficient to prevent post-splenectomy sepsis, it may be an important factor in preventing infections during asplenia. El-Alfy *et al.* studied 318 patients after splenectomy [@pone.0017302-ElAlfy1], and found that 45% was well informed about the risk and prevention of infection, 30% had fair knowledge and 25% had poor knowledge. Patients displaying greatest knowledge had significantly lower prevalence of PSS as compared to those with poor knowledge (1,4% versus 16.5%, p\<0,001). Limitations {#s4c} ----------- Physicians lacking knowledge or affinity regarding asplenia might have been more reluctant to participate in the study, thereby inducing a positive bias. Furthermore, response rates were modest, 47% of GPs and 37% of hospital specialists. This could induce a sampling bias as well. Lastly, although we did find that there was strong agreement amongst the different groups of physicians on perceived barriers; low responses could also negatively influence accuracy of the results. This survey however had a solid research design as both qualitative and quantitative methods were used. The qualitative approach enabled optimal exploration of reasons for nonadherence, after which the identified barriers were quantified in the cross-sectional study. Overall, our results have clear implications for initiatives to improve physician adherence in order to optimize care for asplenic patients. Implications for practice and research {#s4d} -------------------------------------- This survey provides an overview of the range of barriers that prevent physician adherence to post-splenectomy guidelines. Issues that should be addressed according to Dutch hospital specialists and GPs are: improving patient education and increase clarity on responsibilities and implementation of care for asplenic individuals. Education of health care professionals and patients regarding the risk of infection after splenectomy remains a must, perhaps more important than chemoprophylaxis, immunoprophylaxis or any specific surgical intervention. Therefore, the development of a Dutch guideline is urgently required. Conclusions {#s4e} ----------- This study showed suboptimal care delivery for asplenic patients and both identified and quantified physicians\' experienced barriers to comply with best practice recommendations for the management of post-splenectomy. Better informed patients and better transmural collaboration between GPs and hospital based Internists and Surgeons are likely to improve the quality of care of the asplenic patient population. The authors would like to thank the physicians that participated in the focus group discussions and all physicians that completed the questionnaires. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This project was supported by a ZonMw Grant form the Netherlands Organisation of Scientific Research (NWO, grant number 945-17-001). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: AJL KL PS JH. Performed the experiments: AJL KL PS JH. Analyzed the data: AJL KL. Wrote the paper: AJL.

PubMed Central

2024-06-05T04:04:19.775432

2011-3-10

PMC3053368

Introduction {#s1} ============ *Aspergillus nidulans* is a filamentous ascomycete that belongs to the subphylum of Pezizomycotina (previously known as Euascomycotina). Although only pathogenic for immunosupressed individuals, it is closely related to other Pezizomycetes of importance in the areas of medicine (e.g., *A. fumigatus*), industry (e.g., *A. niger* and *A. oryzae*) and agriculture (e.g., *A. flavus*). Because of these close relationships and its ease of manipulation in the laboratory, it has been used worldwide as a model organism for more than sixty years [@pone.0017505-Pontecorvo1]. Many basic eukaryotic developmental mechanisms have been revealed in this model organism through the application of genetic, molecular, physiological and biochemical approaches [@pone.0017505-Martinelli1]. The life cycle of *A. nidulans*, consisting of three main phases: vegetative extension and asexual and sexual reproduction, has been extensively described in the literature (see references within [@pone.0017505-Pggeler1], [@pone.0017505-Virag1]). These programs require the generation, according to sophisticated developmental pathways, of a set of specialized cell types. Vegetative cells or hyphae are tubular syncytia that grow exclusively by polarized extension [@pone.0017505-Fischer1] through the deposition of new material at the tip [@pone.0017505-Momany1]. Changes in environmental conditions, mainly the exposure to the atmosphere and light, but also nutritional and abiotic stresses, induce the generation of asexual spores called conidia [@pone.0017505-Adams1]--[@pone.0017505-Etxebeste1]. These propagules are generated and dispersed in large quantities from asexual microstructures called conidiophores. The architecture of the conidiophore involves the synthesis of five specialized cell types [@pone.0017505-Mims1]. First, the foot-cell is generated in distal cells of specific vegetative hyphae. The foot-cell acts as the base of the second cell type, the stalk, which arises and forms an apical swelling or vesicle. Thirdly, a layer of approximately 60 metulae emerges and, after their division, two phialides per metulae. This fourth cell type is the conidia producing structure [@pone.0017505-Sewall1]. Each phialide can produce a chain of more than 100 conidia. Thus, the architecture of each conidiophore allows for the production of more than 10,000 asexual spores, resulting in an efficient dispersive mechanism. The transduction of environmental cues into intracellular signals that activate the above described morphological transformations is controlled by a signaling cascade in which the transcription factor (TF) FlbB and its interaction partner FlbE play key roles. Deletion of either protein results in a distinctive phenotype characterized by the formation of cottony colonies ('fluffy' phenotype) with a broad delay in the timing of conidiation and a substantial reduction in the number of conidiophores with respect to the wild type strain [@pone.0017505-Wieser1], [@pone.0017505-Garzia1]. FlbB interacts with FlbE at the region that sustains vegetative growth, the hyphal tip [@pone.0017505-Garzia1]. FlbB is the only known TF showing such a localization in *Aspergillus nidulans* [@pone.0017505-Etxebeste2], [@pone.0017505-Etxebeste3]. The association of FlbB and FlbE is thought to form part of an environmental sensing mechanism that transduces signals to nucleus [@pone.0017505-Etxebeste2], where FlbB purportedly activates, in conjunction with additional regulators, the genetic pathway that controls the morphological changes required for conidiophore development [@pone.0017505-Garzia2]. However, further progress towards understanding the roles of FlbB requires uncovering the functional determinants encoded within its sequence. In this paper, we apply *in silico* approaches with the goal of associating particular aspects of FlbB functionality with specific motifs and residues in order to inform the design of site-directed *flbB* mutational strategies. Furthermore, analysis across multiple genomes facilitates the identification of functional determinants gained, lost or modified as species evolved independently. This opens the discussion on how far the functional phenotype of the partially characterized *A. nidulans* FlbB protein penetrates into the Pezizomycotina and which specific functions could be shared with orthologs in other species. Finally, we apply a method to discern co-conservation of motifs that could lead to the identification of the site of interaction between FlbB and FlbE. Results {#s2} ======= Initial characterization of *A. nidulans* FlbB and FlbE {#s2a} ------------------------------------------------------- *Aspergillus nidulans* FlbB is a 426 residue protein that has been the subject of several experimental studies (see references within [@pone.0017505-Garzia2]). The presence of an N-terminal basic region leucine zipper transcription factor domain (bZip) TF domain signature was detected by the National Center for Biotechnology Information (NCBI) Conserved Domain Database search (BRLZ, smart00338) from residues 75 to 126 ([Figure 1 A](#pone-0017505-g001){ref-type="fig"}, bZip). Homology to the carboxy-terminal cysteine-rich domain of the TF Yap1, which regulates the response to mild oxidative stress, [@pone.0017505-Wood1] was established by the Fugue sequence-structure homology recognition server [@pone.0017505-Shi1] in the C-terminal region of FlbB within residues 311 -- 403 ([Figure 1 A](#pone-0017505-g001){ref-type="fig"}, Yap1 C-term) with high confidence (Z-score = 10.8). These structural similarities were previously reported [@pone.0017505-Etxebeste3] although the C-terminal Yap1 similarity was based only on the presence and spacing of cysteines. Between the bZip and C-terminal structured region there is a 165 amino acid central domain with no significant similarities to described motifs or structures. ::: {#pone-0017505-g001 .fig} 10.1371/journal.pone.0017505.g001 Figure 1 ::: {.caption} ###### Sequence conservation, structure prediction and motif locations in nine Eurotiales FlbB and FlbE orthologs. The five-column moving average of entropy values for the nine FlbB (A) and nine FlbE (B) alignments (solid line) was plotted. Gaps in the alignment are indicated by the breaks in the moving average line. Regions and motifs are indicated between the plot and x-axes: structural homologies (structure), predicted order (order), predicted secondary structure (helix and sheet) and conserved areas derived from entropy data (conserved) are indicated in increasing shades of grey. The regions selected for further investigation as putative functional motifs (motifs) are indicated by labeled open boxes (B1 -- B4, E1 -- E5 and 'acidic region'). The three main domains of FlbB described in the text are indicated above the x-axis. Cysteine locations in FlbB are indicated with red diamonds and labeled with Anidu FlbB residue locations. The x-axes displays both Anidu numbering (above) and alignment numbering (below) to account for the fact that the graphed data is from gapped alignments while the regions and motifs pertain only to Anidu. The total length of the FlbE alignment was 292 residues but only the first 235 residues are included in the graph as only the Pchry extended beyond that. :::  ::: Extensive database searches with the *A. nidulans* FlbE sequence revealed neither conserved motifs, functional regions nor structural homologies. Previously, it was noted that it contained two conserved regions, a linker between them and an acidic segment in the C-terminal region [@pone.0017505-Garzia1]. We made predictions of order/disorder and secondary structure to further characterize the structural context of conserved areas. Lacking foreknowledge of the complete structural profile of a given protein, order/disorder predictions can help guide choices for mutational strategies and help to uncover regions of functional importance. Both predictions of secondary structure and order/disorder can help in this regard and agreement between the two serves to increases the likelihood that a particular region is structured. For example, conservation in regions of structure (order) could be related to either structure or other functional aspects of the protein. On the other hand, conservation in unstructured (disordered) regions is not likely to be necessary for maintaining structure but can still encode other functions such as protein-protein interactions and regulation (see below for references). FlbB and FlbE are predicted to be 36% and 43% ordered, respectively, with these regions distributed throughout each protein ([Figure 1 A and B](#pone-0017505-g001){ref-type="fig"}, order). The predominant secondary structural type predicted in FlbB was mostly helical with an extended region coinciding with the bZip domain ([Figure 1 A](#pone-0017505-g001){ref-type="fig"}, helix). A second region of predicted order that includes both predicted helix and sheet is found in the C-terminal Yap1-like domain ([Figure 1 A](#pone-0017505-g001){ref-type="fig"}, order, helix, sheet). Secondary structure prediction for FlbE largely agrees with the order prediction with both sheet and helical regions found in the areas predicted to be structured ([Figure 1 B](#pone-0017505-g001){ref-type="fig"}, helix, sheet and order). The high disordered content of the two proteins (the regions not annotated as ordered) is consistent with their purported signaling and regulatory roles as protein disorder tends to be more prevalent in these types of proteins [@pone.0017505-Iakoucheva1]. There are several attributes of disorder that contribute to its prevalence in protein-protein interactions. For example, energy of binding is reduced for disorder-mediated interactions compared to those mediated by order [@pone.0017505-Dyson1], [@pone.0017505-Oldfield1]. This facilitates the reversibility of interactions necessary for dynamic signaling. Other contributions of the disordered state to signaling functionality have been identified experimentally and theoretically [@pone.0017505-Dunker1]--[@pone.0017505-Cortese1]. Preliminary studies of FlbB orthologs in the NCBI database revealed three short conserved sequence motifs that were present only in genomes containing FlbE proteins that were highly similar to *A. nidulans* FlbE (data not shown). In this comparison, the criteria for a credible FlbE was one with an E value less than 7^-43^ and an overlap of greater than 80% with a PSI BLAST [@pone.0017505-Altschul1] profile generated from a Anidu FlbE seed sequence. This analysis compared 19 FlbBs from species with genomes that contained both proteins with 17 genomes that coded for FlbBs but no credible FlbE. In effect, the comparison was between representative sets of Euromycete and non-Euromycete Pezizomycotina. Because the presence or absence of these motifs was linked to the extent that the FlbE ortholog in that genome diverged from the *A. nidulans* FlbE sequence, we wanted to compare the pattern of sequence motif conservation of both proteins across a broad range of fungal species. Potentially, such a study could lead to the elucidation of a set of similarly conserved motifs. Such co-conservation could be due to their mutual interaction. The rationale for this approach is that interacting residues would necessarily be under evolutionary constraints due to the need to maintain compatibility with the corresponding partner motif and would therefore exhibit increased conservation [@pone.0017505-Sharon1]. For this comparison, assembly of data set of genome-paired of FlbB and FlbE proteins was undertaken. Construction of FlbB and FlbE datasets {#s2b} -------------------------------------- To begin our phylogenetic investigation, we obtained all publicly available sequences with significant similarity to FlbB and FlbE of *A. nidulans* from Pezizomycotina for which both protein sequences were available. [Table 1](#pone-0017505-t001){ref-type="table"} lists the details of the 40 genome-paired FlbB/FlbE sequences used in this study. Note that the five letter genus/species abbreviations given in the last column of [Table 1](#pone-0017505-t001){ref-type="table"} will be used from here onward when discussing individual sequences and species (e.g., Anidu for *A. nidulans*). ::: {#pone-0017505-t001 .table-wrap} 10.1371/journal.pone.0017505.t001 Table 1 ::: {.caption} ###### Sequences used in this study. ::: {#pone-0017505-t001-1} Species and strain Genomic locus/gi Protein gi/ID Protein accession Source Abbr. ------------------------------------------- ---------------------------- --------------- ------------------- -------- ------- **FlbB orthologs** Ajellomyces dermatitidis ER-3 BDCG\_08613 FGI Aderm Arthroderma benhamiae CBS 112371 ARB\_03643 FGI Abenh Aspergillus clavatus NRRL 1 121716128 XP\_001275673 NCBI Aclav Aspergillus flavus NRRL3357 AFL2G\_06507^2^ NCBI Aflav Aspergillus fumigatus Af293 302747314 ADL63138 NCBI Afumi Aspergillus nidulans 165931814 CAM35586 NCBI Anidu Aspergillus niger 145250791 XP\_001396909 NCBI Anige Aspergillus oryzae RIB40 169775489 XP\_001822212 NCBI Aoryz Aspergillus terreus NIH2624 115401688 XP\_001216432 NCBI Aterr Coccidioides immitis RS CIMG\_02371 FGI Cimmi Coccidioides posadasii Silveira CPSG\_00217 FGI Cposa Cochliobolus heterostrophus C5 81856 JGI Chete Fusarium graminearum PH1 FGSG\_01313 FGI Fgram Fusarium oxysporum FOXG\_00073 FGI Foxys Fusarium verticillioides FVEG\_01443 FGI Fvert Magnaporthe oryzae 70-15 MGG\_00342 FGI Moryz Microsporum canis CBS 113480 238844993 EEQ34655 NCBI Mcani Microsporum gypseum CBS 118893 MGYG\_05249 FGI Mgyps Mycosphaerella fijiensis CIRAD86 86007 JGI Mfiji Mycosphaerella graminicola 66896 JGI Mgram Neosartorya fischeri NRRL 181 119481803 XP\_001260930 NCBI Nfisc Neurospora crassa OR74A NCU07379 FGI Ncras Neurospora discreta FGSC 8579 166536 JGI Ndisc Neurospora tetrasperma FGSC 2508 148703 JGI Ntetr Paracoccidioides brasiliensis Pb01 PAAG\_01697 FGI Pb-01 Paracoccidioides brasiliensis Pb03 PABG\_03745 FGI Pb-03 Penicillium chrysogenum Wisconsin 54-1255 255931005 XP\_002557059 NCBI Pchry Penicillium marneffei ATCC 18224 212538897 XP\_002149604 NCBI Pmarn Pyrenophora tritici-repentis Pt-1C-BFP PTRG\_07994 FGI Ptrit Sclerotinia sclerotiorum 1980 SS1G\_08098 FGI Sscle Stagonospora nodorum SN15 SNOG\_04391 FGI Snodo Talaromyces stipitatus ATCC 10500 242820034 XP\_002487435 NCBI Tstip Thielavia terrestris strain NRRL 8126 36483 JGI Tterr Trichoderma atroviride Tatro\_contig\_27^2^ JGI Tatro Trichoderma reesei 57840 JGI Trees Trichoderma virens 15555 JGI Tvire Trichophyton rubram CBS 118892 TERG\_02140 FGI Trubr Trichophyton tonsurans CBS 112818 TESG\_03658 FGI Ttons Trichophyton verrucosum HKI 0517 NW\_003315534^2^ NCBI Tverr Verticillium albo-atrum VaMs.102 VDBG\_06054 FGI Valbo Verticillium dahliae VdLs.17 VDAG\_03214 FGI Vdahl **FlbE orthologs** Ajellomyces dermatitidis ER-3 BDCG\_00996 FGI Aderm Arthroderma benhamiae CBS 112371 ARB 00079^1^ FGI Abenh Aspergillus clavatus NRRL 1 ACLA 021460^1^ FGI Aclav Aspergillus flavus NRRL3357 AFL2G 03377^1^ FGI Aflav Aspergillus fumigatus 44889993 CAF32111 NCBI Afumi Aspergillus nidulans 227433961 ACP28868 NCBI Anidu Aspergillus niger NT\_166524^1^ NCBI Anige Aspergillus oryzae 281311977 BAI58988 NCBI Aoryz Aspergillus terreus NIH2624 NT\_165972^1^ NCBI Aterr Coccidioides immitis RS CIMG\_03365 FGI Cimmi Coccidioides posadasii Silveira CPSG\_02591 FGI Cposa Cochliobolus heterostrophus C5 CocheC5\_1 scaffold\_32^1^ JGI Chete Fusarium graminearum PH1 FGSG\_09567 FGI Fgram Fusarium oxysporum FOXG\_06268 FGI Foxys Fusarium verticillioides FVEG\_04121^1^ FGI Fvert Magnaporthe oryzae 70-15 MGG\_01731 FGI Moryz Microsporum canis CBS 113480 238837821 EEQ27483 NCBI Mcani Microsporum gypseum CBS 118893 MGYG\_01593 FGI Mgyps Mycosphaerella fijiensis CIRAD86 Mycfi2 scaffold 6^1^ JGI Mfiji Mycosphaerella graminicola Mycgr3 chr 6^1^ JGI Mgram Neosartorya fischeri NRRL 181 119495126 XP\_001264355 NCBI Nfisc Neurospora crassa OR74A NCU05255 FGI Ncras Neurospora discreta FGSC 8579 141834 JGI Ndisc Neurospora tetrasperma FGSC 2508 106945 JGI Ntetr Paracoccidioides brasiliensis Pb01 PAAG\_02176 FGI Pb-01 Paracoccidioides brasiliensis Pb03 PABG\_02411 FGI Pb-03 Penicillium chrysogenum Wisconsin 54-1255 255941552 XP\_002561545 NCBI Pchry Penicillium marneffei ATCC 18224 212534004^1^ NCBI Pmarn Pyrenophora tritici-repentis Pt-1C-BFP PTRG\_02296 FGI Ptrit Sclerotinia sclerotiorum 1980 SS1G\_09830 FGI Sscle Stagonospora nodorum SN15 SNOG\_03047 FGI Snodo Talaromyces stipitatus ATCC 10500 242779542^1^ NCBI Tstip Thielavia terrestris strain NRRL 8126 118518 JGI Tterr Trichoderma atroviride Triat2 contig 26^1^ JGI Tatro Trichoderma reesei Trire2 scaffold 11^1^ JGI Trees Trichoderma virens TriviGv29 8 2 scaff 80^1^ JGI Tvire Trichophyton rubram CBS 118892 TERG\_07460^2^ FGI Trubr Trichophyton tonsurans CBS 112818 TESG\_06544 FGI Ttons Trichophyton verrucosum HKI 0517 TRV\_00925^2^ FGI Tverr Verticillium albo-atrum VaMs.102 VDBG\_00858 FGI Valbo Verticillium dahliae VdLs.17 VDAG\_00467 FGI Vdahl Where genomic locus/gi is given, the protein sequence used was derived from genomic DNA using the following procedures: ^1^The longest continuous ORF encoded by the genomic locus, ^2^Fgenesh prediction. ::: Review of gene finding results {#s2c} ------------------------------ Following sequence selection and preliminary analyses, the protein sequences for both sets of orthologs were compared to their genomic loci to check ORF and intron calling. This was necessary because the methodologies that genome projects employ to predict protein sequences vary among organizations and individual projects. Few changes to FlbB orthologs were necessary. The cDNAs for FlbB of Anidu and Afumi were previously sequenced (NCBI accession numbers CAM35586 and ADL63138, respectively) and these formed a basis for verifying the gene finding results of the remaining sequences. Five intron locations were identified in the FlbB orthologs ([Figure S1](#pone.0017505.s001){ref-type="supplementary-material"}). The protein sequence for Aflav FlbB in the Broad Institute and NCBI data bases was significantly shorter than other orthologs. Additional genomic sequence encoding an additional 93 amino acids was obtained and translated. The resulting protein was a good match in length and sequence with the closely related Aoryz ortholog. The protein sequence was derived from Aflav AFL2G\_06507 (Contig7, 124078-125873) using Fgenesh. For FlbE, gene finding was complicated by the existence of two alternative transcripts that have been described for Anidu. NCBI protein data base entries ACP28868 and CAP08290 document intronless and intron forms, respectively, that are both conceptual translations from sequenced cDNA. Examples of alternative splicing, although rare, have been documented in Aspergilli [@pone.0017505-Maruyama1]. The relationship between the alternative forms and FlbE function has not been explored experimentally. The alternative transcripts have different sequence from residue 186 to the C-terminus. This region of Anidu FlbE is outside all conserved regions and those shown to have functional roles [@pone.0017505-Garzia1], [@pone.0017505-Kwon1]. The possibility that FlbE could be subject to alternative splicing [@pone.0017505-Irimia1] opens the question as to which form should be used in our study. The fact that the only other experimentally derived FlbE sequence, Aoryz NCBI accession BAI58988, is intronless, supports using the intronless transcript forms, at least for those species closely related to Anidu and Aoryz. Thus, all Eurotiales FlbE sequences were derived from the longest continuous ORF encoded by the respective genomic sequence. In general, this intronless nature held for the rest of the orthologs in spite of introns being predicted in 20 of the original FlbE sequences obtained from genome project and NCBI databases. The purported intron-containing FlbEs exhibited a scattered phylogenetic distribution with abundant inconsistencies in splice site locations. However, support was found for one cluster of FlbEs with a single intron in the Oxygenales (Ttons, Trubr, Tverr, Mgyps and Mcani; [Figure S2](#pone.0017505.s002){ref-type="supplementary-material"}). For the remaining sequences, support was for an intronless nature. Evidence for support of either outcome included: conserved intron locations in closely related sequences and phylogenetic consistency in terms of homology. Thus, the net effect of the changes made during this process was improvement of the quality of the alignment. Evaluation of structural homologies in full data set {#s2d} ---------------------------------------------------- To evaluate the structural similarities that were found in the experimentally characterized Anidu FlbB sequence in the remaining orthologs, each was evaluated for presence of a N-terminal bZip domain by searching the SMART database [@pone.0017505-Letunic1] and for C-terminal homology to the carboxy-terminal cysteine-rich domain of the Yap1 TF [@pone.0017505-Wood1] using the Fugue sequence-structure homology recognition server [@pone.0017505-Shi1]. The bZip signature was identified at significant E values in every FlbB sequence except Pmarn and Tstip where values were less than significant since both diverge noticeably from the consensus motif in the later part of the domain. As for the homology to the C-terminal region of Yap1, all FlbB orthologs were deemed to have homology equal to or greater than 90% confidence by Fugue except Ntetr and Ncras, which were scored at 50% confidence. These results, along with the facts that they were approximately the same length (396±31 amino acids) and exhibited substantial amounts of identity and similarity throughout the length of the aligned sequences ([Figure S1](#pone.0017505.s001){ref-type="supplementary-material"}), suggested that all 40 were orthologous. High levels of identity and similarity were also observed among the FlbE sequences ([Figure S2](#pone.0017505.s002){ref-type="supplementary-material"}). The topology of the phylogenetic tree generated from the FlbB CLUSTAL alignment did not differ substantially from previously published fungal phylogenies ([Figure 2](#pone-0017505-g002){ref-type="fig"}A)[@pone.0017505-Cornell1]--[@pone.0017505-Sharpton1], suggesting that FlbB evolved unremarkably. The topology of the FlbE tree was quite similar to the FlbB tree with no unusual deviations (data not shown). ::: {#pone-0017505-g002 .fig} 10.1371/journal.pone.0017505.g002 Figure 2 ::: {.caption} ###### Phylogenetic tree and differential features of 40 Pezizomycotina FlbB orthologs. A. Phylogenetic tree of FlbB orthologs generated from pairwise CLUSTAL distances. Four clades equivalent to the fungal classes Eurotiomycetes, Dothideomycetes, Sordariomycetes and Leomycetes are labeled. Two order level subclades within the Eurotiomycetes are also labeled: the Eurotiales and Oxygenales. The nine Eurotiales used for conserved motif discovery are starred. The eight species with putative functionally equivalent FlbB proteins are shaded in yellow. B. Differential functionally-related features identified in the sequences as described in the text. The presence of the six conserved cysteines are denoted by 'C' in the column labeled according to Anidu numbering. 'Yes' indicates that both the B2 and E2 motifs are present in that species. :::  ::: Conservation of critical bZip residues {#s2e} -------------------------------------- Having established that bZip domains were present in 40 FlbB orthologs, we next determined the level of conservation of bZip signature residues. The generalized consensus sequence for the bZip DNA binding domain (DBD) is **[N]{.underline}**\[RK\]x\[**AS**\]\[**ASQ**\]xx**\[SCFY**\]**[R]{.underline}**, with the two underlined residues being invariant and the bold residues contacting the DNA [@pone.0017505-Finn1], [@pone.0017505-Fujii1]. The two invariant residues were present in all 40 DBD-containing FlbB orthologs. Furthermore, taking all the conserved residues into account ([Figure S1](#pone.0017505.s001){ref-type="supplementary-material"}, purple highlight), the bZip DBD subfamily most closely related to FlbB orthologs is the PAP subfamily (consensus sequence [N]{.underline}xxAQxx**F** [R)]{.underline} [@pone.0017505-Fujii1]. The FlbB orthologs, however, match all but one of the five DNA-contacting residues in the PAP signature with the phenylalanine (F, in bold) being substituted with histidine (H) in all 40 orthologs giving the consensus sequence [N]{.underline}xxAQxx**H** [R]{.underline} for the FlbB DBD ([Figure S1](#pone.0017505.s001){ref-type="supplementary-material"}). The fact that all 40 DBD-containing FlbB orthologs contain this histidine suggests that it is important and specific for FlbB function. Although much effort has been expended in the study of bZip transcription factors including comprehensive reviews of the multiple subfamilies [@pone.0017505-Fujii1]--[@pone.0017505-Vinson1], this histidine-containing DBD consensus has not been described. We designated this particular DBD 'H19' in keeping with a systematic designation for describing substitutions in this position [@pone.0017505-Amoutzias2] (see discussion). As this H19 bZip DBD appeared to be novel, we conducted an exhaustive search of public and sequencing project databases to determine the extent of its distribution. The prevalence of the H19 bZip DBD was limited with one in *Pichia stipitis* (subphylum Saccharomycotina), five within fungi but outside Ascomycota, one in phylum Oomycota and one in kingdom Viridiplantae ([Table 2](#pone-0017505-t002){ref-type="table"}). This limited distribution outside Pezizomycotina further supports the hypothesis that the histidine of H19 is important and specific for FlbB function. ::: {#pone-0017505-t002 .table-wrap} 10.1371/journal.pone.0017505.t002 Table 2 ::: {.caption} ###### H19 bZip DNA binding domains found outside Pezizomycotina. ::: {#pone-0017505-t002-2} Organism Kingdom Phylum Subphylum Order Source Accession or reference ------------------------- --------------- --------------- ------------------ ------------------- -------- ------------------------------------------------------ Pichia stipitis Fungi Ascomycota Saccharomycotina Saccharomycetales NCBI gi\|150951570\|ref\|XP\_001387909.2\| Coprinopsis cinerea Fungi Basidiomycota Agaricales NCBI gi\|299746530\|ref\|XP\_001838046 Cryptococcus neoformans Fungi Basidiomycota Tremellales NCBI gi\|134110800\|ref\|XP\_775864.1\| Laccaria bicolor Fungi Basidiomycota Agaricales NCBI gi\|170105495\|ref\|XP\_001883960 Postia placenta Fungi Basidiomycota Polyporales JGI jgi\|Pospl1\|129836\|estExt\_fgenesh3\_pg.C\_1080015 Schizophyllum commune Fungi Basidiomycota Agaricales NCBI gi\|300105220\|gb\|EFI96625.1\| Phytophthora infestans Stramenopila Oomycota Peronosporales NCBI gi\|262097289\|gb\|EEY55341.1\| Volvox carteri Viridiplantae Chlorophyta Chlamydomonadales NCBI gi\|300255540\|gb\|EFJ39839.1\| ::: We have previously demonstrated that the region immediately N-terminal to the highly conserved bZip core is also important for bZip function. The flbB100 allele was identified in a random mutagenesis that sought aconidial ("fluffy") mutations [@pone.0017505-Etxebeste3]. This flbB mutant allele encodes for a FlbB protein with a change from glycine to arginine at position 70 (G70R). Coincident with the inability to produce conidiospores on Aspergillus Minimal Medium ([Figure 3](#pone-0017505-g003){ref-type="fig"}), the G70R mutation results in a remarkable decrease in the capability of the FlbB bZip to bind previously defined DNA targets compared to the wild type protein [@pone.0017505-Etxebeste2]. ::: {#pone-0017505-g003 .fig} 10.1371/journal.pone.0017505.g003 Figure 3 ::: {.caption} ###### Characterization of FlbB-C382A conidiation phenotype. Condidial phenotype of the parental wild type TN02A3 compared to mutant strains ΔflbB, flbB 100, flbB 102 and flbB-C382A. FlbB produced by flbB 100 is truncated after amino acid P305 and that of flbB 102 has a G70R substitution. :::  ::: Both homo- and hetero-dimerizaton of bZip TFs are possible with the dimerization interface composed of leucine repeats located immediately C-terminal to the DBD [@pone.0017505-Amoutzias2], [@pone.0017505-Amoutzias3]. This interface typically contains four to five heptad repeats that mediate binding between two compatible monomers via formation of a coiled-coil. By convention, the residues within each heptad are labeled 'a' through 'g' [@pone.0017505-OShea1]. A leucine zipper is formed primarily by hydrophobic residues in the 'a' and 'd' positions of each heptad. These residues form a hydrophobic interface in which each position of one monomer interacts with its counterpart in the other monomer. However, non-hydrophobic residues, which serve to modify the specificity of the interface, can also be found in these positions [@pone.0017505-Vinson1]. Contrary to the homogeneity found in the DBD, inspection of the CLUSTAL alignment of the 40 orthologs ([Figure S1](#pone.0017505.s001){ref-type="supplementary-material"}) revealed that gaps were present in the bZip dimerization domain, indicating that insertions and/or deletions had occurred in this region as the Pezizomycotina evolved. However, the first three heptads and the 'a' residue of the fourth heptad are in a gap-less region adjoining the DBD and were sufficient to differentiate two different dimerization profiles among the orthologs. Lysines or arginine in the 'a' or 'd' positions function to promote hetero- and disfavor homo-dimerization [@pone.0017505-Vinson1]. As the 'a' and 'd' residues associate with their counterparts along the hydrophobic interface, charged residues in these positions are thought to be a mechanism to disfavor homodimerization [@pone.0017505-Vinson1], [@pone.0017505-Acharya1], [@pone.0017505-Deppmann1]. We know of no studies on the implications of negatively charged residues in the 'd' position but, similar to positive charges in these positions, a similar bias against homo-dimerization would be expected. In Anidu FlbB, the 'd' position of the third heptad is glutamic acid (Glu). This negative charge modifies the hydrophobic character of the dimerization interface and likely has a strong influence on dimerization partner selection. Significantly, all of the Eurotiomycetes have a charged residue in this position except Ttons, which has a glycine ([Figure 2 B](#pone-0017505-g002){ref-type="fig"}; [Figure S1](#pone.0017505.s001){ref-type="supplementary-material"} orange highlight). The most common residue in this group is aspartic acid (Asp) with Glu occurring in eight orthologs. The evolutionary relationship between the Asp- and Glu-substituted orthologs is clearly defined with the Asp- and Glu-containing groups clustering separately ([Figure 2](#pone-0017505-g002){ref-type="fig"}). However, as the charge is conserved between the two possible substitutions, there may be no substantial differences between the dimerization profiles of the Asp- and Glu-containing orthologs. In contrast, the 19 non-Eurotiomycete orthologs lack this negative charge and instead contain a positively charged residue in the 'a' position of the fourth heptad ([Figure 2 B](#pone-0017505-g002){ref-type="fig"}; [Figure S1](#pone.0017505.s001){ref-type="supplementary-material"}, orange highlight). Similar to the conservative substitutions observed for the 'd' position of the third heptad, this charge could be either an arginine (Arg) or a lysine (Lys). This conservative substitution implies the same low likelihood of substantially altered dimerization properties. However, unlike the differential substitutions of the negatively charged residue in the third heptad, the evolutionary relationships among those containing this positive residue are not straightforward. All five Dothideomycetes contain Arg and the single Leotiomycete contains Lys. On the other hand, the Sordariomycetes are mixed, with most containing Lys but with a middle branching group composed of Fgram, Fvert and Foxys containing Arg ([Figure 2](#pone-0017505-g002){ref-type="fig"}, [Figure S1](#pone.0017505.s001){ref-type="supplementary-material"}). The possibility that the conservative Lys-\>Arg substitution arose more than once supports the supposition that no radical changes in dimerization properties are associated with these like-charged substitutions. In some bZips, a second mechanism can influence dimerization properties. Both homo- and hetero-dimers can be stabilized by salt bridge formation between oppositely charged residues in the 'g' position of one heptad in one monomer and the 'e' position of the following heptad in the other monomer [@pone.0017505-Acharya1]. On the other hand, non-complementary charges in these positions disfavor dimerization [@pone.0017505-Deppmann1]. In the dimerization domains of the 40 orthologs, a salt bridge is predicted between the 'g' residue of the first heptad and the 'e' residue of the second heptad ([Figure 2 B](#pone-0017505-g002){ref-type="fig"}; [Figure S1](#pone.0017505.s001){ref-type="supplementary-material"}, light blue highlight). The negatively charged residue of the salt bridge (in the ´ǵposition of the first heptad) is a completely conserved glutamic acid (Glu). According to their respective positions within the leucine zipper heptads, this residue is predicted to form a salt bridge with arginine (Arg) in all the Eurotiomycetes and lysine (Lys) in the Dothideomycetes, Sordariomycetes and Leotiomycetes orthologs ([Figure 2](#pone-0017505-g002){ref-type="fig"}; [Figure S1](#pone.0017505.s001){ref-type="supplementary-material"}, light blue highlight). As the positive charge is conserved with either amino acid, it is likely that the same salt bridge functionality exists in all the orthologs. C-terminal structured region of FlbB {#s2f} ------------------------------------ As discussed above, the C-terminal region of the 40 orthologs have homology to the C-terminal redox responsive domain of Yap1. PSI-BLAST, CLUSTAL, HMMalign and Fugue all align Anidu FlbB cysteine C382 with C598 of Yap1 and C501 of Pap1 (data not shown), both of which are implicated in the redox modulation of stability and localization of Yap1 [@pone.0017505-Tachibana1] and Pap1 [@pone.0017505-Castillo1]. Both proteins are bZip TFs that translocate between the cytosol and nucleus depending on the redox status of the cell. According to the CLUSTAL alignment, all 40 FlbB orthologs contain this cysteine (hereafter referred to as C382; [Figure S1](#pone.0017505.s001){ref-type="supplementary-material"}, red highlight) which, moreover, is within a group of six highly conserved residues. In both Yap1 and Pap1, the positionally equivalent cysteine participates in intramolecular disulfide bond formation in response to oxidative stress by forming bonds with either a C-terminal or a central region cysteine depending on the level of oxidative stress [@pone.0017505-Wood1], [@pone.0017505-Tachibana1], [@pone.0017505-Castillo1]. As C382 was conserved in all 40 FlbB orthologs, we sought to verify that this cysteine was critical for FlbB function in Anidu. A mutant strain expressing FlbB with cysteine 382 substituted by an alanine (C382A) was constructed. The phenotype of the FlbB C382A strain was analysed in relation to the parental wild type, an *flbB* null strain and *flbB102* (truncated at amino acid 305) [@pone.0017505-Etxebeste3] ([Figure 3](#pone-0017505-g003){ref-type="fig"}). flbB-C382A exhibited a fluffy phenotype with sparse conidiation thus confirming that C382 is critical for FlbB function in Anidu. In light of this experimental result and the homology to Yap1, C382 in the FlbB orthologs likely comprises one half of a di-sulfide bond forming pair. Hence, the question arises: Which other FlbB cysteines could form bonds with C382? Based entirely on conservation and the assumption that the FlbB orthologs do, in fact, undergo intramolecular disulfide bond formation, the most likely partner for C382 would be C272 ([Figure 2 B](#pone-0017505-g002){ref-type="fig"}, [Figure 4](#pone-0017505-g004){ref-type="fig"}, [Figure S1](#pone.0017505.s001){ref-type="supplementary-material"}). (Note that Anidu numbering will be used whenever residue positions are reported.) ::: {#pone-0017505-g004 .fig} 10.1371/journal.pone.0017505.g004 Figure 4 ::: {.caption} ###### Alignment of the nine Eurotiales FlbB orthologs used to generate HMM motif profiles. CLUSTAL alignment of the nine Eurotiales FlbB orthologs used to generate conserved motifs. Anidu FlbB is in bold. Motifs B1, B2, B3 and B4 are labeled above the alignment and highlighted in green or yellow. The heptads of the bZip dimerization domain (the first three heptads and the first residue of the third) are identified by brackets with the residue positions labeled a -- g according to convention. Cysteine residues are labeled with Anidu numbering and highlighted in light red. The positions of G70 and H93 in the bZip DBD (Anidu numbering) are also labeled. Residues flanking intron locations are in bold italic. :::  ::: The remaining cysteines exhibit varying levels of conservation with distinct patterns that segregate according to class and order. The presence or absence of the six most relevant conserved cysteines (C236, C272, C280 and C303 in the central region and C382 and C397 in the C-terminal region) are indicated for each species in [Figure 2 B](#pone-0017505-g002){ref-type="fig"}. An additional potentially relevant cysteine, C354 in Anidu numbering ([Figure 4](#pone-0017505-g004){ref-type="fig"}, [Figure S1](#pone.0017505.s001){ref-type="supplementary-material"}), is conserved in seven of the Eurotiales but was not considered here for two reasons. First, it is not found in any of the orthologs outside the Eurotiales and therefore does not contribute phylogenetic information to the study of Anidu FlbB. Second, it is not present in Anidu and therefore does not have a role in the characterized function of Anidu FlbB and thus has no impact on the transfer of experimentally derived information from that species to the other orthologs. Eight Eurotiales, including Anidu, contain all six of the most conserved cysteines ([Figure 2](#pone-0017505-g002){ref-type="fig"}; [Figure 4](#pone-0017505-g004){ref-type="fig"} and [Figure S1](#pone.0017505.s001){ref-type="supplementary-material"}, red highlight). In the other Eurotiales (Pmarn, Tstip and Pchry), five of these cysteines are completely conserved. In the Onygenales there are four completely conserved cysteines, with C236 and C397 missing in all the sequences. There are only two completely conserved cysteines in the Dothideomycetes, C272 and C382, providing additional support for this pair being redox active. All Sordariomycetes have C303, C272 and C382 except Moryz which is missing C272. Extracting conserved motifs from FlbB and FlbE {#s2g} ---------------------------------------------- In order to identify potential functional determinants and clues to potential coevolving interaction sites, we set out to identify the conserved regions of FlbB and FlbE. Although conserved regions were observed in the alignments of both sets of orthologs ([Figure S1](#pone.0017505.s001){ref-type="supplementary-material"}; [Figure S2](#pone.0017505.s002){ref-type="supplementary-material"}), we choose to map areas of conservation within a cohort of nine species most closely related to Anidu. This selection would be more likely to include conserved regions that encoded some of the experimentally characterized functionality of Anidu FlbB. These sequences were chosen based on BLAST scoring with each FlbB ortholog having greater than 60% identity and an E value less than 1^-120^ and each FlbE ortholog having greater than 50% identity and an E value less than 1^-50^. Additionally, the selected sequences were genome-paired. Not surprisingly, the nine species were members of the Eurotiales subclade ([Figure 2](#pone-0017505-g002){ref-type="fig"}, red stars). FlbB motifs were selected based on structural information and entropy analysis. To this end, entropy was calculated for each column (residue position) of the alignments of the nine Eurotiales orthologs ([Figure 4](#pone-0017505-g004){ref-type="fig"}) and a 5-residue moving average was calculated ([Figure 1](#pone-0017505-g001){ref-type="fig"}, Entropy). Initially, conserved motifs were defined as those regions with a moving average of entropy less than 0.5. The motifs were expanded as necessary to include: structural features (bZip and Yap1 C-term similarities), columns with conservative substitutions, columns with high levels of conservation that were not reflected in the moving averages and to combine short stretches of conservation into larger fragments ([Figure 1 A](#pone-0017505-g001){ref-type="fig"}, conserved and motifs). For example, B2 is comprised of a cluster of four smaller conserved regions. B3 and B4 are contiguous but we chose to terminate B3 at the beginning of B4, which was defined by structural homology to Yap1 PDB structure 1SSE as determined by Fugue. In all, four motifs (B1--B4) were identified in the FlbB alignment of the nine Eurotiales ([Figure 1 A](#pone-0017505-g001){ref-type="fig"}, [Figure 4](#pone-0017505-g004){ref-type="fig"}, [Table 3](#pone-0017505-t003){ref-type="table"}). ::: {#pone-0017505-t003 .table-wrap} 10.1371/journal.pone.0017505.t003 Table 3 ::: {.caption} ###### Length and location of regions and motifs in Anidu FlbB and FlbE. ::: {#pone-0017505-t003-3} Protein Designation Length (AA) Residues (Anidu numbering) ------------ --------------- ------------- ---------------------------- Anidu FlbB N-term 144 1--144 Central 166 145--310 C-term 116 311--426 B1 96 57--152 B2 47 157--203 B3 58 253--310 B4 95 311--405 Anidu FlbE E1 41 1--41 E2 13 45--57 E3 16 58--73 E4 12 79--90 E5 29 121--149 Acidic region 25 153--177 ::: As no conserved domains or motifs were found when the Anidu FlbE sequence was used to search conserved domain databases, motifs E1--E5 were based solely on entropy data ([Figure 1 B](#pone-0017505-g001){ref-type="fig"}, [Figure 5](#pone-0017505-g005){ref-type="fig"}, [Table 3](#pone-0017505-t003){ref-type="table"}). Additionally, FlbE orthologs were evaluated for the existence of an acidic region C-terminal to motif E5 that had been previously noted [@pone.0017505-Garzia1]. Since there was little positional conservation in this region, it was scored by calculating average charge (see [Methods](#s4){ref-type="sec"} section). An acidic region was present in all the Eurotiomycetes except Pb-01, Pb-02 and 11 other species dispersed among the Dothideomcetes, Sordariomycetes and Leotiomycetes ([Table 4](#pone-0017505-t004){ref-type="table"}). ::: {#pone-0017505-g005 .fig} 10.1371/journal.pone.0017505.g005 Figure 5 ::: {.caption} ###### Alignment of the nine Eurotiales FlbE orthologs used to generate HMM motif profiles. CLUSTAL alignment of the nine Eurotiales FlbE orthologs used to generate conserved motifs. Anidu FlbE is in bold. Motifs E1, E2, E3, E4 and E5 are highlighted in green or yellow. The acidic region is highlighted in purple. :::  ::: ::: {#pone-0017505-t004 .table-wrap} 10.1371/journal.pone.0017505.t004 Table 4 ::: {.caption} ###### HMM profile scores and other parameters for FlbB and FlbE orthologs. ::: {#pone-0017505-t004-4} FlbB FlbE ----------------- ----------- --------- ----------- ---------- ---------- ---------- ---------- --------- ----------- ---------- --------- ---------- --------- ---------- --------- **Eurotiales** **Aflav** **427** 4.2^-265^ 3.4^-62^ 3.0^-27^ 2.7^-46^ 9.6^-65^ **202** 1.3^-111^ 9.3^-35^ 1.0^-8^ 1.8^-13^ 9.6^-8^ 7.3^-25^ **YES** **Eurotiales** **Aoryz** **427** 1.4^-264^ 3.5^-62^ 2.5^-27^ 3.0^-46^ 8.9^-65^ **202** 1.3^-111^ 9.3^-35^ 1.0^-8^ 1.8^-13^ 9.6^-8^ 7.3^-25^ **YES** **Eurotiales** **Aterr** **434** 1.3^-254^ 8.7^-62^ 5.6^-28^ 4.1^-46^ 5.2^-62^ **226** 9.1^-115^ 2.3^-35^ 4.4^-8^ 2.4^-13^ 7.0^-8^ 9.5^-25^ **YES** **Eurotiales** **Anige** **420** 1.2^-252^ 6.8^-62^ 1.4^-27^ 1.2^-45^ 8.3^-64^ **203** 7.6^-111^ 3.5^-35^ 5.4^-8^ 4.0^-14^ 5.8^-8^ 4.0^-25^ **YES** **Eurotiales** **Afumi** **420** 8.4^-251^ 1.7^-60^ 9.6^-28^ 2.9^-46^ 9.2^-65^ **222** 2.2^-118^ 2.5^-35^ 2.2^-9^ 1.6^-13^ 5.8^-8^ 1.2^-24^ **YES** **Eurotiales** **Nfisc** **391** 4.1^-244^ 6.9^-61^ 7.0^-28^ 5.0^-46^ 1.1^-64^ **222** 2.4^-118^ 5.7^-35^ 2.2^-9^ 1.6^-13^ 5.7^-8^ 1.5^-24^ **YES** **Eurotiales** **Anidu** **426** 1.5^-235^ 2.6^-59^ 3.2^-25^ 2.7^-45^ 2.1^-61^ **201** 2.2^-104^ 6.4^-35^ 2.2^-8^ 1.6^-13^ 9.4^-8^ 9.4^-24^ **YES** **Eurotiales** **Aclav** **397** 7.8^-241^ 1.5^-60^ 3.5^-23^ 6.0^-46^ 1.6^-64^ **205** 5.4^-115^ 1.2^-34^ 2.3^-9^ 3.2^-14^ 4.9^-8^ 9.7^-25^ **YES** **Eurotiales** **Pchry** **410** 8.0^-230^ 3.8^-57^ 9.0^-25^ 9.1^-46^ 1.3^-60^ **280** 1.7^-104^ 4.5^-34^ 1.2^-7^ 8.2^-13^ 9.4^-8^ 1.3^-21^ **YES** Onygenales Cposa 434 2.3^-143^ 3.6^-47^ 6.7^-25^ 3.0^-44^ 187 1.3^-65^ 4.2^-17^ 1.2^-9^ 6.6^-5^ 6.6^-17^ YES Onygenales Cimmi 434 2.5^-143^ 3.6^-47^ 6.6^-25^ 3.0^-44^ 187 1.7^-65^ 4.8^-17^ 1.2^-9^ 6.5^-5^ 6.6^-17^ YES Eurotiales Pmarn 431 1.0^-136^ 1.6^-43^ 7.4^-26^ 8.4^-48^ 208 3.8^-63^ 4.5^-20^ 6.0^-8^ 2.2^-4^ 1.4^-17^ YES Eurotiales Tstip 421 6.5^-143^ 1.9^-45^ 2.2^-3^ 1.3^-27^ 9.1^-47^ 217 1.0^-65^ 6.2^-21^ 5.8^-8^ 1.3^-4^ 9.5^-19^ YES Onygenales Mcani 417 2.0^-130^ 7.2^-44^ 1.6^-21^ 4.9^-41^ 208 1.1^-54^ 2.7^-14^ 1.5^-9^ 2.4^-17^ YES Onygenales Pb-01 408 4.6^-136^ 6.5^-46^ 2.0^-23^ 1.8^-45^ 180 3.1^-57^ 9.6^-16^ 9.9^-9^ 4.5^-8^ 1.2^−11^ Onygenales Pb-03 410 1.7^−134^ 2.4^−45^ 2.0^−23^ 1.9^−45^ 180 1.6^−58^ 1.7^−15^ 3.5^−10^ 4.5^−8^ 1.2^−11^ Onygenales Abenh 384 1.1^−127^ 2.5^−44^ 2.1^−19^ 3.3^−42^ 222 3.9^−52^ 2.0^−14^ 1.5^−9^ 1.4^−16^ YES Onygenales Aderm 424 1.5^−131^ 3.9^−42^ 3.8^−21^ 1.2^−47^ 201 1.6^−59^ 1.4^−16^ 1.6^−10^ 3.7^−8^ 2.5^−12^ YES Onygenales Tverr 409 5.0^−129^ 2.6^−44^ 8.1^−20^ 3.7^−42^ 208 2.8^−51^ 1.8^−14^ 9.1^−7^ 2.3^−15^ YES Onygenales Ttons 429 1.6^−128^ 1.0^−44^ 1.8^−19^ 1.9^−41^ 207 2.1^−52^ 1.8^−14^ 1.3^−9^ 3.6^−15^ YES Onygenales Mgyps 446 6.7^−133^ 1.6^−46^ 2.2^−20^ 1.8^−42^ 208 5.2^−52^ 1.9^−14^ 2.6^−8^ 3.0^−16^ YES Onygenales Trubr 429 5.8^−129^ 2.5^−44^ 3.3^−20^ 2.4^−42^ 208 9.0^−52^ 1.6^−14^ 1.3^−9^ 2.5^−14^ YES Dothideomycetes Chete 354 2.5^−86^ 1.4^−34^ 5.1^−15^ 2.6^−27^ 237 1.3^−43^ 1.5^−14^ 1.2^−10^ 2.8^−14^ Dothideomycetes Ptrit 389 3.3^−85^ 1.3^−34^ 3.2^−15^ 1.9^−26^ 238 1.7^−49^ 4.8^−14^ 7.3^−12^ 3.3^−15^ Dothideomycetes Snodo 405 2.4^−84^ 6.7^−36^ 3.6^−15^ 2.9^−26^ 242 2.0^−46^ 2.1^−14^ 3.3^−12^ 4.6^−15^ Dothideomycetes Mgram 380 4.1^−73^ 3.3^−34^ 3.8^−13^ 1.6^−20^ 213 1.3^−47^ 2.7^−16^ 1.5^−10^ 1.1^−4^ 2.4^−8^ YES Sordariomycetes Tvire 358 2.3^−73^ 2.4^−33^ 5.7^−13^ 1.6^−23^ 269 9.2^−28^ 7.3^−9^ 1.7^−7^ YES Sordariomycetes Ndisc 422 2.0^−69^ 3.8^−32^ 1.0^−14^ 4.0^−22^ 1014 5.5^−20^ 8.1^−9^ 2.2^−4^ 2.1^−8^ YES Sordariomycetes Tterr 427 6.0^−72^ 1.8^−33^ 2.2^−11^ 8.0^−21^ 524 5.2^−21^ 6.6^−11^ 8.6^−5^ 3.9^−8^ Sordariomycetes Vdahl 382 1.1^−67^ 1.6^−34^ 8.^9−8^ 2.5^−23^ 383 2.0^−25^ 8.9^−12^ 4.9^−8^ Sordariomycetes Ncras 365 2.2^−68^ 5.9^−32^ 5.1^−14^ 1.5^−22^ 1001 6.0^−20^ 2.0^−9^ 2.1^−4^ 2.5^−8^ YES Sordariomycetes Tatro 351 9.2^−75^ 1.0^−33^ 4.8^−13^ 9.1^−24^ 269 1.1^−28^ 1.5^−9^ 2.3^−10^ YES Sordariomycetes Ntetr 366 6.1^−69^ 1.2^−32^ 5.2^−14^ 2.9^−22^ 990 1.1^−21^ 2.0^−9^ 2.1^−4^ 2.4^−8^ YES Leotiomycetes Sscle 386 2.9^−73^ 8.4^−32^ 1.1^−16^ 2.1^−21^ 249 2.9^−20^ 1.1^−12^ 5.4^−4^ Sordariomycetes Foxys 362 5.1^−67^ 2.6^−31^ 5.2^−12^ 7.2^−22^ 832 5.3^−18^ 5.5^−10^ YES Sordariomycetes Fgram 361 2.5^−67^ 3.0^−31^ 4.8^−12^ 4.1^−22^ 836 1.3^−19^ 6.2^−10^ YES Sordariomycetes Trees 367 1.9^−70^ 1.0^−33^ 8.5^−13^ 6.2^−22^ 274 2.9^−28^ 4.2^−8^ 1.4^−7^ YES Sordariomycetes Fvert 359 1.8^−67^ 2.9^−31^ 6.5^−12^ 2.4^−22^ 722 1.3^−18^ 4.6^−10^ YES Dothideomycetes Mfiji 368 5.9^−70^ 5.5^−33^ 1.5^−11^ 1.2^−19^ 195 6.6^−47^ 5.9^−16^ 2.8^−11^ 1.1^−4^ 8.4^−13^ Sordariomycetes Moryz 380 5.2^−61^ 6.5^−30^ 8.2^−6^ 3.5^−21^ 294 4.3^−20^ 6.7^−9^ YES \*YES indicates acidic region of FlbE is present. The nine orthologs used for motif discovery are in bold. ::: We then generated Hidden Markov Model (HMM) [@pone.0017505-Eddy1] profiles for the conserved motifs. Full length HMM profiles were also generated so that relative distances between the orthologs and the nine Eurotiales could be scored. The HMM profile scores are reported in [Table 4](#pone-0017505-t004){ref-type="table"}. Full-length sequence comparison with HMM {#s2h} ---------------------------------------- When the sequences were evaluated using the respective full-length HMM profiles, all the FlbB and FlbE orthologs were scored at significant E values ([Table 4](#pone-0017505-t004){ref-type="table"}). Among all the Eurotiomycetes, FlbE full-length HMM E values, in general, followed the trend of the respective FlbB E values (by which [Table 4](#pone-0017505-t004){ref-type="table"} is ordered). On the other hand, full-length HMM E values for the Sordariomycetes and Leotiomycetes FlbEs were higher, exhibited more interspecies variation and greater divergence from the FlbB scored order than those for the Eurotiomycetes. Additionally, there was much more variation in the length of the FlbE orthologs among the Sordariomycetes (e.g., 1014 residues for Ndisc compared to 201 for Anidu, [Table 4](#pone-0017505-t004){ref-type="table"}). These findings suggested that there was substantial variation among the FlbE orthologs outside of the Eurotiomycetes ([Table 4](#pone-0017505-t004){ref-type="table"}) at the species level. Note however, that all the FlbE orthologs had significant full-length HMM E values indicating that they encoded FlbE functionality. Motif conservation among orthologs {#s2i} ---------------------------------- FlbB motifs B1, B3 and B4 were found at significant E values in all 40 FlbB orthologs ([Table 4](#pone-0017505-t004){ref-type="table"}). That B1 and B4 are highly conserved is consistent with their being largely comprised of structural homologies. The fact that B3 is highly conserved suggests that it also encodes a region important for FlbB function. B2, on the other hand, was limited to the nine Eurotiales from which the motifs were derived plus a lower confidence match with Tstip. This limited distribution, at first, would seem to be an artifact of the motif derivatization process since the motif was found to be highly conserved only in the orthologs that were used to define it. However, the three other conserved regions (B1, B3 and B4) had a wide distribution among the FlbB orthologs suggesting that conserved motifs extracted with our methodology were not always exclusive to the nine Eurotiales. Additionally, it is unlikely that insertions and deletions caused the limited distribution of B2 since both B2 and B3 include gapped regions in alignment of the 40 FlbB orthologs ([Figure S1](#pone.0017505.s001){ref-type="supplementary-material"}). One hypothesis for the limited distribution of B2 could be that it encodes a novel function that is not present in the other orthologs. Motif distribution among the FlbE orthologs showed more variation than was seen in the FlbB motifs ([Table 4](#pone-0017505-t004){ref-type="table"}, [Figure 1 B](#pone-0017505-g001){ref-type="fig"}). These differential and unique distributions could indicate that they encode various functionalities that arose at different points in the evolution of FlbE. While FlbE is not the primary focus of this study, the varying levels of motif conservation could be a means to prioritize areas for experimental investigation of FlbE function. Indeed, a portion of E1, the most conserved motif, has recently been shown to be essential for FlbE function [@pone.0017505-Kwon1]. Of particular interest for our study of FlbB functional determinants is that B2 and E2 share the same distribution. This shared distribution suggests that these two motifs may be functionally related. FlbB and FlbE function outside of Eurotiales {#s2j} -------------------------------------------- FlbB, FlbD and FlbE form a branch of the upstream conidiation pathway that has been well studied in Anidu and related fungi. All three of these proteins are crucial for timely conidiation with knockouts of the respective genes exhibiting a strong aconidial fluffy phenotype. The exact function the orthologs of FlbB and FlbE in other species is unknown, however, two orthologs of the FlbB/D/E pathway have been characterized in Ncras. The gross phenotype of a knockout of the FlbB ortholog in Ncras (NCU07379) was characterized by the Neurospora Genome Project (NGP) [@pone.0017505-Colot1] with the results available on the Broad Institute website (<http://www.broadinstitute.org/annotation/genome/neurospora/AlleleDetails.html?sp=S989&sp=S7000006085195119>).This mutant conidiates normally. Unfortunately, the Ncras FlbE ortholog (NCU05255) has not been characterized by the NGP. However, a knock out of the third player in FlbB/D/E branch of the asexual development pathway, FlbD, has been shown to have no discernable phenotype by the Ebbole group [@pone.0017505-Shen1] and to conidiate normally by the NGP (NCU01312). In view of this result, we hypothesize that a null allele of Ncras FlbE would also display a wild type phenotype in terms of conidiation. Discussion {#s3} ========== The 40 FlbBs characterized in this study all contain highly conserved bZip and C-terminal Yap1-like domains, share the same general topology and are similarly sized. Additionally, all the orthologs have highly significant HMM E values for full length FlbB and conserved motif B1 and B4 profiles. These findings, combined with their unremarkable evolutionary relationships suggest that the FlbB proteins of the 40 species are indeed orthologs and likely function as key developmental regulators in a manner similar, but perhaps not identical, to Anidu FlbB (as noted above for Ncras). However, similarities and differences among the sequences identify both putative key functional determinants as well as important differences among the orthologs from the different orders and classes within the Pezizomycotina. The eight Eurotiales that have an intact bZip DBD, Arg as the positive residue in the dimerization salt bridge, Glu in the 3^rd^ heptad 'd' position, all six relevant conserved cysteines and the B2/E2 conserved motif pair ([Figure 2](#pone-0017505-g002){ref-type="fig"}, yellow highlight) are likely to share functionality with Anidu FlbB to the extent that results from experimental characterizations of that protein could apply to each. Additional support for shared functionality is provided by the low E values obtained by all eight orthologs (\<8^-230^) when evaluated with the nine Eurotiales full length HMM profile ([Table 4](#pone-0017505-t004){ref-type="table"}). Follow up experimental work in this regard could involve cross-complementation studies among these eight species. For example, it was recently shown that the Anidu FlbB sequence was able to partially complement an Afumi FlbB^-^ strain [@pone.0017505-Xiao1]. Such experiments could be informative as to the determinants of particular functions. Cross-complementation studies among fungal proteins have proven to be informative and an effective way to transfer characterizations of proteins in one species to other, less genetically amenable, species [@pone.0017505-Shen1]-[@pone.0017505-Bayram1], [@pone.0017505-Yamada1], [@pone.0017505-Fleck1]. However, FlbB orthologs outside Eurotiales likely function differently than those most closely related to Anidu. Differences in function could be due to: alterations in the bZip dimerization domain, differential distribution of conserved cysteines and the lack of the putative interaction motif pair B2-E2. For example, the H19 bZip DBD is present in all the FlbB orthologs but the dimerization profiles of the 22 Eurotiomycetes likely differ from the other 18 orthologs because the charge along the hydrophobic bZip dimerization interface differs in polarity and location in the two groups. The presence of a histidine (Anidu His93) in the DBD, which we designated H19, puts FlbB DBDs outside of any of the described bZip subfamilies [@pone.0017505-Fujii1]--[@pone.0017505-Amoutzias2]. The H19 DBD has remarkably limited distribution, especially considering the ubiquitous nature of bZip TFs in Eukaryotes. Furthermore, of the 22 putative bZips found in Anidu [@pone.0017505-Wortman1], only FlbB contains a H19 DBD (data not shown). These two conservation profiles suggest that H19 plays a specific role in FlbB function. Although DBD position 19 is among the five residues in contact with DNA, there is evidence that it does not always participate in DNA recognition [@pone.0017505-Amoutzias2]. Residues found in position 19 of bZip DBDs include cysteines (C19), serine (S19), tyrosine (Y19) and phenylalanine (F19) [@pone.0017505-Amoutzias2]. It has been shown that serine and cysteine in position 19 do not necessarily contribute to DNA binding [@pone.0017505-Abate1], [@pone.0017505-Deppmann2]. On the other hand, phenylalanine in position 19 has been shown to be important in fungi for Pap1 recognition sequence binding [@pone.0017505-Fujii1]. Contrary to this finding, the H19 containing Anidu FlbB has been demonstrated to bind these same Pap1 sites [@pone.0017505-Etxebeste2], [@pone.0017505-Fujii1]. This apparently unaltered binding site specificity of a F19-\>H19 substitution supports the relative unimportance of position 19 for DNA binding. These differential findings may have arisen because the functional difference between F19 and H19 has more to do with relative affinity rather than absolute site recognition. This could be investigated using reciprocally cross-mutated versions of FlbB and Pap1 or by analyzing the effect of a Phe-\>His substitution in FlbB. In addition to DNA binding site recognition, phosphorylation of S19 [@pone.0017505-Deppmann2], [@pone.0017505-Mahoney1] and oxidation of C19 [@pone.0017505-Abate1], [@pone.0017505-Xanthoudakis1] have been shown to be post translational mechanisms for abrogating DNA binding. A clue as to the functional importance of the histidine in position 19 could be that that histidine, too, can be phosphorylated [@pone.0017505-Puttick1]. This possibility is highly speculative as histidine phosphorylation occurs in only a few specific functions such as phosphorylations in the first step in two-component system signaling, of non-adenine nucleoside diphosphate by nucleoside diphosphate kinase and of histone H4 by histidine kinase [@pone.0017505-Besant1]. Another mechanism by which the DNA binding activity of the H19 DBD could be modulated is that histidine, with a pI of 7.6, could conceivably function as a pH sensor such that small changes in nuclear pH could modulate DNA binding. Although there are no detailed studies on the pH of fungal nuclei, it is known that the cytoplasmic pH of Anidu is maintained at 7.6 under a variety of external conditions [@pone.0017505-Hesse1]. In such an environment, relatively small shifts in pH will alter the charge of histidine and thereby alter its affinity for DNA. To date, no link has been described between FlbB and pH response or with the regulator of this response in Anidu, PacC [@pone.0017505-Pealva1], however, the possibility that DNA binding could be modulated through one or both of these mechanisms makes experimental work to confirm or refute these hypotheses quite compelling. Experimental results show that Anidu FlbB G70 residue is not essential for target DNA recognition in vitro but seems to be necessary to increase the efficiency of the binding. Based on this observation, we could suggest that although the consensus (NxxAQxxHR) sequence defines the specificity of the target DNA binding, auxiliary residues, as G70 in the case of FlbB, modulate the efficiency of the interaction. For example, a Gly residue plays a key role in the formation of the Hap complex in Anidu [@pone.0017505-Tanaka1]. It is part of a region necessary for the recruitment of HapX to the Hap complex and in subsequent binding to the regulatory sequence [@pone.0017505-McNabb1]. As the region containing G70 is conserved in all 40 orthologs, any specific knowledge gained from the study of any of the 40 FlbB orthologs in this respect would likely apply to all of them. The similarities between the FlbB orthologs and the C-terminal region of Yap1 and Pap1 raise two questions. First, although the C-terminal regions of the FlbB orthologs likely function similarly to these two oxidative response regulators (i.e., mechanistic homology) is FlbB, in fact, a functional homolog? That is unlikely as the regulatory role of FlbB is directed towards the induction and control of cellular development [@pone.0017505-Garzia1]-[@pone.0017505-Etxebeste3] rather than specific responses to oxidative stress. Additionally, NapA, a Yap1/Pap1 functional homolog has been characterized in Anidu [@pone.0017505-Asano1]. This protein, besides not being linked to conidiation, has been shown to function as an oxidative response regulator and shares the signature phenylalanine-containing PAP subfamily bZip DBD (F19) [@pone.0017505-Fujii1] with Yap1 and Pap1 rather than the distinctive H19 DBD of FlbB [@pone.0017505-Asano1]. The second question is whether or not this region participates in the modulation of localization and stability of FlbB as it does in Yap1 and Pap1. Such modulation is consistent with the observed alterations in Andiu FlbB nuclear localization at different growth and development stages [@pone.0017505-Etxebeste2], however, further experimental work is needed to establish links between this proposed mechanism and FlbB localization. The distribution of cysteines in the FlbB orthologs provides little information in terms of determining functional cysteine pairs except for two points: 1) the eight orthologs containing all six highly conserved cysteines ([Figure 2](#pone-0017505-g002){ref-type="fig"}, [Figure 4](#pone-0017505-g004){ref-type="fig"}, [Figure S1](#pone.0017505.s001){ref-type="supplementary-material"}) could function similarly to Anidu FlbB in that the same options for intramolecular disulfide bond(s) formation would exist, and 2) a single disulfide bond, such as could be formed between single cysteines pair C272 and C382 in the Dothideomycetes could be sufficient for minimal functionality required for a hypothetical redox control mechanism (at this point, a putative requirement for FlbB function). In this regard, it is worth noting that Yap1 and Pap1 form different disulfide bonds depending on conditions and/or to alter the longevity of the response [@pone.0017505-Tachibana1], [@pone.0017505-Castillo1] but also that a single cysteine pair could be sufficient for function of this mechanism although in this case there would be only one level of response. Support for C272 being the most likely cysteine to form a di-sulfide bond with C382 comes from the findings that 1) it is the second most conserved cysteine in FlbB among the orthologs, 2) it is within the highly conserved B3 motif and a region important for FlbB function [@pone.0017505-Etxebeste3], and 3) the area of B3 that contains C272 contains predicted order and helix secondary structure which is consistent with the possibility of structure formation with C-terminal region in a manner similar to Yap1 [@pone.0017505-Wood1]. It should also be noted that both Yap1 and Pap1 contain functional nuclear localization sequences (NLS) [@pone.0017505-Isoyama1], [@pone.0017505-Umeda1] and nuclear export signals (NES) [@pone.0017505-Kuge1], [@pone.0017505-Kudo1] that function in conjunction with di-sulfide bond-mediated NES masking in order to affect differential nuclear/cytoplasmic localization. Neither NLS nor NES, both of which can be cryptic [@pone.0017505-Boulikas1], [@pone.0017505-Bedard1], have been identified in any of the FlbB orthologs but experiments to elucidate them would be worth pursuing as a means of further understanding the Yap1/Pap1 mechanistic connection. Interpretation of the order predictions relative to the conserved regions suggests that both proteins contain a linker separating regions of structure. In FlbB, the proposed linker lies in the less conserved region that contains few structure predictions between motif B2 and B3 ([Figure 1 A](#pone-0017505-g001){ref-type="fig"}). In FlbE, a region with the same characteristics lies between motifs E4 and E5 ([Figure 1 B](#pone-0017505-g001){ref-type="fig"}). The majority of conserved motifs in both proteins are largely associated with predictions of order (B1, B3, B4, E1, E4 and E5). Motifs B2, E2 and E3 have less structure predictions associated with them suggesting that some of the conservation in these regions is not related to structure. The limited distribution of certain motifs (B2, E2, E3, E4, E5 and the FlbE acidic region) suggests that they may encode functional aspects that are not present in all the orthologs. Furthermore the *shared* limited distribution of conserved motifs B2 and E2 suggests that they may be linked functionally. Since FlbB and FlbE have been shown to interact in vivo and in vitro [@pone.0017505-Garzia1], of particular interest here is the possibility that B2 and E3 may facilitate this interaction. While highly speculative, this hypothesis specifies a starting point for mutational studies towards this end. Conclusions {#s3a} ----------- *In silico* analyses of the FlbB orthologs from 40 closely related filamentous fungi have revealed similarities and differences at the domain, motif and residue level. The 40 FlbB orthologs are highly similar and likely function as key developmental regulators in a manner similar to Anidu FlbB. While all contain structural homologies to bZip and the C-terminal region of Yap1, differences in key residues differentiate some orthologs from others. Changes in the bZip dimerization domain affect specificity and affinity for dimerization partners and could thereby alter the transcriptional activation profile of the functional dimers. The presence or absence of the conserved motif B2 and E2 pair could influence FlbB/FlbE interactions. Differences in the C-terminal cysteine pattern may provide a means for increased and/or differential functionality. Eight species: Anidu, Aterr, Anige, Aflav, Aoryz, Aclav, Afumi and Nfisc contain all the Anidu-centric features identified in this study and would therefore likely be able to functionally complement one another. However, the remaining orthologs lacked one or more of these critical features and could therefore function differently. Indeed, experiments have shown that Ncras FlbB apparently does not play a role in conidiation. However, its role in other cellular processes has not been investigated. Nevertheless, the differentially conserved residues and motifs identified here comprise a list of targets for functionally-directed mutational studies. Future directions {#s3b} ----------------- The abundance of fully sequenced fungal genomes makes in-depth approaches such as presented here feasible for almost any fungal protein. Indeed, more than 100 sequenced genomes are now available with more forthcoming in the near future from efforts like Joint Genome Institute (JGI) and the Fungal Genome Initiative (FGI). This study was focused on Anidu but the approach lends itself to the study of almost any fungal protein. Our Anidu-centric focus was intentional since we were interested in exploring the distribution of the experimentally characterized functionality of that species. Conducting a similar analysis based on Ncras, for example, could yield information about functional features encoded in that species that may not necessarily be present in Anidu. In some cases, this type of analysis could be informative as to the origin of proteins that have arisen through recombination or have been adapted to 'new' functions [@pone.0017505-Vandenbussche1]. Such analyses can be used to augment traditional approaches prior to initiating laboratory experiments but can also elucidate details of protein function and potentially lead to information regarding the origin of proteins and motifs that serve particular functions, protein interrelationships and/or pathway evolution. Methods {#s4} ======= Verification of ORF and exon/inton calling {#s4a} ------------------------------------------ Genomic sequences were manually compared with protein sequences in order to verify gene finding results reported by sequencing projects. When possible, experimentally verified protein sequences were used for reference (i.e., in Anidu and Afumi, both proteins have been sequenced). Phylogenetic relationships in terms of homology and splice site conservation among the orthologs were the main criteria to support the existence or absence of introns. Results from TBLASTN, BLASTX, Genewise (<http://www.ebi.ac.uk/Tools/Wise2/>) and Fgenesh (Softberry) [@pone.0017505-Solovyev1] were considered along with manual verification of splice sites. Shannon entropy calculations {#s4b} ---------------------------- Following alignment with CLUSTAL, column entropy [@pone.0017505-Durbin1] was calculated using the entropy web server [http://www.hiv.lanl.gov/content/sequence/entropy /entropy\_one.html](http://www.hiv.lanl.gov/content/sequence/entropy/entropy_one.html). Amino acid substitutions were not allowed in the calculation. Data for columns in which Anidu contained a gap were not considered for conserved motif selection. Columns in which the five-residue moving average of entropy was less than 0.5 were considered to have significant levels of conservation and were candidates for inclusion in motifs. Structure predictions and homology searches {#s4c} ------------------------------------------- Order was predicted using VLS2B (<http://www.ist.temple.edu/disprot/Predictors.html>) [@pone.0017505-Obradovic1]. GORIV was used for secondary structure prediction (<http://npsa-pbil.ibcp.fr/cgi-bin/npsa_automat.pl?page=npsa_gor4.html>) [@pone.0017505-Garnier1]. NCBI Conserved Domain Database search [@pone.0017505-MarchlerBauer1] results were taken from BLAST for Anidu FlbB. The 40 FlbB orthologs were scored for the presence of conserved domains by directly accessing the SMART database [@pone.0017505-Letunic1] through the website interface (<http://smart.embl-heidelberg.de/>). Conserved and domain search results were considered significant if the E value was less than 0.01. The Homstrad database [@pone.0017505-deBakker1] was queried using the Fugue sequence-structure homology recognition server [@pone.0017505-Shi1] (<http://tardis.nibio.go.jp/fugue/prfsearch.html>) using default settings to evaluate the FlbB orthologs for structural homologies. Searches of NCBI non-redundant database and recently completed genomes from the FGI and JGI were searched for H19 bZip DBD signatures using HMMs with profiles generated from the DBD of 40 FlbB orthologs and, in a recursive manner, profiles generated from extant H19 bZip DBD sequences found outside Pezizomycotina. Both standalone [@pone.0017505-Durbin1] and web-based (<http://mobyle.pasteur.fr/cgi-bin/portal.py>) HMM programs were used. Evaluation of the acidic region of FlbE {#s4d} --------------------------------------- The presence or absence of the C-terminal acidic region in FlbE orthologs was determined by using the Grantham method [@pone.0017505-Grantham1] as implemented on the ExPASy website (<http://www.expasy.org/tools/protscale.html>) [@pone.0017505-Gasteiger1] using a window of 15 residues. The acidic region was deemed to be present if the moving average of polarity scores exceeded 10 for 10 consecutive residues. An additional constraint was that this acidic region needed to be adjacent to the C-terminal end of motif E5. Generation of the FlbB C382A mutant strain {#s4e} ------------------------------------------ A pair of complementary oligonucleotides, flbB-C382A+1 and flbB-C382A-1, was designed bearing a TGC (coding for Cys382)-GCA (coding for an Ala382) substitution ([Table 5](#pone-0017505-t005){ref-type="table"}). These, plus two oligonucleotides flanking the flbB locus (flbB-PP1 and flbB-GSP4) were used to generate products for a fusion-PCR mutation procedure. Briefly, using genomic DNA from a strain expressing FlbB::GFP::pyrG [@pone.0017505-Etxebeste3] as a template, two DNA fragments were amplified: one of 2.9Kb covering the *flbB* promoter plus the corresponding sequence of the *flbB* locus (oligonucleotides flbB-PP1 and flbB-C382A-1) and the second one of 3.5Kb covering the rest of *flbB* locus, *gfp*, *pyrG* and the 3′ untranslated region (oligonucleotides flbB-C382A+1 and flbB-GSP4). Both fragments were fused [@pone.0017505-Yang1], purified and used to transform the wild type strain TN02A3 [@pone.0017505-Nayak1]. ::: {#pone-0017505-t005 .table-wrap} 10.1371/journal.pone.0017505.t005 Table 5 ::: {.caption} ###### Oligonucleotides used to generate the FlbB-C382A allele. ::: {#pone-0017505-t005-5} Designation Sequence (5′- 3′) -------------- -------------------------------------------------- flbB-PP1 GTTTTCTGGTCCTCGGTCAACCGGTGG flbB-GSP4 GAAAGGTGCGTGGGTTCGAATCCCACC flbB-GSP2 TGAATACATCGTCTCATCAGCATGCCGGGT flbB-C382A+1 GAGAACAAGGTGCGC**[GCA]{.underline}**TACGGATTCGGG flbB-C382A-1 CCCGAATCCGTA**[TGC]{.underline}**GCGCACCTTGTTCTC flbB-sek5 GCCGGGAAAACGCAACGC The substituted codon is in bold underlined font. ::: The parental strain allowed homologous recombination events either upstream or downstream of the TGC-\>GCA substitution, and this was reflected at the phenotypic level with transformants showing both wild type and aconidial phenotypes. Transformants were checked by Southern-blotting to confirm the appropriate recombination (data not shown) and the presence of the TGC-\>GCA substitution was confirmed by sequencing. With this aim, a 3.1Kb amplicon covering the *flbB* promoter and the entire coding region was generated using oligonucleotides flbB-PP1/flbB-GSP2, and sequenced using oligonucleotide flbB-sek5. The phenotype of the strain expressing the FlbB allele bearing the C382A substitution was analysed in *Aspergillus* Minimal Media [@pone.0017505-Kfer1] after 72 hours of growth and compared with the parental wild type TN02A3, Δ*flbB*, *flbB100* (G70R point substitution) and *flbB102* mutant strains, the latter being a mutant allele with a truncation after amino acid P305 [@pone.0017505-Etxebeste3]. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### **Alignment of the 40 Pezizomycotina FlbB orthologs used in this study.** Anidu FlbB is in bold. Motifs B1, B2, B3 and B4 are labeled and highlighted in green or yellow in the nine sequences that were used to generate the motifs. The five signature residues of the bZip DNA binding domain are highlighted in purple. The first four heptads of the bZip dimerization domain are identified by brackets with the residue positions labeled a -- g according to convention (only the first residue of the fourth heptad was positively identified). In the bZip dimerization domain, hydrophobic and charged residues in positions 'a' and 'd' (zipper forming residues) are highlighted in grey or orange, respectively, and salt bridge residues are highlighted in light blue. Specific residues discussed in text (Anidu numbering) are labeled above the alignment. Cysteine residues are additionally highlighted in light red. Residues flanking intron locations are in bold italic. (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### **Alignment of the 40 Pezizomycotina FlbE orthologs used in this study.** Anidu FlbE is in bold. Conserved motifs E1, E2, E3, E4 and E5 are labeled and highlighted in green or yellow. The acidic region is highlighted in purple. Residues flanking intron locations are in bold italic. (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was supported by the Basque Government (<http://www.hezkuntza.ejgv.euskadi.net>) through grant IT393-10 to U.U. and by the Spanish Ministerio de Educación y Ciencia ([www.mec.es](http://www.mec.es)) through grant BFU2009-08701 to E.A.E. M.S.C. was a contract researcher of the Ikerbasque program of the Basque Government ([www.ikerbasque.net](http://www.ikerbasque.net)). O.E. was a contract researcher of the University of the Basque Country ([www.ehu.es](http://www.ehu.es)). A.G. held a predoctoral fellowship from the Basque Government. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: MSC OE AG EAE UU. Performed the experiments: MSC. Analyzed the data: MSC OE. Wrote the manuscript: MSC OE. Critical review: OE AG EAE UU.

PubMed Central

2024-06-05T04:04:19.779052

2011-3-10

PMC3053369

Introduction {#s1} ============ Leishmaniasis is a disease caused by different species of protozoa of the genus *Leishmania* that are transmitted by Phlebotomine sandflies. It has traditionally been classified in three different clinical forms, visceral (VL), cutaneous (CL) and mucocutaneous leishmaniasis (MCL), which have different immunopathologies and degrees of morbidity and mortality. VL is caused by *Leishmania donovani* in the Indian subcontinent, Asia, and Africa, *L. infantum* in the Mediterranean basin, and *L. chagasi* in South America and is usually fatal if left untreated [@pone.0017376-Desjeux1], [@pone.0017376-Lukes1]. In the absence of vaccines, pentavalent antimonials like sodium stibogluconate (SSG) and meglumine antimoniate remain the first line therapy for CL and VL for over half a century, and are still in use in many parts of the world. The prolonged treatment requiring parenteral administration, toxicity, and the emergence of significant resistance are all factors limiting the drugs\' usefulness [@pone.0017376-Croft1]. Although few other options are available such as amphotericin B (AmB), miltefosine, paromomycin, and lipid-conjugated formulations of AmB, they also suffer from one or more limitations [@pone.0017376-Croft2]--[@pone.0017376-Sundar1]. Approaches to overcome antimonial resistance include use of alternative drugs which as mentioned earlier, are not devoid of limitations, or combination therapy having synergistic effects to check the further development of resistance [@pone.0017376-Chunge1]--[@pone.0017376-Olliaro2]. Additionally, effective therapies against SSG-resistant parasites include reversing their resistant nature with verapamil or buthionine sulfoximine in combination with SSG [@pone.0017376-Valiathan1], [@pone.0017376-Carter1]. Reports of resistance development towards newer drugs [@pone.0017376-Kumar1], considerable variation in resistance mechanisms in field isolates [@pone.0017376-Mandal1], and lack of progress in drug discovery add to the severity of the problem. A lot of research has focused on understanding the probable mechaninsms of drug resistance in *Leishmania* with main focus on SSG resistance. The reason for the emergence of resistance seems complex and multifactorial. As reviewed by Croft et al. [@pone.0017376-Chunge1], reduction of drug concentration within the parasite, either by decreasing drug uptake or by increasing efflux/sequestration of the drug, constitutes the primary mechanism of antimonial resistance; other potential resistance mechanisms include inhibition of drug activation, inactivation of active drug, and gene amplification. In the light of the above complications, improvement in the mode of administration and action of the existing and proven antileishmanials can become an important strategy in the management of VL [@pone.0017376-Frezard1]. Liposomes are artificially prepared vesicles made of lipid bilayer. They can act as targeted drug delivery agents to macrophages in parasitic infections [@pone.0017376-Owais1]. Liposomal SSG formulations against SSG-responsive parasites have been shown to enhance the antileishmanial activity of free SSG [@pone.0017376-Alving1]--[@pone.0017376-New1]. Carter et al. reported a differential organ dependent activity of non-ionic vesicular formulation of SSG in some clinical isolates of *Leishmania* which did not respond to free SSG [@pone.0017376-Carter3], but information on the development of long lasting protective cure against SSG-resistant parasite infection with liposomal SSG therapy is lacking. Earlier, we reported that cationic liposomes with egg phosphatidylcholine (PC) and stearylamine (SA) had leishmanicidal activity [@pone.0017376-Afrin1], [@pone.0017376-Dey1] and entrapment of SSG (PC-SA-SSG) enhanced their potentiality against chronic VL in in vivo murine model [@pone.0017376-Pal1]. PC-SA liposomes kill *Leishmania* by specific interaction with surface phosphatidylserine (PS) of promastigotes and amastigotes [@pone.0017376-Banerjee1]. We had earlier reported on the successful immunomodulation of the host with a liposomal AmB formulation as a curative strategy in treating SSG-responsive VL infection [@pone.0017376-Banerjee2]. Moreover, liposomes with their slow release property are known to circumvent the membrane efflux pathways thus increasing the intracellular retention of drug [@pone.0017376-Mamot1]. In the present study we compared the potential of single dose PC-SA-SSG therapy with PC-cholesterol formulation (PC-Chol-SSG) and AmB against SSG-resistant *L. donovani* GE1F8R and CK1R infection in BALB/c mice. These data add a new dimension to the therapy of SSG-unresponsive *L. donovani* infection. The insight into the probable mechanisms by which PC-SA associated drug works gives a broader perspective to this study in aiming other infections with similar immunological profiles and drug resistant phenotypes. Materials and Methods {#s2} ===================== Animals and parasites {#s2a} --------------------- BALB/c mice and Golden Syrian hamsters, bred in the animal facility of Indian Institute of Chemical Biology, Kolkata, India, were used for the studies. The studies were approved by the Institute\'s Animal Ethical Committee (147/1999/CPCSEA) and animals were handled according to their guidelines. *L. donovani* SSG-sensitive AG83 (MHOM/IN/1983/AG83) and resistant, GE1F8R (cloned from MHOM/IN/1989/GE1) and CK1R (isolated as MHOM/IN/1995/CK from a SSG-unresponsive patient) strains were maintained in vivo by intracardiac injection into hamsters [@pone.0017376-Bhattacharyya1]. Transformation of amastigotes from infected hamster spleen to promastigotes was carried out at 22°C in medium-199 (Sigma-Aldrich, St. Louis, MO) supplemented with penicillin G sodium (l00 U/ml), streptomycin sulfate (100 mg/ml) and 10% heat inactivated fetal bovine serum (FBS) (Sigma-Aldrich) and subcultured in the same medium [@pone.0017376-Banerjee2]. Entrapment of SSG, rhodamine 123, and 5-carboxyfluorescein in liposomes {#s2b} ----------------------------------------------------------------------- Liposomes were prepared with PC form egg yolk and Cholesterol (Sigma-Aldrich) or SA (Fluka, Buchs, Switzerland) at 7:2 molar ratios, respectively. For encapsulation of SSG (Gluconate Health Limited, Kolkata, India) the lipid film containing 20 mg of PC was mixed with either 2 mg of SA or 3 mg of Chol and dispersed in 0.02 M phosphate buffered saline (PBS) pH 7.4 containing 1 mg/ml of SSG and sonicated in an ultrasound probe sonicator (Misonix, Farmingdale, NY) for 2 min with intervals on ice. Unentrapped SSG was removed by centrifuging thrice at 60,000×g, for 30 min each [@pone.0017376-Pal1]. The amount of SSG entrapped in PC-SA and PC-Chol liposomes as determined colorimetrically [@pone.0017376-Pal1], ranged from 15 to 20 *µ*g/mg of PC and the efficiency of entrapment was approximately 30 to 40%. To entrap rhodamine 123 (Rh 123) (Calbiochem, San Diego, CA) in PC-SA (PC-SA-Rh 123), freshly prepared Rh 123 solution in chloroform was added to the lipids [@pone.0017376-Kang1] and liposome was prepared as described above. For entrapment of 5-carboxyfluorescein (CF) (Sigma), lipid film was dispersed in PBS containing 50 mM CF [@pone.0017376-Mora1] and liposomes were prepared as above. The entrapment efficiency estimated from a standard curve of the respective dyes after disrupting the liposomes with 1% TritonX-100 (Sigma) and measuring the fluorescence intensity with Fluorescence spectrophotometer (F-7000 FL spectrophotometer, Hitachi High Technologies, Japan) (505 nm/535 nm for Rh 123 and 492 nm/517 nm for CF) was found to be nearly 90% for Rh 123 containing liposome and 75% for CF containing liposome. In vitro antileishmanial assay and quantification of intracellular SSG of amastigotes {#s2c} ------------------------------------------------------------------------------------- Peritoneal macrophages isolated from BALB/c mice were pooled and cultured at 37°C in 5% CO~2~ in RPMI-1640 (Sigma-Aldrich) supplemented with FBS and antibiotics as described above. SSG-sensitive AG83 and resistant GE1F8R promastigotes were allowed to infect peritoneal macrophages (10∶1) for 3 h. Infected cells were thereby treated with various doses of liposomal or free SSG, for 72 h at 37°C. The cells were then fixed and stained with Giemsa for microscopic determination of intracellular parasite numbers per 200 host cells [@pone.0017376-Pal1]. In parallel sets of experiments, cells were scraped and amastigotes were freed from their host cells, as described [@pone.0017376-Mukherjee1]. Dried amastigotes were digested with nitric acid overnight and diluted with deionised water for antimony analysis by atomic absorption spectroscopy (AAS) (Perkin-Elmer 4100 ZL, CA, USA) [@pone.0017376-Roberts1]. Dye uptake and retention assay {#s2d} ------------------------------ Rh 123 uptake and retention studies were performed according to Kang et al. with slight modifications [@pone.0017376-Kang1]. *L. donovani* AG83 and GE1F8R promastigotes were washed and resuspended (2×10^5^ parasites/ml) in serum free M-199, and incubated along with different concentrations of free Rh 123 for indicated periods of time at 22°C. For retention studies, parasites were incubated with 250 ng/ml of free or entrapped Rh 123 for 4 h, washed and further incubated in media free of Rh 123. In case of CF, cells were preincubated with 2 µM of the dye [@pone.0017376-Teng1]. At indicated time points, parasites were washed thrice in PBS and finally lysed in 0.1% Triton X-100. The intracellular dye concentrations were determined by measuring fluorescence intensity of the cell lysates and comparing them with the standard curves as above. The percent retention of free *vs* liposomal dyes were calculated with respect to controls. Infection of mice and treatment regimen {#s2e} --------------------------------------- For experimental infections, BALB/c mice (4--6 weeks) were injected via the tail vein with 2.5×10^7^ hamster spleen-derived *L. donovani* amastigotes (in 200 µl 0.02 M PBS/mouse), and for reinfection, the same number of amastigotes were injected 12-wk after the primary infection. Eight weeks postinfection, groups of animals were treated intravenously with either 300 mg/kg of free SSG, or 12 mg/kg of SSG entrapped in PC-SA or PC-Chol liposomes. Amphotericin B deoxycholate (AmB) (kind gift from R. P. Goswami and B. Saha of School of Tropical Medicine, Kolkata, India) was administered intravenously at a single dose of 2 mg/kg/mice. Mice were sacrificed 4-wk post-treatment, and spleen and liver parasitic loads were determined from Giemsa-stained impression smears and reported as Leishman Donovan Units (LDU), calculated as the number of parasites per 1000 nucleated cells x organ weight (in mg) [@pone.0017376-Banerjee2], [@pone.0017376-Stauber1]. In selected groups, a weighed piece of spleen or liver from experimental mice was first homogenized in Schneider\'s Drosophila medium (Invitrogen Corporation, Carlsbad, CA) supplemented with 10% FBS, and then diluted in the same medium to a final concentration of 1 mg/ml. Five-fold serial dilutions of the homogenized tissue suspensions were plated in 96-well plates and incubated at 22°C for 21days, with periodic checking at 7-day intervals for viable and motile promastigotes. The reciprocal of the highest dilution that was positive for parasites was considered to be the parasite concentration per milligram of tissue. The total organ parasite burden was calculated using the weight of the respective organs [@pone.0017376-Banerjee2]. Infection in bone marrow was calculated as parasites/1000 host cell nuclei [@pone.0017376-Banerjee2]. Mice reinfected at 4-wk post-treatment were sacrificed at 20-wk of initial infection and organ parasite burden was determined as above. Detection of IgG isotype levels in the serum {#s2f} -------------------------------------------- Mice were bled 4 wk after treatment, and sera were stored at -20°C until use. Antigen-specific serum immunoglobulin (Ig)-G isotype antibody response was measured by conventional ELISA as described [@pone.0017376-Banerjee2]. Briefly, 96-well ELISA plates (Maxisorp, Nunc, Roskilde, Denmark) incubated overnight at 4°C with 2.5*µ*g/well *Leishmania* membrane antigen (LAg) [@pone.0017376-Banerjee2], were blocked and further incubated with mice sera (1∶1000 dilutions) for 1 h, washed thoroughly, followed by 1 h incubation at 37°C with peroxidase-conjugated goat anti-mouse IgG1 or IgG2a antibodies (BD Pharmingen, San Diego, CA). The plates were then developed for color reaction with substrate solution (*o*-phenylenediamine dihydrochloride, 0.8 mg/ml in phosphate-citrate buffer (pH 5.0), containing 0.04% H~2~O~2~) for 30 min, and absorbance was measured on ELISA plate reader (Thermo Electron Corporation, Waltham, MA) at 450 nm. Splenocyte-proliferation and analysis of cytokines {#s2g} -------------------------------------------------- Splenocyte-proliferation assay was performed for different experimental groups. Cell suspension was prepared by mechanical disruption of spleen, followed by red blood corpuscles (RBC) lysis with 0.14 M Tris buffered NH~4~Cl. After several washings in RPMI 1640 medium, cells were resuspended in complete medium (RPMI 1640 supplemented with 10% FBS, l00 U/ml penicillin G sodium, 100 mg/ml streptomycin sulfate and 50 mM β-mercaptoethanol \[Sigma-Aldrich\]). Viable mononuclear cell number was determined by counting Trypan blue unstained cells in a hemocytometer. Cells were plated in triplicate at 2×10^6^ cells/ml concentrations in 96-well plates (Nunc, Roskilde, Denmark) and allowed to proliferate for 72 h at 37°C in 5% CO~2~ incubator in presence of 12.5*µ*g/ml LAg [@pone.0017376-Afrin2]. Cells were pulsed further for 18 h with 0.5*µ*Ci of \[^3^H\] thymidine/well (Amersham Biosciences, Buckingham-shire, UK), harvested on glass fiber paper, and radioactivity was measured in a liquid scintillation counter (Beckman Instruments, Fullerton, CA). In parallel experiments, cytokine production by splenocytes was determined by ELISA kit (BD Biosciences, San Diego, CA), as per manufacturer\'s instruction. For in vitro depletion, total splenocytes were incubated with 1*µ*g/10^6^ cells of anti-CD4^+^ or anti-CD8^+^ mAbs (BD Biosciences) for 1 h at 4°C. Cells were washed and cultured in LAg as above. The efficacy of depletion was documented on each experiment by flow cytometry. Splenocytes from normal mice were cultured similarly in complete medium with different doses of free SSG or entrapped in PC-SA and PC-Chol liposomes for 48 h with or without LPS (2.5*µ*g/ml) [@pone.0017376-Banerjee2] and cytokine production was determined as above. Quantification of Nitric Oxide {#s2h} ------------------------------ Nitric oxide (NO) content in the culture supernatants from LAg-pulsed splenocytes cultured for 72 h was analyzed by Griess assay method according to Ding et. al., using NaNO~2~ diluted in culture medium as standard [@pone.0017376-Ding1]. Briefly, the mixture of Greiss reagent (1% sulfanilamide and 0.1% *N*-(1-naphthyl) ethylenediamine dihydrochloride in 2.5% H~3~PO~4~) and culture supernatant at 1∶1 ratio was incubated for 15 min at room temperature, and the OD was determined at 550 nm by ELISA reader (Thermo Electron Corporation, Waltham, MA). Statistical analysis {#s2i} -------------------- Statistical analyses were done using GraphPad Prism (GraphPad Software,v.5.0, San Diego, CA) software. A two-tailed Student\'s t-test was used to compare the significance between two groups. A one-way ANOVA (non-parametric) was used to compare more than two groups, followed by Tukey\'s multiple comparison test. Differences were considered statistically significant at p\<0.05. Results {#s3} ======= Cure and protection against reinfection conferred by PC-SA-associated SSG against drug-resistant *L. donovani* infection in mice {#s3a} -------------------------------------------------------------------------------------------------------------------------------- Earlier we reported a profound synergistic activity of SSG entrapped in PC-SA liposome in both in vitro and in vivo models of SSG-non-resistant VL [@pone.0017376-Pal1]. Here, therapeutic potency of PC-SA-SSG was compared with other formulations in susceptible BALB/c mice infected with virulent strains of SSG-resistant *L. donovani*, GE1F8R and CK1R, to validate the strain-independent efficacy of PC-SA-SSG. Progressively infected mice on treatment with free SSG and PC-Chol-SSG failed to suppress parasite load in GE1F8R, although they showed partial activity against CK1R strain ([Figure 1](#pone-0017376-g001){ref-type="fig"}). Interestingly, PC-SA-SSG suppressed liver parasitic load by 93% and 97%, respectively in GE1F8R and CK1R infected mice, which were comparable to AmB therapy ([Figure 1A, D](#pone-0017376-g001){ref-type="fig"}). In contrast to free SSG and PC-Chol-SSG, significant suppression with PC-SA-SSG was also achieved in spleen (98% and 96%) ([Figure 1B, E](#pone-0017376-g001){ref-type="fig"}) and bone marrow (84% and 86%) ([Fig. 1C, F](#pone-0017376-g001){ref-type="fig"}) of GE1F8R and CK1R infected mice respectively (p\<0.05). Strikingly, the more sensitive limiting dilution assay demonstrated superiority of PC-SA-SSG over AmB in suppressing liver and splenic parasite burden ([Figure 1G](#pone-0017376-g001){ref-type="fig"}) in both GE1F8R and CK1R infected mice (p\<0.05). To further strengthen our findings, 8-wk-infected BALB/c mice treated with PC-SA-SSG and AmB were re-infected with GE1F8R intravenously 4-wk after therapy. Age-matched normal mice were also infected simultaneously and were considered as age-matched controls. In comparison to AmB, mice treated with PC-SA-SSG were more resistant to re-infection with GE1F8R (p\<0.05) ([Figure 2A, B](#pone-0017376-g002){ref-type="fig"}). We observed only a slight increase in liver and spleen parasite burden in the PC-SA-SSG-treated group till 20-wk of initial infection. ::: {#pone-0017376-g001 .fig} 10.1371/journal.pone.0017376.g001 Figure 1 ::: {.caption} ###### Parasite burden in treated BALB/c mice after *L. donovani* challenge infection. Parasite loads of liver, spleen, and bone marrow in murine model of established visceral leishmaniasis after treatment with 300 mg/kg of free sodium stibogluconate (SSG), 12 mg/kg of SSG entrapped in phosphatidylcholine-stearylamine (PC-SA) or phosphatidylcholine-cholesterol (PC-Chol) liposomes and 2 mg/kg amphotericin B (AmB). Mice were infected with *L. donovani* amastigotes of either GE1F8R or CK1R strains. At 8 week postinfection (p.i.) mice were treated with a single dose of various drugs by intravenous injection. Mice were sacrificed after 4 weeks of treatment for determination of (A, D) liver, (B, E) spleen, and (C, F) bone marrow parasite loads. Untreated, infected mice were used as controls. Liver and spleen parasite burden were determined by stamp-smear method and expressed as Leishman Donovan Units (LDU), and bone marrow parasite load in cell smear prepared from femur bone marrow and expressed as amastigotes/1000 bone marrow nuclei. Data represent mean ± SEM (n = 5 mice per group), representative of two similar experiments. \* p\<0.05; \*\* p\<0.01; \*\*\* p\<0.001. (G) Eight week infected mice received optimal SSG, PC-SA-SSG and AmB. Parasite burden determined 12 weeks following infection reflects the mean log~10~ parasite burden ± SEM determined by the limiting dilution assay (LDA) (n = 5 mice per group) with PBS treated group as control. Data are representative of two similar experiments. *\** p\<0.05 compared to AmB therapy. :::  ::: ::: {#pone-0017376-g002 .fig} 10.1371/journal.pone.0017376.g002 Figure 2 ::: {.caption} ###### Parasite burden in cured BALB/c mice after reinfection with *L. donovani.* Cured mice (PC-SA-SSG and AmB treated) after 4 weeks of treatment along with naïve age-matched controls were reinfected with similar dose of virulent amastigotes and at 20 weeks of primary infection, were sacrificed and liver (A) and spleen (B) parasitic loads were determined by stamp-smear method and expressed as Leishman Donovan Units (LDU). Data represent mean ± SEM (n = 3-5 mice per group), representative of two similar experiments. \* p\<0.05; \*\* p\<0.01; \*\*\* p\<0.001. :::  ::: In vitro antileishmanial activity and differential accumulation of SSG induced by PC-SA and PC-Chol liposomes {#s3b} ------------------------------------------------------------------------------------------------------------- In accordance to our in vivo findings, PC-SA-SSG could induce significantly higher suppression towards parasite infection (p\<0.05) compared to PC-Chol-SSG treatment in both SSG-sensitive AG83 and SSG-resistant GE1F8R infected macrophages ([Figure 3A](#pone-0017376-g003){ref-type="fig"}). SSG resistance in *Leishmania* involves reduction in intracellular SSG accumulation, either as a result of its reduced influx or increased efflux from the parasite and/or macrophages [@pone.0017376-Croft2]. The suppression in parasite infection as observed in our experiments was associated with an elevated accumulation of SSG in the intracellular amastigotes of both AG83 and GE1F8R (p\<0.05) parasites compared to PC-Chol-SSG as measured by AAS ([Figure 3B](#pone-0017376-g003){ref-type="fig"}). Free SSG on the other hand showed negligible antileishmanial activity, accompanying poor drug accumulation within resistant GE1F8R, compared to sensitive AG83 ([Figure 3C, D](#pone-0017376-g003){ref-type="fig"}) (p\<0.05). ::: {#pone-0017376-g003 .fig} 10.1371/journal.pone.0017376.g003 Figure 3 ::: {.caption} ###### Parasite suppression in *L donovani*-infected macrophages and SSG accumulation in intramacrophagic amastigotes. Both AG83 and GE1F8R infected peritoneal macrophages were treated with (A, B) PC-SA-SSG, PC-Chol-SSG, and (C, D) free SSG for 72 h. (A, B) The treatment-induced percentage suppression of parasites was calculated in comparison to respective untreated-control. (B, D) Total SSG uptake by intramacrophagic amastigotes following various treatments was estimated by atomic absorption spectroscopy (AAS), and represented as nanogram (ng)/10^6^ cells. Data represent the mean ± SEM of three independent experiments. \* p\<0.05 compared to AG83 and GE1F8R strains treated with PC-Chol-SSG. :::  ::: PC-SA incorporation increases retention of rhodamine 123 and 5-carboxyfluorescein in *Leishmania* {#s3c} ------------------------------------------------------------------------------------------------- We and others have demonstrated the increased efficacy of vesicular SSG in the treatment of VL [@pone.0017376-Alving1]-[@pone.0017376-New1], [@pone.0017376-Carter3]. The exact mechanism by which this strategy works however remains to be studied. One of the mechanisms is the entrapment of drugs in liposomes, which results in circumvention of Pgp mediated drug efflux, and is well demonstrated in cancer therapy [@pone.0017376-Kang1]. Drug resistance in *Leishmania* has been variously described and involves active efflux systems working in the parasite and/or macrophages [@pone.0017376-Croft2]. Thus we speculated that entrapment of SSG in PC-SA can prevent its exposure to the active pumps in turn increasing its intracellular concentration. To investigate if PC-SA can circumvent the resistance mechanism active in the parasite, we used Rh 123 and CF, substrates for Pgp and MRP family respectively. Rh 123 concentrations of 250 ng/ml and uptake time of 4 h were chosen from the dose and time kinetic studies using free Rh 123 ([Figure 4A, B](#pone-0017376-g004){ref-type="fig"}). When AG83 and GE1F8R parasites, preincubated with either free Rh 123 or PC-SA-Rh 123, were again incubated in dye-free medium, the intracellular Rh 123 concentration decreased with time regardless of the parasite strain ([Figure 4C](#pone-0017376-g004){ref-type="fig"}). However, the fall was faster and greater in GE1F8R compared to AG83 for free Rh 123. Interestingly, loading Rh 123 into PC-SA liposome although showed no difference compared to free dye in AG83, it significantly increased the retention capacity of GE1F8R (1.6 to1.8 folds between 0.5 to 1.5 h) (p\<0.05 to 0.001). Loading the liposomes with CF produced a similar profile in SSG-unresponsive parasites. When GE1F8R promastigotes preincubated with the free dye were further incubated in dye free medium, 65.5±1.55%, 57.27±1.714%, and 37.57±1.068% dye were retained after 0.5 h, 1 h and 1.5 h respectively. When liposomal CF was used the retention increased to 83.21±1.76%, 74.28±2.73%, and 67.23±1.28%, with a significant rise of 0.8 to 0.6 folds between 0.5 h and 1.5 h (p\<0.01 to 0.001). Free Rh 123 was effluxed out more vigorously then free CF till 1 h of incubation after which CF levels dropped to 37%. ::: {#pone-0017376-g004 .fig} 10.1371/journal.pone.0017376.g004 Figure 4 ::: {.caption} ###### Intracellular retention of free and liposomal rhodamine 123 in *L.donovani* promastigotes. Rhodamine 123 (Rh123) uptake and retention study was performed in AG83 and GE1F8R strains with Rh123. Parasites were incubated with (A) different concentrations of free Rh123 for 4 h and (B) for different time periods. At indicated time points parasites were washed, lysed and fluorescence intensity of the cell lysates measured. Data reperesent mean ± SEM of two independent experiments performed in triplicates and expressed as ng of Rh123/10^5^ cells. (C, D) In another set of experiments, parasites incubated for 4 h with 250 ng/ml of either free or liposomal Rh123 were washed and reincubated in medium free of Rh 123 for different times followed by lysis. The percent retention of Rh123 was calculated for each group compared to respective controls. Data represent mean ± SEM of three independent experiments each performed in duplicate. \* p\<0.05; \*\*\* p\<0.001 :::  ::: Immunomodulatory effects of PC-SA-SSG on normal mice splenocytes {#s3d} ---------------------------------------------------------------- The protective chemotherapeutic response demonstrated by PC-SA-SSG prompted us to investigate its possible immunomodulatory role in vitro. As reported earlier [@pone.0017376-Saha1], SSG could suppress the disease promoting IL-10 production in normal LPS-pulsed splenocytes ([Figure 5A](#pone-0017376-g005){ref-type="fig"}). SSG (0.36 µg/ml) entrapped in PC-SA (20 µg/ml) further brought about a 1.7-fold higher reduction in IL-10 compared to equivalent amount of free SSG ([Figure 5B](#pone-0017376-g005){ref-type="fig"}). PC-Chol on the other hand, masked the ability of entrapped SSG to bring down IL-10 ([Figure 5C](#pone-0017376-g005){ref-type="fig"}). ::: {#pone-0017376-g005 .fig} 10.1371/journal.pone.0017376.g005 Figure 5 ::: {.caption} ###### Immunomodulatory activity of free and vesicular SSG on splenocytes of normal BALB/c mice. Splenocytes of normal healthy mice were incubated with various concentrations of (A) free SSG, (B) PC-SA-SSG, and (C) PC-Chol-SSG with or without LPS (2.5 µg/ml) for 48 h at 37°C with 5% CO~2~. IL-10 was measured from culture supernatants by enzyme-linked immunosorbent assay (ELISA). Each symbol represents mean cytokine level ± SEM, representative of three independent experiments. :::  ::: Humoral response in PC-SA-SSG treated mice {#s3e} ------------------------------------------ IgG2a levels are dependent on IFN-γ, whereas IgG1 levels correlate with IL-4. IgG2a and IgG1 are therefore used as surrogate markers for Th1 and Th2 responses [@pone.0017376-Coffman1]. PC-SA-SSG treatment mounted 2 and 1.3-fold (p\<0.05) higher *Leishmania* membrane antigen (LAg)-specific IgG2a compared to respective GE1F8R, and CK1R-infected controls, and almost steady IgG1 levels concomitant with a steady secretion of IL-4. Similar results were also observed in mice treated with AmB, but not with PC-Chol-SSG treatment ([Figure 6A, B](#pone-0017376-g006){ref-type="fig"}). ::: {#pone-0017376-g006 .fig} 10.1371/journal.pone.0017376.g006 Figure 6 ::: {.caption} ###### Humoral and antigen-specific proliferative response following treatment with PC-SA-SSG. (A, B) Sera from treated mice were analyzed individually by ELISA for detection of IgG1 and IgG2a antibodies in GE1F8R and CK1R-infected groups of mice (3-5 mice/group). The results are representative of two independent experiments and data represent mean ± SEM. \* p\<0.05 compared to PBS treatment. (C, D) Total, and CD4^+^ and CD8^+^ T cell-depleted splenocytes from treated groups were stimulated in vitro with *Leishmania* membrane antigen (LAg) (12.5*µ*g/ml). After 48 h, \[^3^H\] thymidine was added and cells were harvested subsequently. Proliferative index was measured as \[^3^H\] thymidine incorporation in counts per minute. Data represent mean ± SEM (3 mice/group) in triplicates. \* p\<0.05 compared to AmB therapy. :::  ::: PC-SA-SSG-driven antigen-specific proliferative response {#s3f} -------------------------------------------------------- Impairment of cell-mediated immune response in active VL patients is reflected by marked T cell anergy specific to *Leishmania* antigens [@pone.0017376-Banerjee2], [@pone.0017376-Murray1] which is reversed by successful therapy. To investigate whether PC-SA-SSG can do so, we performed LAg-specific T cell proliferation assay. In contrast to PC-Chol-SSG therapy, PC-SA-SSG triggered 11 and 12-fold higher LAg-specific proliferation compared to respective GE1F8R and CK1R infected mice ([Figure 6C, D](#pone-0017376-g006){ref-type="fig"}), which were even higher than AmB treatment (p\<0.05). The proliferative response was significantly blocked individually by both anti-CD4^+^ and anti-CD8^+^ antibodies inferring that marked proliferation was contributed by both subsets of T cells. Cyokine response and NO production {#s3g} ---------------------------------- To compare the type of immunological response in *L. donovani*-infected mice with PC-SA-SSG liposome treated mice, detailed analysis of cytokine production was conducted in splenocytes of differently treated infected animals by ELISA at 12-wk post-infection. A general Th1 dominance in PC-SA-SSG treated mice was evident from 18 and 16-fold higher Interferon (IFN)-γ secretion from LAg-pulsed splenocytes, than respective GE1F8R and CK1R-infected mice ([Figure 7A, C](#pone-0017376-g007){ref-type="fig"}) (p\<0.05) which was comparable with AmB therapy. In contrast, PC-Chol-SSG treated mice expressed no enhancement of IFN-γ. In vitro CD4^+^and CD8^+^ T cell depletion demonstrated significant role of the dichotomous T cell subsets in IFN-γ expression ([Figure 7E](#pone-0017376-g007){ref-type="fig"}). Elevated IFN-γ in PC-SA-SSG and AmB treated mice corresponded with a simultaneous upregulation of interleukin (IL)-12 and tumor necrosis factor (TNF)-α expression. IL-12 levels were elevated by 3.5 and 4-fold (p\<0.05) and TNF-α levels showed 1.5 and 1.8-fold increase in PC-SA-SSG treated mice, compared to GE1F8R and CK1R-infected mice ([Figure 7A, C](#pone-0017376-g007){ref-type="fig"}). ::: {#pone-0017376-g007 .fig} 10.1371/journal.pone.0017376.g007 Figure 7 ::: {.caption} ###### Differential pattern of cytokine and NO production following therapy with antileishmanial formulations in infected BALB/c mice. Splenocytes isolated from GE1F8R and CK1R-infected mice after indicated treatments were plated aseptically, and stimulated with LAg (12.5*µ*g/ml) for 72 h. (A, C) IFN-γ, IL-12, TNF-α and (B, D) IL-10, TGF-β, IL-4 cytokine levels in supernatants of splenocyte cultures were assayed by ELISA. Total, and CD4^+^ and CD8^+^ T cell-depleted splenocytes were stimulated as above and (E) IFN-γ and (F) IL-10 levels were measured after 72 h. Values represent the mean ± SEM (3--5 mice/group). (G) Leishmanicidal NO generation determined by Greiss assay method in supernatants of splenocytes derived from indicated groups. Data represent the mean ± SEM (3--5 mice/group). \* p\<0.05 compared to (A, C) PBS or (B, D,G) AmB therapy. :::  ::: We next investigated the altered expression of immunosuppressive cytokines of treated mice. Despite sustained levels of IL-4, PC-SA-SSG treated mice expressed negligible IL-10 (p\<0.05) and transforming growth factor (TGF)-β compared to GE1F8R and CK1R-infected mice ([Figure 7B, D](#pone-0017376-g007){ref-type="fig"}). In vitro blocking demonstrated CD4^+^ cells as the main source of IL-10 ([Figure 7F](#pone-0017376-g007){ref-type="fig"}). In comparison AmB could only partially suppress the disease promoting cytokines, IL-10 and TGF-β highlighting the better curative response of PC-SA-SSG therapy. Failure to suppress IL-10 and TGF-β accounted for poor outcome of the therapy with PC-Chol-SSG ([Figure 7B, D](#pone-0017376-g007){ref-type="fig"}). NO is the crucial killing effector molecule against leishmaniasis produced by IFN-γ-stimulated and NO synthase-induced classical macrophages. Th1 dominance in PC-SA-SSG treated mice correlated with 6-fold higher NO production than infected control, which was significantly higher than AmB treated group ([Figure 7G](#pone-0017376-g007){ref-type="fig"}) (p\<0.05). The strong ability to suppress disease promoting IL-10 and to effectively trigger macrophage microbicidal molecule NO thus reversing the immunosuppressive condition towards Th1 type immune response, accounts for the radical cure elicited by PC-SA-SSG. Discussion {#s4} ========== In this study we evaluated a new therapeutic approach with cationic liposomal SSG against SSG-resistant *L. donovani* parasites. Therapy with a single dose of SSG in PC-SA liposomes led to the successful cure of progressive SSG-resistant VL in BALB/c mice that was even better than AmB therapy in providing effective antileishmanial immunity and strong protection against reinfection. In contrast, equivalent amount of SSG in PC-Chol liposomes failed to evoke significant cure. Investigation into the probable mechanisms demonstrated that PC-SA-SSG could directly kill parasites irrespective of SSG-sensitivity owing to the leishmanicidal effect of PC-SA combined with higher accumulation of SSG within the amastigotes favored by this formulation, and a simultaneous protective immunomodulation of the host immune system. To our knowledge this is the first demonstration of robust effective treatment against infection with SSG-resistant *Leishmania* parasites in mice with a liposomal SSG formulation. Previously reported anionic or neutral liposomal pentavalent antimonial formulations [@pone.0017376-Alving1]--[@pone.0017376-New1] were largely restricted to SSG-responsive strains alone, except one where differential organ dependent response was observed in primary infection model with SSG-resistant parasites [@pone.0017376-Carter3]. As an extension to our previous observations on the therapy of PC-SA-SSG against SSG-sensitive *L. donovani* [@pone.0017376-Pal1], we herein observed an equipotent effect of this therapy against non-healing infection with SSG-resistant parasites in BALB/c mice. Resistance mechanisms in clinical isolates differ from those active in laboratory generated strains [@pone.0017376-Mastroianni1]. PC-SA-SSG was equally effective against differentially originated GE1F8R and CK1R reflecting its strain independent antileishmanial activity. A single dose of 12 mg/kg of PC-SA-SSG was remarkably effective not only against liver but also splenic and bone marrow parasites which was even better than AmB therapy. There was a 10^5^-fold fall in viable parasites with less than 100 viable parasites observed in the organs after PC-SA-SSG therapy, which is a clear indication of nearly complete healing. In comparison, PC-Chol-SSG exhibited incompetence against SSG-resistant parasites. It is well known in VL that a successful therapy may not clear all the parasites from sites of infection but most T cell intact patients show long lasting clinical cure despite the presence of residual intracellular parasites [@pone.0017376-Haldar1]. Whereas PC-SA-SSG effectively controlled parasite visceralization on a secondary attack, AmB showed only a partial protection to reinfection, further strengthening the superiority of this formulation. Prior experimental and clinical observations have pointed towards the importance of T cell mediated protective post-treatment mechanism in VL [@pone.0017376-Haldar1]. We tested the immunological outcome after chemotherapy in SSG-resistant *L. donovani*-infected mice treated with various therapies. Disease severity in BALB/c mice infected with SSG-resistant strains was associated with significantly hampered Ag-specific T cell proliferation, low expression of IL-12, TNF-α, IFN-γ and upregulation of suppressive cytokines IL-10 and TGF-β. Detailed immunological analysis of AmB and PC-SA-SSG treated mice showed enhanced T cell proliferation, persistent IgG1 levels, probably maintained by continued secretion of IL-4 [@pone.0017376-Alexander1], [@pone.0017376-Basu1] along with increased IgG2a and upregulated IL-12 and IFN-γ production in LAg-pulsed splenocytes, which was lacking in PC-Chol-SSG treated group. Treatment with PC-SA-SSG also increased the level of another proinflammatory cytokine, TNF-α, which probably stimulates IL-12 driven IFN-γ secretion. Strong IL-12 driven IFN-γ and TNF-α triggering in PC-SA-SSG versus PC-Chol-SSG treatment suggests that these cytokines might be involved in the observed upregulated NO secretion for providing impressive levels of protection. Although AmB induced significant Th1 responses, it failed to sufficiently suppress IL-10 and TGF-β production. On the other hand, PC-SA-SSG led to strong suppression of IL-10 and TGF-β production that correlated with successful resolution of infection. A growing body of literature correlates IL-10 and TGF-β with susceptibility to *Leishmania* infection [@pone.0017376-Ghalib1]--[@pone.0017376-BarralNetto1], [@pone.0017376-Saha1]. Since dominant host immunity over persistent infection was achieved by an ongoing Th1 response in the absence of immunosuppressive cytokines, animals treated with PC-SA-SSG could effectively prevent reinfection. Such a profound leishmanicidal potential of SA-bearing SSG formulation against SSG-resistant parasites prompted us to investigate the probable mechanisms for this phenomenon. PC-SA-SSG treatment of infected macrophage cultures demonstrated a direct killing of parasites concomitant with a higher SSG accumulation inside the amastigotes. The leishmanicidal activity of the cationic PC-SA-SSG may be due to their preferential uptake by the macrophages [@pone.0017376-Nakanishi1] followed by their cytotoxic action in the parasitophorous vacuoles. Killing of parasites by PC-SA occurs through specific interactions of the liposomes with the parasite membrane phosphatidylserine (PS), leading to membrane disruption and depletion in cytosolic ATP levels [@pone.0017376-Banerjee1]. The ATP-energized efflux transporters on parasite membranes responsible for the intracellular accumulation of drug [@pone.0017376-Ashutosh1] probably are rendered nonfunctional by PC-SA-SSG, enhancing SSG accumulation within resistant parasites. Additionally, liposomes are known to circumvent the action of membrane associated efflux pumps providing an alternate strategy to overcome drug resistance by increasing intracellular drug accumulation [@pone.0017376-Mamot1], [@pone.0017376-Kang1]. There are also reports that SA-bearing liposomes are less leaky towards cationic drugs [@pone.0017376-Webb1]. We used two dyes of different nature, Rh 123 which is entrapped in the lipid phase of the liposome and CF which is entrapped in the aqueous core. In addition, our aim behind using Rh 123 and CF was to assess their release profiles and see if PC-SA incorporation can increase their retention within resistant parasites, as the former is a well known Pgp substrate [@pone.0017376-Kang1] and latter is known to be transported by proteins of the MRP transporter family [@pone.0017376-Teng1]. Efflux of SSG in SSG-unresponsive parasites has been attributed to similar cellular processes [@pone.0017376-Ashutosh1]. Interestingly, PC-SA enhanced the intracellular retention of both the dyes, which are otherwise vigorously thrown out of the resistant cells [@pone.0017376-Kang1], [@pone.0017376-Kok1]. Although we didn\'t perform any experiments to confirm which efflux pumps is actually blocked, inhibited or downregulated in this case, but the retention studies clearly indicate a general decrease in efflux rates when liposomal incorporation was done. Thus, increased retention and slow release of the entrapped drug at the site of action add to the therapeutic advantage of PC-SA-SSG. Cationic SA-bearing liposomes, in addition to having antileishmanial activity, greatly enhance the immunogenicity of associated antigens [@pone.0017376-Afrin2]. Antimonials can act effectively in the presence of a propicious immune response [@pone.0017376-Murray2]. In view of our in vitro experiments on normal mice, it is clear that SSG exerted an inhibitory effect on IL-10 production that was augmented by incorporation in PC-SA liposomes. In vitro studies with only PC-SA also showed downregulation of IL-10 which was not the case with PC-Chol (data not shown). We have earlier observed successful downregulation of IL-10 and TGF-β along with an increase in IFN-γ levels in in vivo experiments on normal mice with only PC-SA [@pone.0017376-Banerjee2], [@pone.0017376-Banerjee3]. The exact mechanism for the IL-10 downregulation and effective generation of NO, however, remains unclear. SSG treatment is known to activate mitogen-activated protein kinase p38 (MAPK p38) and subsequently release TNF-α, which result in the production of NO in macrophages [@pone.0017376-MookerjeeBasu1]. Recent reports suggest that resistant parasites modulate the host immunity to induce unresponsiveness to SSG therapy [@pone.0017376-Haldar2]. Moreover, SSG availability in resistant parasites is less, rendering the drug ineffective in the generation of host microbicidal molecules. As reported earlier [@pone.0017376-Iwaoka1], PC-SA liposomes may also activate the MAPK p38 pathway, in addition to increasing the intracellular retention of SSG. Any drug or drug combination activating the macrophage CD40 induced MAPK p38 pathway can also boost the IL-12 mediated antiparasitic function and ameliorate *Leishmania* infection by reinstating Th1 response [@pone.0017376-Mathur1]-[@pone.0017376-Feng1]. At the doses used, it is possible that PC-SA, along with SSG induces a MAPK mediated signalling pathway, downregulating the pro-parasitic IL-10 secretion, ultimately leading to killing of the parasite and restoration of the protective immunity. Promising therapeutic effect and protection against SSG-unresponsive *L. donovani* infection can be attributed to multiple driving forces exerted by the combined activity of SSG and PC-SA leading to enhanced accumulation of SSG within parasites, direct killing of parasites induced by PC-SA and the switch of immunosuppressive humoral and cell-mediated responses to a protective Th1 type. This synergistic approach gains importance in today\'s scenario where it can help save some of the potent drugs against infectious diseases which are facing extinction due to emergence of resistance, since resistance to a combination therapy is less likely to occur. We thank Manik Saha for AAS work, Athar Alam for fluorimetry, and Soumik Basuray, K. Nasrin Nisha, Janmenjoy Middya, Smriti Mondal, Saumyabrata Mazumder, Manjarika De and Md Asad for their assistance with these studies. We thank Siddhartha Roy, Director, IICB for supporting this work. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This study was supported by grants from the Indian Council of Medical Research, and University Grants Commission, Government of India. JR was a Senior Research Fellow of Council for Scientific and Industrial Research. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: JR RS NA. Performed the experiments: JR RS. Analyzed the data: JR RS NA. Contributed reagents/materials/analysis tools: JR RS NA. Wrote the manuscript: JR RS NA. [^2]: ¤ Current address: Department of Ophthalmology and Visual Sciences, Washington University School of Medicine, St. Louis, Missouri, United States of America

PubMed Central

2024-06-05T04:04:19.788386

2011-3-10

PMC3053370

Introduction {#s1} ============ Aromatic compounds are widely distributed in ecosystems mainly released from plant materials and by anthropogenic activities. Bacteria have evolved to metabolize diverse aromatic compounds including environmental pollutants, using these compounds as carbon and energy source. A broad range of aromatic compounds is degraded by bacteria through peripheral pathways that funnel into few central pathways [@pone.0017583-Harwood1]--[@pone.0017583-Chain1]. In peripheral pathways, activation of the aromatic ring is commonly mediated by monooxygenases or dioxygenases that produce dihydroxylated intermediates. Central pathways involve the fission by a dioxygenase of the dihydroxylated aromatic metabolic intermediates at *ortho*- or *meta*-position and lead to the formation of Krebs cycle intermediates [@pone.0017583-Harwood1]--[@pone.0017583-Jimnez1]. Bacterial metabolism of aromatic compounds has been extensively studied [@pone.0017583-Chain1]--[@pone.0017583-Saavedra1]. The metabolic reconstruction of aromatic compounds pathways in in model environmental bacteria has been achieved by genome sequence analyses [@pone.0017583-Chain1], [@pone.0017583-McLeod1]--[@pone.0017583-Seeger2]. However, further studies are needed to establish gene function and to elucidate functional aromatic metabolic pathways of bacteria. *B. xenovorans* LB400 is a model bacterium for the degradation of PCBs and other aromatic compounds [@pone.0017583-Chain1], [@pone.0017583-Seeger3], [@pone.0017583-Seeger4]. Strain LB400 has one of the largest bacterial genomes (9.73 Mbp), distributed in a major chromosome (C1), a minor chromosome (C2) and a megaplasmid (MP). Genome analyses of strain LB400 revealed the presence of genes encoding eleven central pathways and twenty peripheral pathways for aromatic compounds degradation [@pone.0017583-Chain1]. The genes encoding the homogentisate and the homoprotocatechuate central pathways were identified in *B. xenovorans* LB400 genome [@pone.0017583-Chain1]. The *hmgABC* and *hpaGEDFHI* gene clusters are located at C1 and C2, respectively. Additional *hmg* gene copies were identified within LB400 genome [@pone.0017583-Chain1]. Both ring-cleavage pathways are used in aromatic amino acid metabolism in *Bacteria* and *Eucarya* [@pone.0017583-Sparnins1]--[@pone.0017583-SchmalerRipcke1]. The homogentisate pathway has been described as the central route for the degradation of L-tyrosine and L-phenylalanine in bacteria, fungi and mammals [@pone.0017583-AriasBarrau1], [@pone.0017583-SnchezAmat1], [@pone.0017583-FernndezCan1]. The homogentisate central pathway is encoded by the *hmgABC* operon in *Pseudomonas putida* strain U [@pone.0017583-AriasBarrau1]. Homogentisate (HMG) cleavage is catalysed by the HMG 1,2-dioxygenase HmgA producing maleylacetoacetate (MA). MA is isomerized by a maleylacetoacetate isomerase (HmgC) into fumarylacetoacetate, which is further converted by a fumarylacetoacetate hydrolase (HmgB) into fumarate and acetoacetate [@pone.0017583-Chapman1]. The homoprotocatechuate pathway has been described as a central route for the catabolism of aromatic amino acids and related compounds in *Klebsiella pneumoniae, P. putida*, and *Escherichia coli* [@pone.0017583-Daz1], [@pone.0017583-Cooper1]--[@pone.0017583-Gibello1]. The homoprotocatechuate pathway is encoded by the *hpaGEDFHI* gene cluster in *E. coli* strain W. Homoprotocatechuate (HPC) is *meta*-cleaved by a HPC 2,3-dioxygenase encoded by the *hpaD* gene. The product 5-carboxymethyl-2-hydroxymuconic semialdehyde (CHMS) is then converted to 5-carboxymethyl-2-hydroxy-muconic acid (CHM) and subsequently degraded by the *hpaF*, *hpaG*, *hpaH* and *hpaI* gene products into Krebs cycle intermediates [@pone.0017583-Daz1], [@pone.0017583-Sparnins1], [@pone.0017583-Prieto1], [@pone.0017583-Gibello1]. Degradation of phenylacetate hydroxylated-derivatives is channeled into the HMG or the HPC central pathways [@pone.0017583-Sparnins1], [@pone.0017583-SchmalerRipcke1], [@pone.0017583-Prieto2], [@pone.0017583-AriasBarrau2]. Assimilation of 3-hydroxyphenylacetate (3-HPA) through the homogentisate pathway in *P. putida* U has been reported. The *mhaAB* genes encode a 3-HPA 6-hydroxylase that hydroxylates 3-HPA at C-6 producing HMG [@pone.0017583-AriasBarrau1], [@pone.0017583-AriasBarrau2]. In addition, *P. putida* strain U degrades 4-hydroxyphenylacetate (4-HPA) via the homoprotocatechuate pathway [@pone.0017583-Olivera1]. *E. coli* strains W, B and C are able to degrade both 3-HPA and 4-HPA via the homoprotocatechuate central route, which is encoded by the inducible *hpaGEDFHI* operon [@pone.0017583-Daz1], [@pone.0017583-Prieto1], [@pone.0017583-Cooper1], [@pone.0017583-Roper1]. The *hpaBC* gene cluster encodes a 4-HPA 3-hydroxylase involved in the hydroxylation of 4-HPA and 3-HPA to yield HPC [@pone.0017583-Prieto1]. The *hpaBC* and *hpaGEDFHI* gene clusters constitute a single operon in *E. coli* strain W [@pone.0017583-Daz1], [@pone.0017583-Prieto1]. The aims of this study were to determine functional homogentisate and homoprotocatechuate central pathways in *B. xenovorans* strain LB400 and to establish their role in 3-HPA and 4-HPA peripheral pathways, which has not been described previously in this bacterium. This study showed that both homogentisate and homoprotocatechuate central pathways are involved in 3-HPA and 4-HPA degradation in strain LB400. Materials and Methods {#s2} ===================== Chemicals {#s2a} --------- 3-hydroxyphenylacetic acid (\>99% purity), 4-hydroxyphenylacetic acid (98% purity) and 3,4-dihydroxyphenylacetic acid (homoprotocatechuate; 98% purity) and 2,5-dihydroxyphenylacetic acid (homogentisate; 98% purity) were obtained from Sigma-Aldrich (Saint Louis, MO, USA). Bacterial strain and culture conditions {#s2b} --------------------------------------- *Burkholderia xenovorans* LB400 was cultivated at 30°C in mineral M9 medium with trace solution and glucose (5 mM) [@pone.0017583-Agull1], 3-HPA (5 mM) or 4-HPA (5 mM) as the sole carbon and energy source. Growth was determined by measuring turbidity at 525 nm and by counting colony-forming units (CFU). Aliquots taken from cultures were diluted and plated on Luria-Bertani agar medium. CFU per milliliter values were calculated as the mean ± SD of at least three independent experiments. Resting cell assays {#s2c} ------------------- Resting cells (turbidity ~525nm~ = 2.0) in 50 mM sodium phosphate buffer (pH 7.0) were incubated with HMG or HPC (0.03 mM) at 30°C. Aliquots of cell suspensions were taken at different incubation times and centrifuged (19,283 *g* for 2 min). Assays with boiled cells and without cells were used as controls. Cell-free supernatants were analyzed, using a Waters liquid chromatograph model 515 equipped with a UV detector and a RP-C18/Lichrospher 5-µm column (Supelco, Bellefonte, USA). The mobile phase contained 20% acetonitrile, 20% methanol, 60% water and 0.1% phosphoric acid. The flow rate was 0.5 mL min^-1^. HMG and HPC had retention times of 4.6 and 5.1 min, respectively. HMG and HPC were quantified using calibration curves with authentic standards. Resting cells experiments were performed in triplicate. Preparation of cell extracts {#s2d} ---------------------------- Cells harvested at exponential phase of growth (turbidity~525\ nm~ = 0.40--0.45) were centrifuged (10,733 *g* for 10 min) at 4°C and washed with 50 mM sodium phosphate buffer (pH 7.0). Bacterial cells were disrupted using an ultrasonic cell disruptor Microson™ at 4°C. Lysate was clarified by centrifugation (19,300 g for 15 min) at 4°C. Protein concentration was determined using Qubit™ fluorometer (Invitrogen). Dioxygenase activity assays {#s2e} --------------------------- HMG dioxygenase activity was determined spectrophotometrically by measuring the formation of MA at 330 nm as previously described [@pone.0017583-FernndezCan2]. The assay volume (1 ml) contained: 50 mM phosphate buffer (pH 7.0), 50 µM FeSO~4~, 2 mM ascorbate, crude extract (100 µg of protein) and 0.3 mM HMG. The reactions were carried out at 30°C and initiated by the addition of HMG. One enzyme milliunit corresponds to the transformation of 1 µmol of HMG to MA per min at 30°C under the conditions described above. HMG dioxygenase activity was calculated using the molar extinction coefficient of MA, 13,500 M^−1^ cm^−1^ [@pone.0017583-Seegmiller1]. HPC dioxygenase activity was measured by monitoring the formation of 5-carboxymethyl-2-hydroxymuconate semialdehyde (CHMS) at 380 nm [@pone.0017583-Arunachalam1]. LB400 cells cultivated using 3-HPA, 4-HPA or glucose were harvested at a turbidity~525\ nm~ of 0.45, washed and concentrated. HPC dioxygenase activity of LB400 cell suspension (turbidity~525\ nm~  = 2.0) was determined by incubation with 0.3 mM HPC at 30°C. At intervals, the A~380nm~ values of supernatants were measured. Two-dimensional polyacrylamide gel electrophoresis (2-DE) {#s2f} --------------------------------------------------------- 2-DE gels with non-equilibrium pH gradient electrophoresis (pH 3--10) were performed as previously described [@pone.0017583-Agull1], [@pone.0017583-Martnez1]. Cell cultures were grown using glucose, 3-HPA or 4-HPA (5 mM) as sole carbon source. Cells were harvested at exponential phase (turbidity~525nm~  = 0.40--0.45) by centrifugation, disrupted with a sonicator and concentrated in a speed-vac concentrator. Dried samples were resuspended in lysis buffer (9.5 M urea, 2% v/v IGEPAL CA-630, 2% ampholytes (1.6% ampholytes pH 5--7, 0.4% ampholytes pH 3--10, Bio-Rad) and 5% β-mercaptoethanol. NEPHGE gels were electrophoresed for 6.5 h at 400 V. The second dimension was performed in 11% polyacrilamide-SDS gels. Proteins were visualized with Coomassie brilliant blue R-250. The volume (Int × mm^2^) of the spots was analyzed, using Quantity One image analysis software (Bio-Rad Laboratories). To standardize the quantification of independent experiments, the ratio of the intensity of the protein to a reference protein was calculated. Protein sequencing and identification {#s2g} ------------------------------------- To perform the mass spectrometric analysis, protein spots were recovered from gels and prepared as described [@pone.0017583-Agull1]. For matrix-assisted laser desorption ionization-time of flight (MALDI-TOF) analysis, the peptide extracts were analysed with a MALDI-TOF UltraXex apparatus (Bruker, Bremen, Germany). The peptide fingerprints obtained by mass spectrometry MALDI-TOF were used for searches in the NCBI protein database with the Matrix science MASCOT search tool. The complete sequences of each protein were obtained and BLAST searches were executed with the FASTA tool for the identification and similarity data analyses. Isolation of total RNA and RT-PCR {#s2h} --------------------------------- Total RNA was isolated from LB400 cells in the stationary growth phase using an RNeasy mini kit (Qiagen, Hilden, Germany) according to the manufacturers\' recommendations. DNase I treatment was carried out using the RNase-Free DNase Set (Qiagen, Hilden, Germany) to degrade any residual DNA. The RNA concentration was quantified using a Qubit™ fluorometer (Invitrogen, Carlsbad, CA, USA). Reverse transcription-PCR (RT-PCR) was carried out with sequence-specific primers design in this study for *hmgA1* (BxeA2725), *hmgA2* (BxeA3900) and *hpaD* (BxeB2031) genes by using SuperScript™ One-step RT-PCR with Platinum®*Taq* (Invitrogen, Carlsbad, CA, USA). The BxeA2725 gene (*hmgA1*) was amplified using the primers HmgA1-f (5′-ATTTGCGACCGAAACGCTGCC-3′) and HmgA1-r (5′-CGACGGGTTGAAGTGCTTCC-3′). Amplification of the BxeA3900 gene (*hmgA2*) was performed with specific primers HmgA2-f (5′-TACGCCGAACACTGTCAGTTCG-3′) and HmgA2-r (5′-GCCCCGAAATTCTCGCAGATGTA-3′). Reverse transcriptase reaction for *hpaD* mRNA was performed using the primers HpaD-f (5′-CTCGCGTTGGCAGCAAAAGTG-3′) and HpaD-r (5′-GCAGAGGCGTAACCGGAAAG-3′). Amplification of the 16S rRNA gene was used as control for DNA contamination using the primers 27f (5′-AGAGTTTGATCMTGGCTCAG-3′) and 1492r (5′-TACGGYTAC CTTGTTACGACTT-3′) [@pone.0017583-Weisburg1]. Results {#s3} ======= Analysis of *hmg* and *hpa* gene clusters in strain LB400 {#s3a} --------------------------------------------------------- The *hmgABC* gene cluster (BxeA2725; BxeA2724 and BxeA2723; hereafter *hmgA1*, *hmgB1* and *hmgC1*, respectively) is located at C1 of strain LB400 genome. Copies of the *hmgA1* gene and *hmgB1* genes (BxeA3900 and BxeA3899, respectively) were identified in a 1,300 kb distant region from *hmgA1B1C1* cluster at C1 (hereafter *hmgA2* and *hmgB2*). A *hmgC* gene copy (BxeA4141; hereafter *hmgC2*) is located in a different region of C1 [@pone.0017583-Chain1] ([Fig. 1A](#pone-0017583-g001){ref-type="fig"}). Sequence analyses of the chromosomal region flanking the *hmgA1B1C1* gene cluster in the strain LB400, revealed that the products of the open reading frames (ORFs) BxeA2727 and BxeA2726 shared 47% and 35% sequences identity with the MhaA and MhaB proteins from *P. putida* strain U, respectively [@pone.0017583-AriasBarrau2] ([Fig. 1A](#pone-0017583-g001){ref-type="fig"}). In *P. putida* U the enzyme 3-HPA 6-hydroxylase (MhaAB) is a two- component FAD-dependent monooxygenase that hydroxylates 3-HPA to produce HMG ([Fig. 2](#pone-0017583-g002){ref-type="fig"}). The large component MhaA is a flavoprotein and the small component MhaB is a coupling protein [@pone.0017583-AriasBarrau2]. In this study, the presence of the *mhaAB* genes located adjacent of *hmgA1B1C1* gene cluster in LB400 was revealed, suggesting a role of *mhaAB* genes in 3-HPA peripheral reaction leading to HMG. ::: {#pone-0017583-g001 .fig} 10.1371/journal.pone.0017583.g001 Figure 1 ::: {.caption} ###### Organization of genes encoding the homogentisate and homoprotocatechuate central pathways and related peripheral pathways in *B. xenovorans* LB400. A, The *hmg* genes encoding the homogentisate central pathway, and *mha* genes encoding a 3-HPA peripheral pathway located in the major chromosome (C1). B, The *hpa* cluster encoding the homoprotocatechuate central pathway, and *hpaBC* genes encoding a 4-HPA peripheral pathway located in the minor chromosome (C2) of strain LB400. Genes encoding ring-cleavage dioxygenases are indicated with black arrows. :::  ::: ::: {#pone-0017583-g002 .fig} 10.1371/journal.pone.0017583.g002 Figure 2 ::: {.caption} ###### Phylogenetic tree showing the relatedness of FAD-dependent monooxygenases. The dendogram was constructed by the neighbor-joining method using MEGA 4.1 based on sequence alignments calculated by Clustal W. Sequences of deduced proteins from BxeA2727 and BxeB2308 genes from *B. xenovorans* L400 are highlighted (black circles). Proteins and their accession numbers are: HpaB (Bxe\_B2308) *B. xenovorans* LB400 (YP\_553029); HpaB *E. coli* (CAA82321); HpaB *K. pneumoniae,* (AAC37120); HpaB *T. thermophilus* (BAD70783); HpaB *Geobacillus sp.* PA-9 (AAT28189); HpaB *P. fluorescens* Pf-5 (YP\_260461); HpaB *Y. pestis* (YP\_651053); HpaB *Rhizobium sp.* NGR234, (YP\_002824209); HpaB *P. aeruginosa* PAO1 (NP\_252780); HpaB *R. jostii* RHA1, (YP\_701753); HpaB *B. ambifaria* IOP40-10 (ZP\_02892641); HpaB *B. pseudomallei* MSHR346 (ZP\_04521853); HpaB *C. necator* JMP134 (YP\_293474); HpaB *B. cenocepacia* HI2424 (YP\_837331); MhaA (BxeA2727) *B. xenovorans* LB400, (ABE30237); MhaA *P. putida* U (AAY16572); MhaA *C. necator* JMP134 (AAZ64667); MhpA *K. pneumoniae* 342 (YP\_002238037); MhpA *B. cenocepacia* HI2424 (YP\_840344), MhpA *C. necator* JMP134 (YP\_294496); MhpA *B. xenovorans* LB400 (YP\_553009); PheA *Pseudomonas sp.* EST1001 (AAC64901); PheA1 *R. jostii* RHA1 (YP\_702477); TbuD *R. pickettii* (AAA25992); PobB *B. xenovorans* LB400 (ABE30926); PobA *P. putida* KT2440 (AAN69138); PobA *P. fluorescens* (CAA48483); Reut\_B5020 *C. necator* JMP134 (AAZ64368); PobA *Rhizobium sp.* NGR234 (YP\_002823464); PobA *Ruegeria sp.* R11 (ZP\_05090233); PobA *R. jostii* RHA1 (YP\_702502); PobA *K. pneumoniae* 342 (YP\_002240587); SalA *B. xenovorans* LB400 (ABE36893); NahG *P. putida* PpG7 (AAA25897); SalA *P. reinekei* (ABH07020); SalA *C. necator* JMP134 (YP\_296720); SalA *R. jostii* RHA1 (YP\_701838). :::  ::: The gene cluster *hpaG1G2EDFHI* encoding the homoprotocatechuate pathway is located at C2 in strain LB400 genome (BxeB2028, BxeB2029, BxeB2030, BxeB2031, BxeB2032, BxeB2033 and BxeB2034) ([Fig. 1B](#pone-0017583-g001){ref-type="fig"}). Downstream and adjacent to the *hpa* gene cluster, a divergent gene putatively encoding the HpaR repressor protein of the HPC catabolic pathway was identified ([Fig. 1B](#pone-0017583-g001){ref-type="fig"}) [@pone.0017583-Chain1]. Genes encoding the 4-HPA catabolic pathway were not found in the neighborhood of *hpa* gene cluster in strain LB400. The degradation of 4-HPA in *E. coli* strain W involves enzymes encoded in the 4-HPA hydroxylase *hpaBC* operon and the *meta*-cleavage operon *hpaGEDFHI* [@pone.0017583-Prieto1]. The *hpaBC* cluster of strain W encodes a two-component protein responsible for 4-HPA hydroxylation to yield HPC [@pone.0017583-Prieto1], [@pone.0017583-Prieto2]. The BxeB2309 gene product of strain LB400 was identified as the 4-HPA 3-monooxygenase coupling protein HpaC located downstream to the *hpa* cluster (at 309 kb). The protein encoded by the BxeB2308 gene is highly related to the HpaB proteins from *B. cenocepacia* strain HI2424 and *R. jostii* RHA1 ([Fig. 2](#pone-0017583-g002){ref-type="fig"}). Based on its relatedness with FAD-dependent monooxygenases and its location next to a putative *hpaC* gene, the gene BxeB2308 may encode the main component of the closely related 4-HPA 3-monooxygenase (HpaB). However, further analyses are required to establish the 4-HPA 3-monooxygenase peripheral pathway encoded by BxeB2308 and BxeB2309 genes in LB400. The presence of genes putatively encoding peripheral reactions leading the homogentisate and homoprotocatechuate central pathways suggested functional HPA peripheral pathways leading to these ring-cleavage pathways in strain LB400. Strain LB400 growth on HPAs {#s3b} --------------------------- Sequence analyses of strain LB400 genome revealed the presence of putative genes encoding 3-HPA and 4-HPA peripheral pathways. To this end, the growth of strain LB400 on 3-HPA and 4-HPA as sole carbon source was studied. Strain LB400 was able to grow using 3-HPA and 4-HPA as sole carbon and energy source reaching a high biomass (data not shown). Strain LB400 attained at stationary phase (25 h), slightly higher biomass using 4-HPA (3.21× FU 10^7^ mL^−1^) than using 3-HPA (1.28 1× CFU 10^7^ mL^−1^) and glucose (2.09× CFU 10^7^ mL^−1^) as sole carbon source. The growth on HPAs indicate that strain LB400 possess functional 3-HPA and 4-HPA catabolic pathways. HMG and HPC degradation by strain LB400 {#s3c} --------------------------------------- In order to determine the functionality of the homogentisate and homoprotocatechuate central pathways, the degradation of these compounds was analysed. HMG and HPC degradation by strain LB400 were studied using resting cell assays. LB400 cells grown either in 3-HPA or 4-HPA were able to degrade HMG and HPC compounds. Cells grown in 3-HPA and 4-HPA degraded after 10 h of incubation 100% and \>95% of HMG, respectively ([Fig. 3A](#pone-0017583-g003){ref-type="fig"}). It is worth noting that, during HMG degradation assays, a brown colored medium was only observed in control assays with boiled cells. In contrast, no pigmented medium was observed in assays with metabolically active cells. The brown pigmentation of the medium has been reported in bacteria and fungi that lack the homogentisate pathway, and thus HMG accumulation and its chemical oxidation were observed [@pone.0017583-AriasBarrau1], [@pone.0017583-SchmalerRipcke1], [@pone.0017583-RodrguezRojas1]. LB400 cells grown in 3-HPA and 4-HPA degraded after 22 h almost completely HPC ([Fig. 3B](#pone-0017583-g003){ref-type="fig"}). A slightly faster degradation of HPC was observed with 3-HPA-grown cells than with 4-HPA-grown cells. These results showed that strain LB400 degraded HMG and HPC, indicating that the corresponding ring-cleavage pathways are active in 3-HPA- and 4-HPA-grown cells and are involved in their degradation. ::: {#pone-0017583-g003 .fig} 10.1371/journal.pone.0017583.g003 Figure 3 ::: {.caption} ###### Degradation of HMG and HPC by resting cells of *B. xenovorans* LB400. LB400 cells were grown in 3-HPA (squares) or 4-HPA (triangles) as sole carbon source and incubated with HMG (0.03 mM) (A) or HPC (0.03 mM) (B). Control assays with boiled cells showed no degradation of HMG and HPC (data not shown). Each point is an average of results from at least three independent assays. :::  ::: HMG and HPC ring-cleavage activities {#s3d} ------------------------------------ To further determine functional homogentisate and homoprotocatechuate central pathways, the enzymatic activities of HMG dioxygenase and HPC dioxygenase were determined. HMG dioxygenase activity was measured in crude extracts of LB400 cells cultured in 3-HPA, 4-HPA and glucose. A high HMG dioxygenase activity was observed during growth on 3-HPA (82 mU/mg protein) and 4-HPA (75 mU/mg protein) ([Fig. 4A](#pone-0017583-g004){ref-type="fig"}). HMG dioxygenase activity was slightly higher on 3-HPA-grown cells than on 4-HPA-grown cells, whereas no activity was observed in glucose-grown cells. HPC dioxygenase activity was measured by monitoring the formation of CHMS after whole cells incubation with HPC. Cells grown on 3-HPA and 4-HPA showed high HPC dioxygenase activity. A moderately higher HPC dioxygenase activity was observed in 4-HPA-grown cells than in 3-HPA-grown cells. In contrast, a low HPC dioxygenase activity was measured in glucose grown cells ([Fig. 4B](#pone-0017583-g004){ref-type="fig"}). These results indicate an induction of the key enzymes of the homogentisate and homprotocatechuate central pathways during growth of strain LB400 on 3-HPA and 4-HPA. ::: {#pone-0017583-g004 .fig} 10.1371/journal.pone.0017583.g004 Figure 4 ::: {.caption} ###### Homogentisate and homoprotocatechuate ring-cleavage dioxygenase activities of *B. xenovorans* LB400. Cells were grown in minimal medium using 3-HPA (squares), 4-HPA (triangles) and glucose (empty circles) as sole carbon source. A, HMG dioxygenase activity measured by maleylacetoacetate (MA) formation in crude extracts; B, HPC dioxygenase activity measured 5-carboxymethyl-2-hydroxy-muconic semialdehyde (CHMS) product formation in resting cells of strain LB400. :::  ::: *hmgA* and *hpaD* expression analyses by RT-PCR {#s3e} ----------------------------------------------- Further analyses were conducted in order to determine the role of the homogentisate and homoprotocatechuate pathways in 3-HPA and 4-HPA catabolism. The expression of *hmgA* and *hpaD* genes encoding the key enzymes HMG and HPC dioxygenases was analyzed during exponential growth of strain LB400 in 3-HPA and 4-HPA. Additionally, transcriptional analyses of two gene copies encoding HMG 1,2-dioxygenase were performed using specific primers for *hmgA1* and *hmgA2* genes. Expression of the *hmgA1* gene was observed at early exponential growth phase of strain LB400 in 3-HPA, whereas no expression was detected during growth on 4-HPA or glucose ([Fig. 5A](#pone-0017583-g005){ref-type="fig"}). This indicates that *hmgA1* gene is up regulated by 3-HPA in strain LB400. On the other hand, expression of the *hmgA2* gene was detected in 3-HPA, 4-HPA and glucose grown-cells. These data also suggests lower *hmgA2* transcription in glucose-grown cells ([Fig. 5A](#pone-0017583-g005){ref-type="fig"}). ::: {#pone-0017583-g005 .fig} 10.1371/journal.pone.0017583.g005 Figure 5 ::: {.caption} ###### Expression of *hmgA* and *hpaD* genes during growth on HPAs. LB400 cells were grown on glucose (lanes 1) 3-HPA (lanes 2) and 4-HPA (lanes 3) as sole carbon source. A, Expression of *hmgA1* (BxeA2725); B, expression of *hmgA2* (BxeA3900); C, expression of *hpaD*. RT-PCR assays were performed using RNA from LB400 cells collected at early exponential growth phase. :::  ::: Expression analysis of *hpaD* gene encoding HPC 2,3-dioxygenase was performed. The *hpaD* transcripts were detected at early exponential phase of strain LB400 during growth on 3-HPA, 4-HPA and glucose as sole carbon source ([Fig. 5B](#pone-0017583-g005){ref-type="fig"}). Higher *hpaD* expression during growth on 3-HPA and 4-HPA was observed. These results indicate an expression of the dioxygenase-encoding genes during 3-HPA and 4-HPA catabolism in strain LB400. Furthermore, these results indicate that during 3-HPA degradation, two HMG 1,2-dioxygenase encoding *hmgA* genes are expressed. Proteomic analyses during 3-HPA and 4-HPA degradation {#s3f} ----------------------------------------------------- A proteomic analysis of LB400 cells grown in 3-HPA and 4-HPA as sole carbon source was performed in order to elucidate catabolic enzymes involved in HPAs degradation. 2-DE analysis revealed the induction of an enzyme from the homogentisate central pathway during growth of strain LB400 in 3-HPA and 4-HPA ([Fig. 6](#pone-0017583-g006){ref-type="fig"}). The polypeptide was identified by MALDI-TOF as fumarylacetoacetate hydrolase HmgB, which catalyzes the conversion of fumarylacetoacetate into fumarate and acetoacetate. The protein sequence possesses 91% identity with HmgB protein from *Burkholderia phytofirmans* PsJN. The induction of HmgB during growth on 3-HPA, was 3.5-fold compared to glucose-grown cells. During growth on 4-HPA, induction of this protein was 2.5-fold compared to cells grown on glucose. The corresponding *hmgB2* gene (BxeA3899) is adjacent to *hmgA2* in strain LB400 genome. This result correlates with the expression of *hmgA2* gene observed under the same growth conditions by RT-PCR. The induction of an enzyme belonging to the homogentisate ring-cleavage pathway is in accordance with the degradation of HMG and the increased HMG dioxygenase activity in strain LB400 during growth in 3-HPA and 4-HPA. Inducible expression of two enzymes from the homogentisate pathway observed during growth of strain LB400 in HPAs indicate that this catabolic pathway is involved in 3-HPA and 4-HPA degradation by strain LB400. The lower synthesis of HmgB protein in *B. xenovorans* LB400 grown in glucose indicates a basal expression of the catabolic *hmgB1* gene during growth on this carbon source. ::: {#pone-0017583-g006 .fig} 10.1371/journal.pone.0017583.g006 Figure 6 ::: {.caption} ###### Induction of fumarylacetoacetate hydrolase HmgB in *B. xenovorans* LB400 during growth on HPAs. Cells were grown on glucose 5 mM (A), 3-HPA 5 mM (B) and 4HPA 5 mM (C). Proteins were separated by 2-D gel electrophoresis and stained with Coomassie blue. A segment of each 2-D gel is shown. :::  ::: Discussion {#s4} ========== This study has shown that the homogentisate and the homoprotocatechuate central pathways are involved in 3-HPA and 4-HPA catabolism by *B. xenovorans* strain LB400. 3-HPA and 4-HPA isomers are used by *B. xenovorans* LB400 as sole carbon and energy source for growth, indicating active peripheral and central catabolic pathways. Bacterial catabolism of hydroxyphenylacetates, particularly 3-HPA and 4-HPA compounds have been described to channel either the homogentisate or the homoprotocatechuate central pathways. Interestingly, this report showed that LB400 growth on 3-HPA and 4-HPA separately funnels into both central pathways. Similarly, in *P. putida* strain U, different catabolic pathways were reported for 4-HPA and 3-HPA degradation. It is worth noting that strain LB400 showed a faster growth in 3-HPA (5 mM) than *P. putida* strain U in 3-HPA (10 mM) [@pone.0017583-AriasBarrau2]. In this study, strain LB400 reached high biomass using 4-HPA as sole carbon source (turbidity~525\ nm~ of 1.1) (data not shown). In contrast, in similar media with 4-HPA as carbon source *P. putida* U attains lower biomass (turbidity~540\ nm~ of 0.3) [@pone.0017583-Olivera1]. The MhaAB protein from strain U is highly specific for 3-HPA substrate to yield HMG [@pone.0017583-AriasBarrau2], which is further degraded by the *hmgABC* gene products [@pone.0017583-AriasBarrau1], [@pone.0017583-AriasBarrau2]. Interestingly, upstream to the *hmgA1B1C1* cluster in the LB400 genome, two ORFs (BxeA2727 and BxeA2726) with high identity to the *mhaAB* genes from strain U were found. The MhaA flavoprotein component (566 aa) and the MhaB coupling protein (60 aa) possess a similar length to the aminoacid sequence deduced from BxeA2727 (560 aa) and BxeA2726 (60 aa) genes of strain LB400. In a previous study, the BxeA2727 was annotated as a putative 3-(3-hydroxy-phenyl)propionate hydroxylase (MhpA) [@pone.0017583-Chain1]. We propose that BxeA2727 and BxeA2726 of strain LB400 encode the two components of the hydroxylating enzyme MhaAB that is involved in catabolism of 3-HPA via HMG ([Fig. 7A](#pone-0017583-g007){ref-type="fig"}). However, further analyses are required to confirm the involvement of the *mhaAB* genes in 3-HPA degradation by strain LB400. ::: {#pone-0017583-g007 .fig} 10.1371/journal.pone.0017583.g007 Figure 7 ::: {.caption} ###### Model of 3-HPA and 4-HPA catabolic pathways in *B. xenovorans* LB400. Proposed main routes for hydroxyphenylacetates oxidation in LB400 are indicated with thick arrows, whereas the not preferentially oxidation route is indicated with a thin arrow. A, Hydroxyphenylacetates catabolism via *ortho*-cleavage pathway (continuous line). The substrates and products are: 3-HPA (3-hydroxyphenylacetate); 4-HPA (4-hydroxyphenylacetate); HMG (homogentisate); MA (maleylacetoacetate) and FA (fumarylacetoacetate). The enzymes are MhaAB (3-HPA 6-hydroxylase); HmgA (HMG 1,2-dioxygenase); HmgC (maleylacetate isomerase), HmgB (fumarylacetate hydrolase). B, Hydroxyphenylacetates catabolism via *meta*-cleavage pathway (dotted line). The metabolites are: HPC (homoprotocatechuate), CHMS (5-carboxymethyl-2-hydroxy-muconic semialdehyde), CHM (5-carboxymethyl-2-hydroxy-muconic acid), OPET (5-oxo-pent-3-ene-1,2,5-tricarboxylic acid), HHDD (2-hydroxy-hept-2,4-diene-1,7-dioic acid), OHED (2-oxo-hept-3-ene-1,7-dioic acid), and HHED (2,4-dihydroxy-hept-2-ene-1,7-dioic acid). The enzymes are HpaBC (4-HPA monooxygenase), HpaD (HPC 2,3-dioxygenase), HpaE (CHMS dehydrogenase), HpaF (CHM isomerase), HpaG (OPET decarboxylase), HpaH (OHED hydratase), HpaI (HHED aldolase), and Sad (succinic semialdehyde dehydrogenase). We propose that in strain LB400 3-HPA is preferentially channeled into the homogentisate central pathway, whereas 4-HPA is channeled actively in both homogentisate and homoprotocatechuate central pathways. :::  ::: Alternatively, 4-HPA degradation in *P. putida* strain U is mediated by a 4-HPA 3-hydroxylase yielding HPC [@pone.0017583-Olivera1]. Interestingly, the results presented above indicate that both 4-HPA and 3-HPA compounds are degraded via HPC in strain LB400. However, based on *hmgA* and *hpaD* gene expression analyses and HMG and HPC dioxygenase activities, we postulate that 3-HPA in strain LB400 is preferentially channelled into the homogentisate central pathway. A previous report of 4-HPA catabolism showed that 4-HPA 3-hydroxylase from *E. coli* strain W hydroxylates 4-HPA and also 3-HPA yielding HPC [@pone.0017583-Prieto2]. A search of *hpaB* and *hpaC* genes in LB400 genome showed a gene (BxeB2309) with high identity to the *hpaC* gene encoding the 4-HPA 3-hydroxylase coupling protein of *E. coli* strain W located 309 b distant from the *hpaGEDFHI* gene cluster in strain LB400. The adjacent gene (BxeB2308) probably encodes a FAD-dependent monooxygenase distantly related to the HpaB hydroxylase component from *E. coli* W [@pone.0017583-Prieto2]. We propose that the BxeB2308 and BxeB2309 genes encode the two components of a hydroxylase involved in 3-HPA and 4-HPA catabolism ([Fig. 7B](#pone-0017583-g007){ref-type="fig"}). Additionally, this study demonstrated that both chromosomally encoded homogentisate and homoprotocatechuate central pathways are functional in strain LB400. Resting cell assays indicated that strain LB400 degraded HMG and HPC compounds. Moreover, a high HMG dioxygenase activity was measured in 3-HPA and 4-HPA-grown cells (82 and 75 mU/mg of protein, respectively), whereas no activity was detected in glucose-grown cells ([Fig. 4A](#pone-0017583-g004){ref-type="fig"}). A similar HMG dioxygenase activity (68 mU/mg of protein) was previously reported in crude extracts of *P. putida* U cells grown on tyrosine or phenylalanine [@pone.0017583-AriasBarrau1]. LB400 cells grown on 3-HPA and 4-HPA showed high HPC dioxygenase activities, whereas a lower activity was measured in glucose-grown cells ([Fig. 4B](#pone-0017583-g004){ref-type="fig"}). These results indicate that: i) key ring-cleavage dioxygenases of the homogentisate and the homoprotocatechuate pathways are functional, and ii) the homogentisate and the homoprotocatechuate central pathways are induced in strain LB400 during growth on 3-HPA and 4-HPA. During growth of strain LB400 on 3-HPA, expression of the *hmgA1* gene (BxeA2725) and *hmgA2* gene (BxeA3900) were observed by RT-PCR. The *hmgA1* gene is part of the *hmgABC* gene cluster in LB400, whereas the *hmgA2* gene is clustered only with *hmgB2* gene. In contrast, during 4-HPA degradation, the expression of *hmgA2* gene but not of *hmgA1* gene was observed. The results indicate that: i) the *hmgA1B1C1* cluster of strain LB400 is up regulated during 3-HPA metabolism but not during 4-HPA metabolism and, ii), two *hmgA* gene copies are transcribed during growth on 3-HPA. It is worth noting that these two genes have not identical DNA sequence. It is likely that one of the *hmgA* genes was acquired via horizontal gene transfer, which has been an important source of genes in strain LB400 [@pone.0017583-Pieper1]. These data indicates that growth on 3-HPA and 4-HPA are mediated by different cellular enzymatic components. During degradation of 3-HPA, it is likely that two *hmgA* gene copies are used, suggesting that enhanced catabolic abilities are required. Probably, the homogentisate pathway in strain LB400 is regulated at the level of HMG dioxygenase to prevent accumulation of catabolic intermediates that may have toxic effects on the cell. Expression of two gene copies encoding key enzymes may enhance catabolic machinery of strain LB400. The presence of multiples *hmg* gene copies spread in LB400 genome probably enhance its catabolic capabilities. The *hmgAB* arrangement, not linked with *hmgC* gene has also been observed in *Silicibacter pomeroyi*, *Pseudomonas syringae*, *Ralstonia solanacearum*, *Bordetella bronchiseptica* and *Cupriavidus necator* [@pone.0017583-PrezPantoja1], [@pone.0017583-AriasBarrau1]. The use of the homogentisate pathway encoded by the *hmgAB* gene cluster has also proposed during growth on 3-HPA and 4-HPA in *C. necator* strain JMP134; in this strain the homoprotocatechuate pathway is absent [@pone.0017583-PrezPantoja1]. The presence of multiples genes copies encoding chlorocatechol 1,2-dioxygenase in *C. necator* JMP134 are required for efficient chlorocatechol degradation during growth of strain JMP134 on 3-chlorobenzoate [@pone.0017583-PrezPantoja2]. Enhanced degradation avoids accumulation of toxic metabolic intermediates such as catechols [@pone.0017583-Cmara1], [@pone.0017583-Schweigert1]. The toxicity of aromatic compounds and their metabolic intermediates for strain LB400 has been reported [@pone.0017583-Cmara2], [@pone.0017583-Agull1], [@pone.0017583-Martnez1]. In addition to the expression of *hmgA* genes, the *hpaD* gene encoding HPC dioxygenase was expressed in LB400 during growth on 3-HPA and 4-HPA. A lower *hpaD* expression on glucose was observed. This in accordance with high HPC dioxygenase observed during growth in 3-HPA and 4-HPA and a lower activity observed in glucose-grown cells. In *Gammaproteobacteria*, 4-HPA is commonly degraded through the homoprotocatechuate pathway. 4-HPA degradation via HPC has been studied in some *E. coli* strains such as strain W that lacks the homogentisate pathway [@pone.0017583-Daz1], [@pone.0017583-Prieto1], [@pone.0017583-Prieto2]. As mentioned above, the 4-HPA hydroxylase from *E. coli* strain W is able to hydroxylate a wide range of substrates, including 3-HPA [@pone.0017583-Prieto2]. Thus, we propose that the homoprotocatechuate pathway is an additional convergent route involved in 3-HPA and 4-HPA degradation in *B. xenovorans* LB400. Proteomic analyses of LB400 cells grown on 3-HPA and 4-HPA provided additional evidence that the homogentisate central pathway is involved in the catabolism of 3-HPA and 4-HPA. The induction of fumarylacetoacetate hydrolase HmgB by 3-HPA and 4-HPA was observed. The HmgB protein was induced in 3-HPA-grown cells (3.5-fold) and in 4-HPA-grown cells (2.5-fold) compared to glucose-grown cells. As suggested above, 3-HPA hydroxylation mediated by the *mhaAB* gene products from strain LB400 is in accordance with 3-HPA catabolism via HMG revealed by proteomic analyses. 4-HPA degradation also has been reported to funnel into the homogentisate route, by hydroxylation at carbon 1 in *Pseudomonas acidovorans* and *Azoarcus evansii* [@pone.0017583-Hareland1], [@pone.0017583-MohamedMel1]. Therefore, it is likely that 4-HPA is metabolized through the homogentisate pathway in LB400, in accordance with HmgB2 induction during 4-HPA metabolism observed by 2D-PAGE. The *hmgB2* gene is adjacent to the *hmgA2* gene (BxeA3900) in C1 of strain LB400. This is in accordance with *hmgA2* expression observed during growth in 3-HPA and 4-HPA. The induction of HmgB protein suggests an up-regulation of the homogentisate pathway during 3-HPA and 4-HPA degradation in strain LB400. Moreover, HMG dioxygenase activity measured in crude extracts of strain LB400 suggested also induction of *hmg* catabolic genes. The HmgR regulator protein from *P. putida* strain U is an IclR-type regulator divergently transcribed from *hmgABC* cluster and acts as a repressor of *hmg* genes [@pone.0017583-AriasBarrau1]. However, a gene similar to the HmgR protein from *P. putida* U was not identified in the *hmgA1B1C1* cluster neighborhood and other regions of the genome of strain LB400. The transcription analyses suggested that expression of the homogentisate and homoprotocatechuate pathways are controlled at transcriptional level in LB400, in which the presence of specific substrates such as 3-HPA and 4-HPA are required. A gene encoding a HpaR regulator is located adjacent to the *hpaGEDFHI* gene cluster in LB400 genome. It has been determined that the HpaR protein of *E. coli* strain C regulates transcription of the *hpaGEDFHI* cluster and it is induced by 4-HPA or HPC [@pone.0017583-Roper1]. It is likely that 4-HPA, 3-HPA or metabolic intermediates such as HPC may regulate expression of *hpa* cluster in LB400 mediated by HpaR ([Fig. 1B](#pone-0017583-g001){ref-type="fig"}). Genome characterization of *B. xenovorans* LB400 revealed interesting catabolic capabilities. In this work, we provided evidence on the functionality of the two chromosomally encoded homogentisate and homoprotocatechuate central pathways. In addition, it was shown that the 3-HPA and 4-HPA catabolic peripheral pathways are active. Moreover, two functional *hmgA* genes were used by LB400 during 3-HPA catabolism. Based on growth and degradation assays, gene expression analyses, protein synthesis and key dioxygenases activities, we postulate that in strain LB400 3-HPA and 4-HPA separately funnel into both homogentisate and homoprotocatechuate central pathways. Probably, 3-HPA is preferentially channelled into the homogentisate central pathway, whereas 4-HPA is channelled actively in both central pathways. This study reveals the wide aromatic catabolic repertoire of strain LB400, linking the abilities predicted *in silico* with those observed at functional level. Furthermore, the utilization of both *ortho*- and *meta*-cleavage pathways in the degradation of two isomers reflects the amazing metabolic plasticity of this bacterium. The authors thank Bernd Hofer for helpful discussions and Manfred Nimtz for protein sequencing. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**M.S. gratefully acknowledges financial support from FONDECYT (1070507, 7090079 and 1020221) (<http://www.fondecyt.cl>), MILENIO P04/007-F (MIDEPLAN) (<http://www.iniciativamilenio.cl>) Center for Nanotechnology and Systems Biology (<http://www.usm.cl>) and USM (130836, 130948) (<http://www.usm.cl>) grants. The funders had no role in study design, data collection and analyses, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: VM LA MS. Performed the experiments: VM MG. Analyzed the data: VM LA MG MS. Contributed reagents/materials/analysis tools: MG MS. Wrote the paper: VM MS.

PubMed Central

2024-06-05T04:04:19.791777

2011-3-10

PMC3053371

Introduction {#s1} ============ The prediction of a protein\'s function is one of the most valuable contributions of bioinformatics to biological research. Next to providing functional prediction for experimental design, the functional annotation of entire proteomes is nowadays a basic task of genome database providers. Among the most used resources for functional annotations are conserved domains, which are distinct structural and functional units of a protein [@pone.0017568-Lawrence1]. In general, family members of conserved domains are collected and deposited in profile databases such as Pfam, SMART or CDD [@pone.0017568-MarchlerBauer1], [@pone.0017568-Finn1], [@pone.0017568-Letunic1]. These databases can be searched by a number of different algorithms including Hidden Markov Models (HMMs) [@pone.0017568-Eddy1], RPS-BLAST [@pone.0017568-MarchlerBauer1] or Pattern Matching [@pone.0017568-Gattiker1]. Although these methods work very well when sufficient sequence similarity is present, they tend to miss more divergent family members, which lie within and below the so-called twilight zone of below 20% sequence similarity. This is in many cases the result of a lack of divergent members in domain profiles resulting in profile definitions that are too strict. Consequently, in automated conserved domain searches that are applied to entire proteomes, sensitivity has to be sacrificed for the benefit of reliable predictions. When proteins are analyzed manually, even more sensitive methods can be applied. PSI-BLAST searches [@pone.0017568-Altschul1], for instance, which use a profile of homologs as input to iterative database searches, as well as the detection of divergent superfamily- or conserved domain- members using profile-profile comparisons [@pone.0017568-Hofmann1], [@pone.0017568-Soding1] can greatly enhance the sensitivity and therefore provide new or additional information to functional predictions of individual proteins. The HHPred-server [@pone.0017568-Soding1] as an example allows the user to build a profile of an input sequence and performs profile-profile comparisons to conserved domain databases or profile resources of fold classes like SCOP [@pone.0017568-Murzin1] or CATH [@pone.0017568-Orengo1]. HHPred works extremely well for the detection of remote sequence similarity [@pone.0017568-Hildebrand1], yet it has not been adapted for genome-scale searches. The Superfamily database [@pone.0017568-Gough1], as another example, uses sensitive profile-based searches (SAM-T99 HMM [@pone.0017568-Karplus1]) to provide structural annotation of genomes based on SCOP families. As several, overlapping profiles are used to represent a single SCOP family and SAM-T99 HMM is used that is specifically strong in detecting remote sequence similarity, this method is able to detect remote homologs to known structural families [@pone.0017568-Apic1]. In the initial version, the main information provided was the predicted structural fold of a protein sequence. Meanwhile, extensive functional information is provided in addition (based on Gene Ontology), thereby making the database useful for functional classification of proteins based on fold classes [@pone.0017568-Wilson1]. Alternative to sequence-based searches, structural features have been recognized as useful in detecting remote sequence similarity. As the function of a protein is to a great extent defined by its structure, its fold is generally better conserved than its sequence [@pone.0017568-Chothia1], [@pone.0017568-Goldstein1]. Fold recognition (also known as threading) approaches have proven most successful in this area [@pone.0017568-Mooney1], [@pone.0017568-Ashburner1]. In most fold recognition applications, a protein sequence is compared to a set of three-dimensional protein structures and is scored based on statistical approaches [@pone.0017568-Chothia1], [@pone.0017568-Goldstein1], [@pone.0017568-Sippl1], [@pone.0017568-Jones1], [@pone.0017568-Bauer1]. Though fold recognition techniques are constantly improving, it is nevertheless still difficult to reliably score a threading hit. For many approaches, one sequence can produce multiple threading hits that are hard to distinguish by score alone [@pone.0017568-Moult1]. One common approach is to include information from other sources like putative biological function to reliably determine true positive hits. Threading tools like 3D-PSSM [@pone.0017568-Kelley1] (now Phyre [@pone.0017568-Kelley2]), TASSER [@pone.0017568-Zhang1], which uses PROSPECTOR [@pone.0017568-Skolnick1], [@pone.0017568-Skolnick2] derivatives as the threading engine, or MUSTER [@pone.0017568-Wu1] all take into account sequence-, as well as structural (secondary and/or tertiary) features when scoring hits and therefore outperform purely structure-based techniques. The SAMD method as another example, utilizes neural networks together with predicted structural properties to predict structural folds within the twilight zone [@pone.0017568-Mooney1]. However, none of the above described methods except for Threader [@pone.0017568-McGuffin1] have been used systematically for genome-wide annotations. Here we introduce the genome-wide HMMerThread resource of remotely conserved domains. Based on a much improved HMMerThread algorithm we have previously published [@pone.0017568-Bradshaw1], we have predicted remotely conserved domains at a proteome-wide level for eight model organisms including human. Through a combination of relaxed conserved domain database searches with subsequent fold recognition steps to eliminate false positive predictions due to high E-value settings, we provide accurate predictions of conserved domains that are well within and beyond the twilight zone of sequence similarity. We use orthology information, as well as information on key functional residues, if available, to validate remotely conserved domains. Our pipeline has achieved an accuracy of 90%, making HMMerThread an accurate application to detect remotely conserved domains. We provide genome-wide data on remotely conserved HMMerThread domains in a relational database, which is openly accessible at <http://vm1-hmmerthread.age.mpg.de>. Among the remote conserved domain hits in our dataset we find a number of interesting new or additional function(s) that could be assigned to proteins associated with a selected number of biological processes and human diseases. Our data allow for many predictions that can be functionally tested and thus open up completely new avenues in experimental research. In conclusion, with the HMMerThread database we have created a rich and accurate resource of remotely conserved domains of great value to experimental biological and medical research. Results {#s2} ======= Major modifications and improvements of the new HMMerThread software {#s2a} -------------------------------------------------------------------- The HMMerThread software searches for remotely conserved domains in proteins by a combination of relaxed sequence-based conserved domain searches with a subsequent fold recognition step to eliminate false positive domain hits resulting from high E-value thresholds [@pone.0017568-Bradshaw1] ([Figure 1](#pone-0017568-g001){ref-type="fig"}). We have adapted the HMMerThread algorithm to handle entire proteomes. For genome-wide annotations, conserved domain searches were carried out using the HMMER2 software (version 2.3.2) [@pone.0017568-Eddy1] with an E-value threshold of 50, which allows for the detection of sequence relationships well beyond statistically significant thresholds. A single HMMerThread run works as follows: first, an HMMER2-search is run against the Pfam database. In case a conserved domain with an E-value above the significance threshold of 1e-04 is detected, a subsequent fold recognition step is carried out. Provided that the structure(s) of the expected conserved domain is (are) positively identified, the conserved domain is scored as a positive hit. If the E-value of the conserved domain search is greater than 0.1, a validation procedure is carried out to ensure correct identification of a weak conserved domain hit. Validation steps include the identification of the same conserved domain in at least two of the orthologs of 3 related species, if available (see Supplemental [Table S1](#pone.0017568.s004){ref-type="supplementary-material"}), as well as identification of essential functional residues provided by the CD database [@pone.0017568-MarchlerBauer1]. ::: {#pone-0017568-g001 .fig} 10.1371/journal.pone.0017568.g001 Figure 1 ::: {.caption} ###### Architecture of genome-wide HMMerThread searches. Each protein of a species\' proteome is sent to a conserved domain search using HMMER2 against the Pfam database with an E-value threshold of 50. If a conserved domain with an E-value below 1e-04 is detected, it is positively scored. In case an identified domain has an E-value above 1e-04, a pre-processing and fold recognition step is performed. In case of a positive identification (p\<0.001), a conserved domain is scored, if the HMMER2 E-value of the conserved domain is below 0.1. If the HMMER2 E-value is above 0.1 and the associated fold has been scored positively, a cross-species validation is performed and essential residues are flagged for a confident assignment of a conserved domain. :::  ::: ### Improvement of the threading module {#s2a1} We needed to take two factors into consideration, when choosing a threading engine for genome-wide HMMerThread searches: First, we needed an algorithm, which was parallelizable and readily adaptable for a high-performance computing setting. Second, we wanted to ensure good performance of the threading engine. Among the tools available, only few algorithms fulfilled the first criterion. Next to Threader3.5 [@pone.0017568-Jones1] we have used previously, OpenProspect [@pone.0017568-Kim1] was readily useable and adjustable for the available high-performance computing setting. We chose to use it for genome-wide HMMerThread searches, as it outperformed Threader3.5 in our tests (data not shown). Searches were furthermore carried out against the ASTRAL structural library [@pone.0017568-Brenner1]. ### Improved scoring function of the fold recognition module in HMMerThread {#s2a2} Our attempt on unsupervised fold recognition required the development of a reliable scoring of a threading hit. The Z-score of threaded structures in OpenProspect cannot be easily interpreted and very diverse structure families can give similar Z-scores for the same sequence. Yet, as our approach combines sequence similarity searches together with fold recognition, we could take advantage of knowing *a priori,* which structural folds to look for. Therefore, we considered two factors for scoring hits from fold recognition in HMMerThread: 1. We considered the Z-score from OpenProspect, as it represents a guide to the strength of a hit. Given a certain Z-score, we wanted to deduce a p-value reflecting the probability that this Z-score can be treated as significant. We therefore constructed a cumulative distribution function of the Z-scores for all structures in SCOP for a random set of 1000 human proteins. This gave us a naïve probability for each possible Z-score. The threshold at 0.5% of all threaded structures (∼top 60 hits) was determined as stringent after examining results of test sets containing known domains (E-values\<0.0001) and remotely conserved domains. 2. We considered the number of structures associated with a conserved domain that were positively identified by OpenProspect. As many conserved domains have generally more than a single structural representative in the SCOP database, we used the hypergeometric probability function to determine, whether a given set of structures associated with a conserved domain was significantly overrepresented. This method enabled us to discriminate between significantly scoring structures due to their high frequency and those that are truly overrepresented and therefore a true positive hit. The top 60 hits we considered approximately equals 90% of the hits at a hypergeometric p-value threshold of 0.05 (Supplemental [Figure S1](#pone.0017568.s001){ref-type="supplementary-material"}). Our final scoring scheme combines the p-value of the top representative conserved domain based on its Z-score -- therefore ensuring that a structural hit is truly significant -- with the p-value of the structure being seen by chance due to its representation in the structural library. For this combined probability, we determined a significance threshold of 0.001. Based on our benchmarking of the HMMerThread software discussed below, we found that this combined probability is a robust scoring mechanism for remotely conserved HMMerThread domains. Benchmarking of the improved HMMerThread software {#s2b} ------------------------------------------------- We looked at two aspects of the new HMMerThread software: First, we tested the threading engine for its ability to detect fold classes. We considered this important as the threading module we used has a limited statistical framework and could also lead to a loss of true positive remote conserved domains. Second, we tested the performance of HMMerThread by calculating precision, recall and accuracy of the software. ### Detection of conserved domains with statistically significant sequence conservation {#s2b1} First, we analyzed the performance of the fold recognition module of HMMerThread by testing its ability to find known conserved domains with a statistically significant HMMER2 E-value. We took all detected conserved domains from the human proteome with an E-value between 1e-20 and 1e-04 and submitted them to OpenProspect for fold recognition. As shown in [Figure 2](#pone-0017568-g002){ref-type="fig"}, we could retrieve 88% of all known conserved domains. We also investigated whether HMMerThread could score all domain types and found that we can positively identify 60% (1920 of 3192) of domain types with OpenProspect. This is in accordance with the observation that ∼30% of domain types only occur in the prokaryotic kingdom and are not found in eukaryotes [@pone.0017568-Apic1]. It also highlights that we cannot identify all domain types with the threading algorithm and the scoring scheme we use. ::: {#pone-0017568-g002 .fig} 10.1371/journal.pone.0017568.g002 Figure 2 ::: {.caption} ###### Performance of the OpenProspect software. Comparison of positive identifications of conserved domains using either HMMER2 alone (grey bars) or HMMerThread (red bars). We have tested an E-value range between 1e-20 and 1e-04 for positive identification of conserved domains by HMMerThread and 88% of conserved domains could be positively identified. :::  ::: ### Calculating Precision, Recall and Accuracy of the HMMerThread software {#s2b2} We next set out to calculate the precision, recall and accuracy of the HMMerThread software. In order to do this, we had to identify a sufficiently high number of proteins containing remotely conserved domains that we knew were true positives. We decided to benchmark the HMMerThread software by a hide and seek procedure that involved different versions of the Pfam conserved domain database. The rationale behind this approach was that with growing domain families, more distantly related members become associated with a conserved domain profile, thereby relaxing the profile and making it more sensitive. According to this hypothesis, we would find a number of conserved domains in proteins that did not score with a significant E-value in older Pfam releases (in our case Pfam10, released in July 2003), but achieved a significant score in the newer one (Pfam22, released in July 2007 and the version used as the conserved domain database in this study). We created an overlapping dataset of the Pfam10 and Pfam22 releases, resulting in 5266 conserved domain profiles that we could test. This dataset we refer to from hereafter as the Pfam10∶22 remotely conserved domain set. We ran HMMER2 searches against both profile versions of the conserved domains using the human proteome and selected those domains, which scored with an E-value\>0.1 using Pfam10, while having an E-value\<0.1 in Pfam22 and where the difference between the two E-values was greater than or equal to a 10-fold change. This resulted in a total of 408 conserved domains that could be considered as true positive, weakly conserved hits in Pfam10. For these domains, we performed HMMER2 searches against Pfam10, which generated 1520 possible, overlapping hits. These, we could analyze for true positive (TP), false negative (FN), true negative (TN) and false positive (FP) identification (see Supplemental [Table S2](#pone.0017568.s005){ref-type="supplementary-material"}). Domain profiles that belonged to the same clan were excluded. Based on the Pfam22 profiles, 390 conserved domains were correct hits, only 18 of which we could not identify, resulting in 372 (or 95%) TPs and 5% FN. From the remaining domains, we obtained 142 (14%) FPs, which resulted in a precision of 74%. The 14% false positive predictions with the HMMerThread algorithm is an obvious concern for automatic prediction of remotely conserved domains. In order to reduce the percentage of false positive predictions, we would have to accept too many false negative predictions (see Supplemental [Figure S2](#pone.0017568.s002){ref-type="supplementary-material"} and also Supplemental [Table S3](#pone.0017568.s006){ref-type="supplementary-material"}). To reduce the number of false positive predictions to for instance 3%, we would only be able to retrieve 48% of true positives (see Supplemental [Table S3](#pone.0017568.s006){ref-type="supplementary-material"}). In order to address this problem, we wanted to know whether cross-species validation could reduce the number of false positive predictions, while retaining the good performance of the HMMerThread algorithm in recall. This also reflects the intended usage of the HMMerThread resource, where only cross-species validated domains are considered as reliable predictions. To this end, we repeated the searches with orthologs of the Pfam10∶22 dataset from mouse, dog and chicken. Validation of HMMerThread results in one species led to only a slight reduction of the false positive rate from 14% to 11%. At the same time, it reduced the number of true positive hits to 274 and increased the number of false negative predictions (note that for 79 of true positive hits, no suitable ortholog could be found for validation, which reduced the proteins that could be scored to 311 proteins with weakly conserved domains). Validation against 2 species reduced the false positive rate to 8% and the accuracy to 90%. Detection of a remotely conserved domain in all 3 orthologs pushed down the number of false positive predictions to 3%, however with a recall of only 68%. We therefore concluded that a validation against 2 species provided the best compromise in true positive and false positive predictions (see [Table 1](#pone-0017568-t001){ref-type="table"} and Supplemental [Table S2](#pone.0017568.s005){ref-type="supplementary-material"}). As is shown later, our dataset induced a very similar number of false positive predictions with other, very reliable algorithms like Superfamily [@pone.0017568-Gough2]. The relatively high number of false positives therefore seems to be an inherent feature of the Pfam10∶22 dataset. In many cases, falsely identified domains contain for instance short domains like Zn-fingers, which are difficult to distinguish correctly (see Supplemental [Table S2](#pone.0017568.s005){ref-type="supplementary-material"}). In order to ensure low false prediction rates, we are marking domains with a length shorter than 50 amino acids in the HMMerThread database. ::: {#pone-0017568-t001 .table-wrap} 10.1371/journal.pone.0017568.t001 Table 1 ::: {.caption} ###### Comparison of the new HMMerThread to its predecessor, GenThreader and the Superfamily resource. ::: {#pone-0017568-t001-1} HMMerThread (new version, 2 species validation) HMMerThread (old version, based on Threader 3.5, ProFAT) ------------------------------------------------- ------- ---------------------------------------------------------- ------- True Positives 217 True Positives 186 False Negatives 47 False Negatives 197 False Positives 76 False Positives 53 True Negatives 845 True Negatives 781 False Negative Rate 18% False Negative Rate 51% False Positive Rate 8% False Positive Rate 6% *Precision* *74%* *Precision* *78%* *Accuracy* *90%* *Accuracy* *79%* *Recall* *82%* *Recall* *49%* **GenThreader** **Superfamily** True Positives 121 True Positives 271 False Negatives 256 False Negatives 91 False Positives 5 False Positives 29 True Negatives 773 True Negatives 565 False Negative Rate 68% False Negative Rate 25% False Positive Rate 2% False Positive Rate 5% *Precision* *96%* *Precision* *90%* *Accuracy* *77%* *Accuracy* *87%* *Recall* *32%* *Recall* *75%* ::: Taken together, we conclude that HMMerThread is a very powerful technique to identify true positive, remotely conserved domains and is moreover highly efficient in discriminating true positive from false positive hits. We could confirm the good performance of HMMerThread, when we searched for remotely conserved domains that have been described in literature before and which we were well familiar with: We could confirm the existence of a BAR domain (Bin-Amphiphysin-RVS) in most of the proteins that we had previously described [@pone.0017568-Habermann1] (data not shown, please refer to [BAR]{.underline} domains in human). We could also confirm most of the remotely conserved domains in proteins discussed in the original manuscript describing the ProFAT and HMMerThread server [@pone.0017568-Bradshaw1]. The RNA-Recognition-Motif (RRM\_1) domain was found in the LOC84060 protein however automated HMMerThread searches without considering overlapping remote conserved domain hits did not reveal the presence of the RRM\_1 domain in the Parn proteins ([LOC84060]{.underline} and [Parn]{.underline}). The SAM domain (for Sterile Alpha Motif) was verified in the epidermal growth factor receptor pathway substrate 8 protein families [EPS8]{.underline} and [EPS8L3]{.underline} and we automatically detected the Acetyltransferase domain (Acetyltransf\_1) in the LOC79969 proteins using the novel HMMerThread server ([LOC79969]{.underline}). Interestingly, a large-scale screen of protein-protein interactions in worm revealed that the *C. elegans* ortholog of LOC79969, W06B11.1, interacts with a methyltransferase (C01B10.8), suggesting that this protein is part of a larger chromatin-remodelling complex [@pone.0017568-Li1]. We also found the previously described Acetyltransf\_1 domain in the protein Eco1 from *Saccharomyces cerevisiae* [@pone.0017568-Ivanov1] (Eco1). This domain was also found in the worm and fly orthologs, though we did not detect it in vertebrates or *Schizosaccharomyces pombe* ([fission yeast]{.underline}, [fly]{.underline}, [worm]{.underline}, [zebrafish homolog 1]{.underline}, [zebrafish homolog 2]{.underline}, [mouse homolog 1]{.underline}, [mouse homolog 2]{.underline}, [human homolog 1]{.underline}, [human homolog 2]{.underline}). HMMerThread also confirmed the presence of winged-helix domains in two proteins required for meiotic recombination, Mnd1 and Hop2 [@pone.0017568-Mochizuki1]. Finally, HMMerThread identified a remotely conserved CARD domain in the Death Receptor 6 (DR6, aka [TNFRSF21]{.underline}), the presence of which was also shown by structural analysis (pdb-code 2dbh, Inoue M, Koshiba S, Kigawa T, Yokoyama S, unpublished). Comparison of HMMerThread to its predecessor, the GenThreader and Superfamily algorithms {#s2c} ---------------------------------------------------------------------------------------- In order to estimate the advancement of HMMerThread in predicting remotely conserved domains compared to its previous version, as well as to other existing resources, we used the Pfam10∶22 remotely conserved domain set to calculate recall, precision and accuracy of the old HMMerThread algorithm, GenTHREADER [@pone.0017568-McGuffin1] as well as the algorithm used by Superfamily [@pone.0017568-Gough2] (see [Table 1](#pone-0017568-t001){ref-type="table"} and Supplemental [Table S2](#pone.0017568.s005){ref-type="supplementary-material"}). ### Significantly improved performance of the novel HMMerThread software {#s2c1} When we compared the precision, accuracy and recall of the old versus new version of HMMerThread, we find that our modifications have significantly improved the software (see [Table 1](#pone-0017568-t001){ref-type="table"} and Supplemental [Table S2](#pone.0017568.s005){ref-type="supplementary-material"}). While we achieved a recall of 82% with the new method, we only reached 49% with the old version. The new version of HMMerThread gets a slightly worse precision with 74% versus 78% of the old version, as we have a slightly higher false positive rate (8% vs 6%, respectively). Both results are due to the very different scoring scheme of the old versus new version of the algorithm. In the old version of HMMerThread, the scoring of a remotely conserved domain consisted of identification of a positive structural hit within the first 25 identified structures. As one is to expect a slightly higher false positive rate, one also can expect a much lower false negative rate with the novel approach we have taken for scoring HMMerThread hits. Finally, the accuracy of the old versus new version of HMMerThread is 79% versus 90%. Based on these data, we conclude that we could significantly enhance the performance of the HMMerThread algorithm. ### Comparison of the new HMMerThread to existing software for detecting remote conservation {#s2c2} We decided to compare the novel HMMerThread algorithm to two existing resources that provide information on remote conservation between proteins. For one, we looked at GenTHREADER [@pone.0017568-McGuffin1], which uses a fold recognition pipeline to predict the putative three-dimensional structures of proteins on a genome-wide scale. In order to estimate the performance of GenTHREADER, we used the Pfam10∶22 remotely conserved domain set and applied GenTHREADER to identify the correct fold of a remotely conserved domain (for detailed description, see [Methods](#s4){ref-type="sec"}). The performance of GenTHREADER was very comparable to the old version of HMMerThread, with a very low false positive rate (2%), but also a quite high false negative rate (68%). The recall of GenTHREADER was therefore only 32%, precision reached 96% and the overall accuracy was with 77% very similar to old version of HMMerThread (see [Table 1](#pone-0017568-t001){ref-type="table"} and Supplemental [Table S2](#pone.0017568.s005){ref-type="supplementary-material"}). Superfamily [@pone.0017568-Gough2] was the second resource we compared the HMMerThread database to. After mapping of the SCOP-IDs of a Superfamily to conserved domains of the Pfam database using PDBMAP, we chose to score a true positive hit, if a Superfamily was reported that included the correct remote conserved Pfam hit; all related Pfam families found within a SCOP family were ignored for false positive predictions, in addition to the excluded CLAN members that we used for HMMerThread or GenThreader. Superfamily was able to identify 271 of our positive conserved domain set, which resulted in a recall of 75%. It achieved a higher precision (90%), with a false positive rate of only 5%. The overall accuracy of Superfamily was 87% and therefore very similar to the new version of HMMerThread ([Table 1](#pone-0017568-t001){ref-type="table"} and Supplemental [Table S2](#pone.0017568.s005){ref-type="supplementary-material"}). As noted earlier, it seems to be a built-in feature of our dataset that even precise algorithms like GenThreader or Superfamily show a higher than usual false positive rate (2% and 5%, respectively). We therefore decided that a false positive rate of 8% with our dataset was acceptable for the new HMMerThread algorithm. We conclude from this data that HMMerThread with its unique approach to consider not only sequence-, but also structural information outperforms other methods currently available in identifying remotely conserved domains. Genome-wide HMMerThread Searches {#s2d} -------------------------------- We have carried out HMMerThread searches against the proteomes of the eight most common model organisms including human and detected a total of 58330 weakly conserved domains with an E-value above 0.1 (see [Table 2](#pone-0017568-t002){ref-type="table"}). About 13000 of these were validated in at least one additional species and ∼6000 in two. Many of the model organisms we chose lack a third species with a reasonable phylogenetic distance. Therefore, only ∼2000 domains can be found in three other species with ∼1000 of these identified in human alone. Nearly all fold classes of the current SCOP release (1.71) could be identified using genomic HMMerThread searches, with globular domains being the vast majority of structures found ([Table 3](#pone-0017568-t003){ref-type="table"}). ::: {#pone-0017568-t002 .table-wrap} 10.1371/journal.pone.0017568.t002 Table 2 ::: {.caption} ###### Statistics of HMMerThread weakly conserved domains in the 8 proteomes analyzed. ::: {#pone-0017568-t002-2} Genome Total proteins Remotely conserved domains 3-species validation 2-species validation 1-species validation ----------------- ---------------- ---------------------------- ---------------------- ---------------------- ---------------------- H. sapiens 33466 12038 1031 2672 4492 M. musculus 34981 11460 636 1873 3541 D. rerio 29720 11422 \- 872 1728 C. elegans 23518 7664 \- \- 1741 D. melanogaster 19,388 7430 249 556 1075 S. cerevisiae 5868 1919 41 87 360 S. pombe 5004 1506 \- \- \- D. discoideum 13501 4891 \- \- \- TOTAL: 165446 58330 1957 6060 12937 ::: ::: {#pone-0017568-t003 .table-wrap} 10.1371/journal.pone.0017568.t003 Table 3 ::: {.caption} ###### SCOP classes (version 1.71) identified in the 8 genome-wide HMMerThread searches. ::: {#pone-0017568-t003-3} SCOP Class Count Percentage ------------------------------------------------- ------- ------------ Small proteins 20297 24.46% Alpha and beta proteins (a+b) 18250 21.99% All beta proteins 15273 18.40% All alpha proteins 13407 16.16% Alpha/beta proteins (a/b) 10935 13.18% Membrane and cell surface proteins and peptides 3185 3.84% Peptides 677 0.82% Multi-domain proteins (alpha and beta) 578 0.70% Coiled coil proteins 377 0.45% Designed proteins 7 0.01% ::: The results of the genome-wide HMMerThread searches are presented in a web-based relational database, which includes the results from all eight proteomes analyzed. The database can be queried by gene name, protein ID or Pfam conserved domains. The HMMerThread domains are shown along with associated annotation for the given gene from diverse sources like the NCBI, SGD, Wormbase, Flybase or HPRD. Though remote conserved domain searches were not carried out against the Pfam24 database [@pone.0017568-Finn2], we provide Pfam24 domain annotation in the HMMerThread database. We show the HMMerThread domains graphically in direct relation to conserved domains from InterProScan [@pone.0017568-Quevillon1], which gives the user an immediate overview of the presumed functions of the protein under study. Next to the Pfam domains, we integrate PROSITE sequence-based features from pattern matches underneath the domain. Remotely conserved HMMerThread domains are colour-coded based on their validation status (green  =  no validation required (HMMER2 E-value\<0.1), red  =  present in 3 validation species, orange  =  present in 2 validation species and yellow  =  present in 1 validation species, grey  =  no validation data available). The validation information is furthermore provided by a mouse-over popup on the domain images (Supplemental [Figure S3](#pone.0017568.s003){ref-type="supplementary-material"}). We extract species-specific information for further annotation of entries; records from human for instance contain NCBI gene summaries, GO terms and HPRD interactions, while entries from *S. cerevisiae* contain summaries and GO terms from SGD. For human records, we integrate iPfam data along with domains that may explain interactions in a separate table. We also provide a "live search" feature, where sequences that have not been processed by us can be searched using the HMMerThread pipeline with an updated version of the Astral database (the currently used version is 1.73). The database is publicly available at <http://vm1-hmmerthread.age.mpg.de>. Novel functional predictions based on remotely conserved domains {#s2e} ---------------------------------------------------------------- We set out to search for novel functional predictions based on detected HMMerThread domains. To do this, we followed several strategies: 1) we searched for undiscovered, remotely conserved domains in proteins that were detected in genome-wide functional screens. We discuss the overall statistics of a genome-wide functional dataset on factors involved in Hepatitis C Virus replication in human cells and describe one more detailed example with potential mechanistic insight. 2) We looked for remotely conserved domains in proteins associated with a biological process. We focused on mitotic and meiotic genes, as well as genes associated with diseases using the OMIM resource (NCBI) [@pone.0017568-Amberger1]. We present two examples for each category. 3) We searched for domains within domains, as weak functionally conserved domains with a known function are often found within DUF (Domain of Unknown Function) domains. 4) We looked at domain-domain interaction data that might shed light on the binding sites and mode of interaction between proteins. Remotely conserved domains in proteins identified in functional screens {#s2f} ----------------------------------------------------------------------- We have chosen a functional screen for cofactors of Hepatitis C Virus replication in human cells [@pone.0017568-Tai1], whose hits we have annotated using HMMerThread in addition to InterProScan domains. Among the genes that were involved in viral replication, those associated with Golgi vesicle binding, organization and biogenesis were overrepresented (see Supplemental [Table S4](#pone.0017568.s007){ref-type="supplementary-material"}). In this dataset, 29 remotely conserved HMMerThread domains are found, which are predominately involved in protein binding activities. ### The transcriptional repressor Nab1 contains a remotely conserved SAM domain {#s2f1} Among the hits that showed significant reduction of viral replication with more than two independent silencing triggers was the gene [Nab1]{.underline} (NGFI-A binding protein 1). Nab1 is a transcriptional co-repressor that interacts directly with early growth response transcription factor 1 (Erg1) and thereby either positively or negatively modulates transcriptional activation of early response genes [@pone.0017568-Russo1], [@pone.0017568-Sevetson1]. Egr1 itself has been implicated in Hepatitis Virus C infection through the activation of IGF-II (insulin growth factor II) gene expression, which is a critical factor during the formation of hepatocellular carinoma (HCC) [@pone.0017568-Lee1]. The fact that Nab1, a stable interactor and transcriptional co-factor of Erg1 is found in a screen for viral replication of Heptatitis C virus raises the possibility that Nab1/Erg1 is already actively assisting Hepatitis C pathogenesis by helping viral reproduction. Moreover, it opens the possibility that not only transcriptional activation, but also repression via Erg1 might be required for efficient replication of the virus. This is in accordance with the observation that Hepatitis C virus not only induces, but also represses the transcription of a set of genes [@pone.0017568-Ray1], [@pone.0017568-Ray2]. We detected an N-terminal SAM (Sterile Alpha Motif) domain in the Nab1 proteins, which lies within their N-terminal NCD1 domain ([Figure 3](#pone-0017568-g003){ref-type="fig"}). The SAM domain of Nab1 shows sufficient sequence conservation for detection by PSI-BLAST searches [@pone.0017568-Altschul1] (data not shown). SAM domains are thought to be protein interaction domains and are found in a number of proteins that are involved in different developmental processes throughout the eukaryotic kingdom [@pone.0017568-Schultz1]. The simplest functional implication of the presence of a SAM domain in Nab1 is that this domain provides the interaction interface to Egr-1. However, SAM domains among others are also found in transcriptional repressors such as the TEL protein. Transcriptional silencing by TEL has been proposed to involve oligomerization of its N-terminal SAM domain, building a proteinaceous core around which the DNA is wrapped, thereby enabling spreading of the repressor activity [@pone.0017568-Kim2]. Although it is not clear, whether the SAM domain of Nab1 is capable of oligomerization, the presence of this conserved SAM domain in Nab1 suggests that Nab1 plays an essential role during Hepatitis C-induced transcriptional repression or activation. ::: {#pone-0017568-g003 .fig} 10.1371/journal.pone.0017568.g003 Figure 3 ::: {.caption} ###### Multiple sequence alignments of remotely conserved domains in proteins identified in functional screens. Multiple sequence alignment of the Nab1 family with the SAM domain family (taken from CDD). Residues that are conserved between the two families are highlighted in yellow, those found in only one of them are highlighted in blue and green, respectively. Essential, functional residues retrieved from the CD database are indicated by hash keys. Accession numbers of sequences can be found in Supplemental [Table S7](#pone.0017568.s010){ref-type="supplementary-material"}. :::  ::: Prediction of molecular mechanistic function to generate testable hypotheses for proteins involved in cell division and proteins associated with human diseases {#s2g} --------------------------------------------------------------------------------------------------------------------------------------------------------------- In order to provide examples of the predictive power of the HMMerThread method, we set out to search for remotely conserved domains in proteins involved in mitosis and/or meiosis, as well as genes associated with human diseases (taken from OMIM) that could elucidate their molecular mechanism. ### Yeast Ssp2 harbours a RNA-binding domain and might be involved in mRNA localization during sporulation {#s2g1} Among the weak, conserved domain hits was an RRM\_1 domain we found in [Ssp2]{.underline}. This protein is required for the proper formation of the prospore membrane (PSM) and the spore wall (SW) during sporulation ([Figure 4 A](#pone-0017568-g004){ref-type="fig"}). We could confirm this remotely conserved domain by PSI-BLAST [@pone.0017568-Altschul1] searches (data not shown). Yet, why is a RNA-binding domain found in a protein involved in spore wall formation? Ssp2 is specifically required for vesicle fusion during formation of the PSM, as the *ssp2* null mutant can be partially rescued by overexpression of proteins from the vesicle fusion machinery, namely the phospholipase D Spo14, and the t-SNARE protein required for meiosis, Sso1. At least Spo14, together with Ssp2 is specifically localized to the PSM after the second meiotic division. Interestingly, when Oyen and colleagues [@pone.0017568-Oyen1] analyzed the sporulation-specific functions of Sso1, they found that next to functional domains within the protein itself, the 3′UTR of the sso1 mRNA is essential for sporulation and this function cannot be rescued by the 3′UTR of its close paralogue, sso2, which does not have a meiosis-specific function. In the same report, Oyen et al. tested for expression levels of sso1 and the sso2 paralogue, and found no difference between the two genes, which suggests that translational control does not play a role in the sporulation-specific function of the sso1 3′UTR. This raises the possibility that proper localization of sso1 mRNA is essential for the function of the Sso1 protein in late meiotic events, suggesting that potentially mRNA localization plays a crucial role in sporulation. The RNA-binding protein Ssp2 might therefore be essential for the proper localization of the mRNA of one -- or multiple -- genes during late meiotic stages to the PSM. ::: {#pone-0017568-g004 .fig} 10.1371/journal.pone.0017568.g004 Figure 4 ::: {.caption} ###### Multiple sequence alignments of remotely conserved domains in proteins associated with mitosis and meiosis. (**A**) Multiple sequence alignment of the Ssp2 family with the RRM\_1 domain. (**B**) Multipe sequence alignment of the Wapl/Rad61 family with the SAP domain family. Residues that are conserved between the two families are highlighted in yellow, those found in only one of them are highlighted in blue and green, respectively. Essential, functional residues retrieved from the CD database are indicated by hash keys, those retrieved from literature (SAP domain) with stars. Accession numbers of sequences can be found in Supplemental [Table S7](#pone.0017568.s010){ref-type="supplementary-material"}. :::  ::: ### A putative SAP domain was found in the C-terminus of the WAPL proteins {#s2g2} The Wapl (*Wings-apart like*) protein was first described as a heterochromatin organizer in fly, whose loss leads to chromosome missegregation in meiosis [@pone.0017568-Verni1]. Subsequently, Wapl was identified as an essential player in chromosome segregation in mitosis, as it is physically associated with two cohesin subunits (Pds5 and Scc3) and associates with DNA at the same location as cohesin [@pone.0017568-Kueng1]. The cohesin complex is a ring-like structure that entraps sister chromatids and thereby ensures the accurate segregation of chromosomes at the metaphase to anaphase transition and enables efficient repair of DNA double-strand breaks (DBS) in G2 [@pone.0017568-McNairn1], [@pone.0017568-Nasmyth1]. In vertebrate cells, Wapl is required to remove cohesin from chromosome arms during prophase and prometaphase and promotes rapid turnover of cohesin during interphase. Loss of Wapl leads to 'hypercohesed' mitotic chromosomes [@pone.0017568-Kueng1], [@pone.0017568-Gandhi1]. Loss of the yeast ortholog of Wapl, Rad61/Wpl1, in contrast, results in weak impairment of cohesion [@pone.0017568-Rowland1], [@pone.0017568-Sutani1], which led the authors to speculate that the Pds5/Scc3/Wpl1 complex helps in the maintenance of cohesion ring closure around DNA [@pone.0017568-Rowland1]. In order to reconcile the two opposing functions of Wpl1/Wapl in vertebrate and yeast cells, Peters [@pone.0017568-Peters1] proposed that the Wapl proteins perform multiple roles in DNA-cohesion interaction, cohesion establishment, maintenance and dissociation and that depending on the cellular system, one or the other function of Wapl will display a phenotype. So far, it is not understood how Wapl enables either the stabilization of the cohesin ring around DNA molecules or how it promotes loss of cohesion during prophase. It is also not known, which other functions Wapl might fulfil. We found a C-terminal SAP domain in the vertebrate [Wapl]{.underline} proteins ([Figure 4](#pone-0017568-g004){ref-type="fig"} B). SAP domains occur in a multitude of DNA binding proteins that contain a diverse set of other functional conserved domains. They bind to AT-rich chromosomal regions and were first described in chromosomal organization [@pone.0017568-Aravind1]. We propose that the predicted SAP domain of Wapl could be directly associated with chromosomal DNA. The SAP domain, being very short and if a true positive, degenerate in Wapl proteins, can in this case not be verified using sequence-based methods. When threading this region using Phyre [@pone.0017568-Kelley2], however the majority of the identified structures are DNA-binding domains with a helix-loop-helix topology. The highly conserved Lysine residue in the loop region of the two helices, as well as the conserved positive charge in the first helix are present in most, yet not all Wapl orthologs. The putative SAP domain of the yeast Rad61 protein shows for instance low conservation on sequence level and some of the phenotypic differences between species upon loss of this protein might be explained by this observation. ### A putative CUT-like Helix-Loop-Helix (HLH) domain was found in the C-terminus of AKAP10 {#s2g3} The dual-specific A-kinase anchor protein 10 (AKAP10) is member of a diverse protein family, which binds to the regulatory subunit of protein kinase A (PKA, for a review on AKAP10, see [@pone.0017568-BurnsHamuro1]). Unlike other AKAP family members, it contains two central RGS (Regulator of G-protein Signaling) domains, which are usually found in GTPase activating proteins (GAPs) for G-proteins [@pone.0017568-Ross1]. The RGS domains in AKAP10 interact with the recycling small GTPases Rab4 and Rab11 [@pone.0017568-Eggers1], making this protein a switch point between signalling and endocytosis. The protein is expressed in all tissues and seems to be enriched in mitochondria [@pone.0017568-Chatterjee1]. An Isoleucine to Valine mutation in the C-terminal PKA interacting motif that leads to three-fold higher affinity for PKA, has been associated with higher mortality [@pone.0017568-Kammerer1]. Humans carrying this mutation show an increased basal heart rate and decreased heart rate variability. Mice carrying the same allele show cardiac arrhythmias and die prematurely [@pone.0017568-Tingley1], which suggests that AKAP10 plays an essential role in the control of heart rhythm and which makes it an interesting medical target. We found a C-terminal CUT-like HLH-domain in the AKAP-10 proteins of [human]{.underline} and [mouse]{.underline} ([Figure 5 A](#pone-0017568-g005){ref-type="fig"}) right adjacent to the PKA interacting motif. CUT domains are DNA-binding domains that either bind alone or in combination with homeodomains, many of which are actually found in the same protein [@pone.0017568-Lannoy1]. The weakly conserved HLH like fold (the HMMER2 E-value is 6.9) can again be verified by PSI-BLAST searches, which identify HLH domain proteins like microphthalmia-associated transcription factor (for instance the human protein NP\_006713) and transcription factors EB from diverse species. Interestingly, no true CUT domain can be found by PSI-BLAST searches (not shown), which opens the possibility that the putative DNA-binding domain of AKAP10 might be member of a different HLH family. Yet, what function does a DNA-binding domain perform in a protein predominantly localized to mitochondria and presumably involved in signal transduction and recycling? It has been shown that AKAP10, together with other proteins carrying a RGS domain undergoes nuclear/nucleolar translocation upon mild heat, proteotoxic stress or the overexpression of Heat Shock Transcription Factor 1 (HSF1). Some members of the RGS-domain family that also carry winged-helix DNA binding domains are thought to be involved in stress-induced gene expression (see [@pone.0017568-Chatterjee1] and references therein) and it is possible that AKAP10 either alone or in combination with another homeodomain transcription factor is involved in regulating stress-related gene expression. AKAP10 therefore might again represent a multifaceted switch-point in the cell that combines signalling, recycling and transcriptional response. ::: {#pone-0017568-g005 .fig} 10.1371/journal.pone.0017568.g005 Figure 5 ::: {.caption} ###### Multiple sequence alignments of remotely conserved domains found in proteins associated with human diseases. (**A**) Multiple sequence alignment of the AKAP10 family with the CUT-like HLH domain family. (**B**) Multiple sequence alignment of the Lba protein family with the VHS domain family. Residues that are conserved between the two families are highlighted in yellow, those found in only one of them are highlighted in blue and green, respectively. Essential, functional residues retrieved from the CD database are indicated by hash keys. Accession numbers of sequences can be found in Supplemental [Table S7](#pone.0017568.s010){ref-type="supplementary-material"}. :::  ::: ### A VHS domain in the chs1/beige protein Lba is implicated in immune deficiency {#s2g4} The maturation of B-cells, monocytes and dendritic cells is induced by Lipopolysaccharide (LPS) that stimulates the response of the cells to bacterial pathogens. The protein Lba (for LPS-responsive, beige-like anchor gene, aka Lrba) has been identified as a gene that is expressed in response to LPS and is involved in the maturation of immune cells [@pone.0017568-Wang1]. Lba has also been associated with Chediak-Higashi syndrome (CHS), which is characterized by a severe immune defect among other symptoms [@pone.0017568-Wang1]. Like other members of the *chs1/beige* family, mutations in Lba lead to perturbed intracellular trafficking and it seems that Lba function is essential for polarized transport of intracellular trafficking. Lba also carries features of AKAP (A kinase anchor proteins), namely the ability to bind to Protein kinase A (PKA). The Lba-GFP fusion protein translocates from cytoplasm to intracellular vesicles upon stimulation with LPS and can be found on Golgi membranes, lysosomes, ER, plasma membrane and endocytic vesicles [@pone.0017568-Wang1]. It is so far unknown, which domain is required for the association of Lba with membranes. *In vitro* experiments precluded binding of the PH-BEACH domain to phospholipids [@pone.0017568-Gebauer1]. We discovered a VHS-like domain in the N-terminal region of the [Lba]{.underline} proteins ([Figure 5 B](#pone-0017568-g005){ref-type="fig"}). Though sequence similarity is very weak, we can verify the VHS domain by profile-profile comparisons (hhpred, [@pone.0017568-Soding1]) and fold recognition (Phyre, [@pone.0017568-Kelley2]) (data not shown). VHS domains have been implicated in cargo recognition and vesicle trafficking [@pone.0017568-Lohi1] and are found in a variety of proteins involved in intracellular transport [@pone.0017568-Lohi2]. When found in GGA (Golgi-localized, γ-ear containing, ARF-binding) proteins, the VHS domain binds to a subset of sorting receptors that move and transfer cargo between the trans-golgi network (TGN) and the endosomal compartment [@pone.0017568-Nielsen1]; [@pone.0017568-Puertollano1]; [@pone.0017568-Zhu1]; [@pone.0017568-Takatsu1]. Though VHS domains occur in combination with different accessory conserved domains, their molecular function is presumably identical. Interestingly, they are mostly found in the very N-terminal regions of the proteins and this localization is thought to contribute to their function, though no experimental proof for this hypothesis exists [@pone.0017568-Lohi1]. As the predicted VHS domain of Lba is not in the very N-terminus of the protein, we propose that this domain is VHS-like and might have at least a subset of similar functionality to the N-terminal VHS domains in cargo sorting required for polarized membrane trafficking. It is furthermore possible that splice variants of the full-length Lba protein lacking the N-terminal part might exist that thus harbour the VHS domain in their very N-terminus. Moreover, the presence of the PKA interacting motif in the Lba family indicates that these proteins serve as linkers between signalling and trafficking. Via the proposed VHS domain, they might ensure proper channelling of extracellular stimuli within the intracellular membrane network. Novel, weakly conserved domains within conserved domains of unknown function (DUF) {#s2h} ---------------------------------------------------------------------------------- Domains of unknown function (DUF) are often annotated in case a clear block of sequence similarity is found within a protein family with unknown function. Domain profiles of DUF-domains are often restrictive, as not many members have been assigned to these conserved domain families. When analyzing the data from HMMerThread searches, we frequently found remotely conserved domains of known function within DUF-domains. For example, a Calponin Homology (CH-) domain lies within the N-terminal region of the DUF1042 domain (see for instance the Spef1 protein in [human]{.underline}, [mouse]{.underline} or [fish)]{.underline}. Other examples are methyltransferase\_11 domains hidden in the DUF689 domains ([DrXP\_684963]{.underline} or [DrCiapin1]{.underline}), which are easily verifiable by PSI-BLAST searching and the DUF738 domain, which contains an Acetyltransferase domain (Acetyltransf\_1) that was already described in an earlier section of this manuscript (see for instance the LOC79969 proteins from [human]{.underline}, [mouse]{.underline}, [fish]{.underline} or [worm]{.underline}). A complete list of conserved domains with an associated function within DUF domains can be found in Supplemental [Table S5](#pone.0017568.s008){ref-type="supplementary-material"}. Identification of potential interaction sites and prediction of interactors due to novel, weakly conserved domains {#s2i} ------------------------------------------------------------------------------------------------------------------ Interaction between proteins often takes place via conserved domains and this type of data is stored in the iPfam database [@pone.0017568-Finn3]. We used remotely conserved HMMerThread domains in the human proteome to search for interaction sites of previously known interaction partners that we have extracted from the HPRD resource [@pone.0017568-Prasad1]. The presence of a remotely conserved domain can also reveal potential interactors that have so far not been predicted. A full list of known interactors and their interacting domains based on remotely conserved domains can be found in Supplemental [Table S6](#pone.0017568.s009){ref-type="supplementary-material"}. Among the HMMerThread hits we found in the human interactome was a weakly conserved RPH3A\_effector domain (HMMER2 E-value of 11) in the protein [exophilin 5]{.underline} (Exph5, aka Slac-2), which overlaps with the PROSITE Rab-binding pattern. The RPH3A\_effector domain was initially described as a Rab3 interaction motif found in the Rabphilin-3A protein [@pone.0017568-Ostermeier1] and is structurally related to the Slp homology domain (SHD). Slac-2/Exph5 uses this domain to interact specifically with Rab27A [@pone.0017568-Fukuda1]. HMMerThread could successfully detect this weak sequence relationship. Another remotely conserved HMMerThread domain is the RhoGEF/DH domain of the protein [Als2cL]{.underline}, a closely related protein to Alsin (Als2), which lacks the N-terminal RCC (regulator of chromosome condensation) domain [@pone.0017568-Hadano1], [@pone.0017568-SuzukiUtsunomiya1] ([Figure 6](#pone-0017568-g006){ref-type="fig"}). Like Als2, Als2cL has a C-terminal VPS9 domain, which acts as a GEF for Rab5a. The protein was shown to form homodimers, which are able to interact with Als2 oligomers and these complexes localize to vesicular structures within the cell. Though Als2cL and Als2 share extensive sequence similarity over large parts of their sequence, their molecular functions seem distinct. Als2cL, for instance negatively modulates the endosome enlargment phenotype observed in Als2 mutants that have constitutive Rab5 GEF activity and rather leads to tubulation of endosomal compartments [@pone.0017568-Hadano1]. Next to its function in endosomal compartment dynamics, Als2 also regulates Rac-PAK signalling in neurite outgrowth [@pone.0017568-Tudor1] and it does so by acting as a GEF for Rac via its central RhoGEF domain. So far, binding to a Rho-type GTPase like Rac has not been reported for Als2cL. Furthermore, though the presence of the RhoGEF and PH domain in the N-terminus of the protein has been stated [@pone.0017568-SuzukiUtsunomiya1] and though the presence of this domain can be verified using PSI-BLAST or BLAST searches [@pone.0017568-Altschul1] (data not shown), this domain is not detected via standard domain search programs as it has an E-value of 4. We propose that like Als2, Als2cL will also interact with and act as a RhoGEF for a Rho-type GTPase, as there are few amino acid exchanges between Als2 and Als2cL in the essential residues. ::: {#pone-0017568-g006 .fig} 10.1371/journal.pone.0017568.g006 Figure 6 ::: {.caption} ###### Multiple sequence alignment of the Als2cL family with the RhoGEF domain family. Residues that are conserved between the two families are highlighted in yellow, those found in only one of them are highlighted in blue and green, respectively. Essential, functional residues retrieved from the CD database are indicated by hash keys. Accession numbers of sequences can be found in Supplemental [Table S7](#pone.0017568.s010){ref-type="supplementary-material"}. :::  ::: Discussion {#s3} ========== With the HMMerThread method we attempt to provide conserved domain predictions beyond the statistical threshold of purely sequence-based methods. By relaxing significance thresholds of sequence-based conserved domain searches and selecting for true positive hits by subsequent fold recognition, we can go far into and beyond the twilight zone of sequence similarity to detect remotely conserved domain members. We can significantly improve the precision of weak conserved domain predictions by cross-species validation of HMMerThread hits. The new implementation of HMMerThread shows a clearly superior performance to our previously published version of the software. We have raised the accuracy of our predictions from 79% to 90% and in this surpass existing methods of genome-scale detection of remote conservation between proteins that are either based on sequence, or structure alone. We could positively identify a number of remotely conserved domains previously reported in literature. We have discussed a number of highly interesting examples of weak, conserved domain hits that are associated with specific functional screens, cellular processes or human diseases. Our predictions could explain in part the observed phenotypes and open up new avenues for experimental studies. In this, the HMMerThread resource provides a rich resource of sensitive, functional annotations of proteins for all major model organisms. In the human proteome, we found ∼12000 remotely conserved domains with an E-value above 0.1. Of those ∼4500 could be validated in at least one other species and ∼2700 were validated against 2 species. This data enhances greatly the ability to functionally characterize many proteins and demonstrates that our knowledge of protein functions can be increased based on more sensitive searches against the current databases. One improvement to the previous version of HMMerThread was the implementation of a reliable scoring scheme for HMMerThread hits so that the software could be applied to entire genomes without manual interference. In contrast, the Z-score of OpenProspect does not effectively discriminate between true- and false- positives and the confidence measure the authors describe for Prospect II [@pone.0017568-Kim1] is not incorporated in the available OpenProspect software version. For the new scoring approach in HMMerThread, we have taken into consideration not only the Z-score of the threading run, but also the number of structures positively identified within a given Z-score threshold. Our data suggests that this combined p-value has strong discriminative power to distinguish between false positive and true positive hits. However, the strength of the p-value we derive from the hypergeometric distribution and therefore also of the combined p-value depends on the number of structures associated with a conserved domain. In order to improve the quality and the reliability of our conserved domain predictions, we chose to validate remotely conserved domains by confirming their presence in the orthologs of other, related species. We found that this is a very good measure for the reliability of weak conserved domain predictions and employed this strategy when using HMMerThread domains for annotating proteins in genome-wide screens. This procedure however is highly dependent on a) the availability of the complete and annotated genome of at least one related species and b) the quality of the genome annotations. We do not try to predict genes in genomic sequences and are relying on the predicted CDS provided by genome databases. Clearly not all genes are correctly predicted, if predicted at all, in less common model organisms such as the mosquito *Anopheles gambiae*, *Fugu rubripes* or chicken. Moreover, we do not find any usable proteome information for the close relatives of some of the chosen model organisms, like *Schizosaccharomyces pombe* or *Dictyostelium discoideum*. In these cases, we do not have any validation information based on orthologous sequences. These problems will however be solved in the future, as more genomes are being sequenced and as the annotation status of genomes from non-model organisms improves over time. A second approach we take is to validate remotely conserved domains by looking for the presence of essential residues provided by the CDD resource [@pone.0017568-MarchlerBauer1]. This method is often restricted by a lack of annotation and - in many cases -- by the lack of knowledge on functionally critical residues. This verification step however, can be of high value, as it can be used to discriminate between a certain sequence just adopting a particular fold rather than actually fulfilling the associated function(s). Contrary to purely structure-based techniques, HMMerThread can only detect remotely conserved domains, whose structure has been solved. Given this fact, we limit ourselves to conserved domains that have an associated three-dimensional structure. We currently can cover about 35% of the conserved domain sequence space. With newly solved three-dimensional structures, we can update our database with low effort, as we can specifically look for newly added structures. Due to the much smaller database sizes, we can therefore greatly reduce the required processing time for updates. Another limitation we have chosen to accept is to ignore overlapping conserved domains and limit the fold recognition step to the top hit of the HMMER2 search. This was purely due to limitations in computational resources. Based on statistics from the yeast proteome, in which we have attempted to discriminate the true positive conserved domain hit in a set of overlapping domains, we estimate that we miss roughly 54% true positive hits in the other organisms, which we could only retrieve through an eight-fold increase in run-time. This data again demonstrates the power of our approach, as in more than 50% of the cases, the true positive hit is not the first one that is detected in the sequence-based search. We are currently working on an updated version of the database that includes overlaps in all organisms presented. Finally, updating of the HMMerThread database with novel software releases will result in the highest cost concerning computing time. Meanwhile, the Pfam24 database has been released and we have included this data in our resource. Live searches using HMMerThread already use the new release of the Pfam database. We will furthermore incorporate future releases of Pfam in updates of the HMMerThread database and we will do the same for the fold library, SCOP. Likewise, we will use HMMER3 for updates, once it is out of beta testing. Remotely conserved HMMerThread domains in the HMMerThread resource are a valuable guideline for further experimental studies of protein function. Often, a remotely conserved HMMerThread domain is the sole information available for a protein under study and it provides clues for experimental design to elucidate the mechanistic function of a protein. Moreover, HMMerThread has demonstrated high precision, recall and accuracy. Yet, it is clear that conserved domain prediction based on weak sequence similarity is essentially a prediction and will need further verification. Moreover, as we partly rely on fold recognition, HMMerThread predictions have to be considered as clan-based predictions of conserved domains. All remotely conserved domains that are discussed in this manuscript have been verified by independent methods like PSI-BLAST, profile-profile comparisons or pure fold recognition using algorithms other than Prospect II/OpenProspect. We suggest, when analyzing proteins in low throughput, to use remotely conserved HMMerThread domains as a starting point for functional prediction and -- especially when looking at remotely conserved domains with very low sequence similarity -- to proceed with an independent verification step. We are currently working on a downloadable version of the HMMerThread tool. Provided that validation data can be generated from a related species, HMMerThread will prove to be a highly useful approach for sensitive conserved domain annotation in entire genomes. Materials and Methods {#s4} ===================== HMMerThread application {#s4a} ----------------------- HMMerThread was implemented using Pfam Release 22 and SCOP release 1.71. The pipeline was implemented in 4 components and in the Perl 5.8 scripting language without dependencies to allow for execution on various HPC platforms. The 4 components include domain searches (HMMER2.3.2, [@pone.0017568-Eddy1]), pre-processing (PSIPred secondary structure prediction [@pone.0017568-Jones1], SEG detection of low complexity [@pone.0017568-Wootton1], NCoils detection of coilded-coil regions [@pone.0017568-Lupas1]), threading (OpenProspect [@pone.0017568-Kim1]), post-processing (scoring). Domain search {#s4b} ------------- The first step of the HMMerThread pipeline is to search for Pfam domains in the full-length sequence. Genome-wide runs were done using an HMMER2 in global search mode and with an expect value threshold of 50. Once identified domains were extracted, they were ranked according to their e-values. Overlapping domains were removed leaving only the best scoring domains for each region of the sequence. For all conserved domains with score higher than 1e-04, the PDBMAP was consulted to see if a structure exists for the given conserved domain. If a structure was present, the domain was sent to pre-processing. Pre-Processing {#s4c} -------------- For pre-processing, secondary structure prediction with PSI-Pred was performed on the entire protein sequence. After this, SEG and NCoils were run to remove regions of low complexity and coiled coils from the input sequence. Data from these 3 programs were collated into a single sequence (with "X" for regions of low complexity and coiled coils) and the domains from the domain search step were retrieved from of the pre-processed sequence to be sent for threading. Threading {#s4d} --------- Threading was performed with OpenProspect on the input sequences from pre-processing. Searches were done on a high-performance computing (HPC) system. Settings included the use of "full" Z-scores (option -zscore\_full) and 100 Z-cycles (option -zcycles 100). Runtime for an input file varied in accordance with the sequence length. The average runtime was ∼ 4 hours on a single core of a 2.6 GHz AMD x85 Opteron processor. HPC was provided by the ZIH (TU-Dresden) in the form of a PC Farm of 2,584 cores. The processing of all 8 species took ∼3,000,000 CPU hours including cross-species validation. Post-processing {#s4e} --------------- Post-Processing was performed in 2 steps. Firstly, the results of the Threading run were processed. This involved extracting the key parameters (Z-Score and position) from the output file. These parameters were ranked producing a hit list for all SCOP domains (12,430 domains for SCOP 1.71). Secondly, scoring was performed on this ordered list based on two factors. The first was on the p-value from the naïve probability generated from a cumulative distribution function (CDF) of the Z-Scores from 1,000 OpenProspect runs (∼12 million Z-Scores) and the second is based on the cumulative probability from a hypergeometric distribution. Therefore: where pCDF is the p-value of the Z-Score from the scoring of expected structure, N  =  all structures threaded, n  =  top 60 structures threaded, m  =  expected structures threaded and k  =  expected structures in hit list. The p-value threshold for considering a domain as a hit was 0.001. Validation through orthologs {#s4f} ---------------------------- Validation through orthologs relied first on the accurate detection of orthologs. This was possible through the use of the Homologene database (where available, [@pone.0017568-Sayers1]) and the Inparanoid 2.0 software [@pone.0017568-Remm1]. Once orthologs were determined, the orthologous sequences were submitted to the HMMER2 domain search with a higher expect value threshold of 100. If the domain that was found in the original species was also detected in the validation sequence, this region was sent for fold recognition. The scoring procedure for threading was identical to that of the genome-wide runs. If the domain did also positively score in the close ortholog, it was marked as such. The species, score and original hit information were retained for storage in the database. Validation through essential, functional residues {#s4g} ------------------------------------------------- Data on functional residues was taken from the CD database. For each HMMerThread weakly conserved domain hit, the corresponding CD alignment was obtained through the use of RPS-BLAST against the CDD [@pone.0017568-MarchlerBauer1]. For this alignment, each functional residue was evaluated against the expected functional residue from the CD consensus alignment. Residues were marked as 1) identity if they were the same, 2) similarity, if they had a positive score from comparison in the BLOSUM62 matrix, 3) null, if they do not fall into the first two categories. For the evaluation, a threshold of 25% similarity was used. Calculating recall, precision and accuracy of HMMerThread (old and new), GenTHREADER and Superfamily {#s4h} ---------------------------------------------------------------------------------------------------- To evaluate the performance of HMMerThread, two versions of the Pfam database were obtained: Pfam10 (July 2003) and Pfam22 (July 2007). Common conserved domains seen in both versions were extracted resulting in 5248 domains. The resulting HMM databases were calibrated and searched against the human proteome set (RefSeq, September 2007) using hmmpfam. Conserved domains were selected, if they scored \< 0.1 in Pfam22, \>0.1 in Pfam10 and had an e-value difference of greater than 10 fold. These conserved domains were considered to be True Positives (TPs) in the following analysis. For each conserved domain region, HMMerThread was run with all overlapping domains enabled (up to an e-value of 50) against the Pfam10 profiles. This provided us with ∼1520 potential domains for HMMerThread to distinguish between True Positives (TPs), False Positives (FPs), False Negatives (FNs) and True Negatives (TNs). The True Negative dataset was derived from Pfam10∶22 searches that did not score significantly in either Pfam10 or Pfam22 according to our criteria. Clan members were furthermore excluded from false positive calculations. Formulas for TPs, FNs, FPs and TNs were as follows: TP  =  (BestHit ∈ 10∶22PosDS) && (p-value\< = 0.001) FN  =  (BestHit ∈ 10∶22PosDS) && (p-value\>0.001) FP  =  (BestHit ∉ 10∶22PosDS) && (p-value\< = 0.001) TN  =  (BestHit ∉ 10∶22PosDS) && (p-value\>0.001), whereby BestHit is the conserved domain hit discovered as top hit in HMMerThread, 10∶22PosDS are all conserved domains from the Pfam10∶22 dataset qualifying as true positives (see above), and p-value represents the combined probability developed for scoring HMMerThread hits. Accuracy, recall and precision were determined as follows: Accuracy  =  (TP+TN)/(TP+TN+FP+FN); Recall  =  TP/(TP+FN); Precision  =  TP/(TP+FP). The old version of HMMerThread was used on the same dataset with standard settings and a hit-depth of 25. The p-value of positively identified conserved domains was set to 0.0000001, negatively identified domains received a p-value of 1. The according p-values were used for calculating TPs, FPs, TNs and FNs. All other procedures were done as described as above. The local version of Genthreader (pgen 8.2) was used on the Pfam10∶22 remotely conserved domain set. PSIPred 2.5 was used with the uniref90 database for secondary structure prediction. Threading was performed against the SCOP fold library provided from 20th July 2009. In order to score conserved domains, SCOP structures that were scored as "CERTS" were mapped to Pfam domains using the PDBMAP mapping provided by Pfam. These were then compared directly with the Pfam10∶22 list as with HMMerThread (old and new versions). Calculations of TPs, FPs, TNs and FNs followed the above formulas, except that instead of the p-value, the presence (as a CERT domain) or absence of a domain was used as the second criteria. CLAN members were again excluded from false positive calculations. The local version of Superfamily was downloaded from the Superfamily website and setup according to the instructions on the site using HMMER 2.3.2 as the HMM search program and SCOP 1.73 as the models database. Processing the sequences produced 766 unique superfamily hits, most of which were scored via multiple superfamily models. In order to determine domain level hits, the SCOP IDs from the models that encompassed the hits for each superfamily were used to map to Pfam domains through the PDBMAP mapping provided by Pfam. True positives were scored, if any of the pfam-IDs associated with a superfamily were identified. All other related Pfam families found in the same region were excluded from false positive calculations next to the CLAN members also used in GenThreader and HMMerThread searches. Formulas for calculating TPs, FPs, TNs and FNs followed the above formulas, again using the presence or absence of a domain as the second criteria. HMMerThread Database {#s4i} -------------------- The HMMerThread Database was implemented in Python 2.4 and MySQL. The web-service is provided by Apache. Annotation {#s4j} ---------- Annotation for the database was obtained from species specific sources either from ftp downloads or from HTTP downloads. Annotation in the form of InterPro Domains [@pone.0017568-Mulder1] and CDD domains [@pone.0017568-MarchlerBauer1] were obtained by running the stand-alone applications InterProScan [@pone.0017568-Zdobnov1] and RPS-BLAST [@pone.0017568-MarchlerBauer1] against the sequences in the database. Databases used include NCBI [@pone.0017568-Pruitt1], SGD [@pone.0017568-Cherry1], Wormbase [@pone.0017568-Stein1], Flybase [@pone.0017568-Tweedie1], PombeDB [@pone.0017568-HertzFowler1]. HPRD Overlay {#s4k} ------------ Protein-protein interactions from the HPRD were extracted for each *H. sapiens* protein in the HMMerThread database. For each interaction partner, conserved domains (HMMerThread and Pfam) were matched with known domain-domain interactions from iPfam [@pone.0017568-Finn3]. If domains in each of the proteins were known to interact, these are displayed as the potential interaction surface that explains the protein-protein interaction in the HPRD. Live HMMerThread {#s4l} ---------------- The *live* version of HMMerThread was implemented in Python 2.4. The only difference to the Perl implementation is that all of the steps for processing are combined and the handling of web jobs is added. Furthermore, the HTML output capability was added directly in a manner similar to the HMMerThread Database. The HMMerThread *live runner* uses the additional threading module for Python to allow for the submission of jobs simultaneously on different threads. Furthermore, it relies on the SMP capabilities of PSI-BLAST (4 CPUs) and HMMER2 (4 CPUs) along with the MPI implementation in OpenProspect (32 CPUs) to reduce the runtime of the jobs. Other bioinformatics methods {#s4m} ---------------------------- PSI-BLAST searches [@pone.0017568-Altschul1] and Phyre runs [@pone.0017568-Kelley2] were carried out using standard settings. hhpred searches [@pone.0017568-Soding1] were carried out using only orthologs of the analyzed families shown in [Figures 4](#pone-0017568-g004){ref-type="fig"} to 7. Multiple sequence alignments were done using ClustalW [@pone.0017568-Thompson1] and/or Mafft [@pone.0017568-Katoh1] and manually refined. Figures were prepared in Illustrator. Multiple sequence alignments of conserved domain families were taken from NCBI CD-database [@pone.0017568-MarchlerBauer1]. For comparison between Pfam22, Pfam24 and the two HMMER releases (HMMER2.3.2 and HMMER3b3), we removed the top 20, promiscuous conserved domains, as well as Zinc Finger domains for the analysis, as HMMerThread has difficulties of identifying the correct family of Zinc Fingers. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### **Cumulative Distribution Function of threading Z-scores.** When a hypergeometric p-value threshold \< 0.05 is used, 90% of the expected conserved domain structures fall within the top 60 positions of threading hits with a Z-score \<2.38. (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### **ROC curve of the performance of the HMMerThread algorithm.** True positives were plotted against false positive predictions of the HMMerThread algorithm. The optimal p-value range corresponds to our chosen cutoff (1E-03), resulting in 14% false positive rate and a 95% true positive rate (see also supplemental [Table S3](#pone.0017568.s006){ref-type="supplementary-material"}) (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S3 ::: {.caption} ###### **Screenshots of the HMMerThread Database.** (**A**) Overview of one entire record in the HMMerThread database, in this case showing *H. sapiens* APPL1 with all associated annotation. Those include links to original database entries (NCBI), interaction partners, interacting domains and literature including GeneRIFs, Gene Summaries, Gene Ontology information, as well as known sequence-based domains. (**B**) HMMerThread domains image with the validated BAR domain (3 species), displayed by mouse over. The associated results of remotely conserved domains are shown in the HMMerThread hits table and the HMMer alignment of all remotely conserved HMMerThread domains are provided below the hit table. (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S1 ::: {.caption} ###### **Species used for cross-species validation of remotely conserved HMMerThread domains.** (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S2 ::: {.caption} ###### **Comparison of performance of the old and new version of HMMerThread, GenTHREADER and Superfamily.** (XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S3 ::: {.caption} ###### **False positive and true positive HMMerThread predictions using different p-value settings.** (XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S4 ::: {.caption} ###### **Conserved domains (InterProScan, HMMerThread) of hits from a genome-wide screen for cofactors of Hepatitis C Virus replication in human cells** (XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S5 ::: {.caption} ###### **HMMerThread remotely conserved domains found in DUF domains** (XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S6 ::: {.caption} ###### **list of known interactors and their interacting domains based on remotely conserved domains** (XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S7 ::: {.caption} ###### **Accession numbers of sequences used in** [**Figures 3**](#pone-0017568-g003){ref-type="fig"} **-** [](#pone-0017568-g004){ref-type="fig"} [](#pone-0017568-g005){ref-type="fig"} [**6**](#pone-0017568-g006){ref-type="fig"} **.** (XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: We thank Marino Zerial, Pavel Tomancak, Rebecca Wade, Sabine Bernauer and Michael Volkmer for helpful discussion. Our thanks also go to members of the high performance computing center (ZIH) at the Technical University of Dresden for computational support and resources, especially Guido Juckeland and Wolfgang E. Nagel. We also thank the members of the computing department of the MPI-CBG, especially Matt Boes and Jeffrey Oegema for their support on computational matters. Finally, we thank Pavel Tomancak, Gregory O\'Sullivan, Thierry Galvez, Unal Coskun, Ian Henry and Ines Wagner for critical reading of the manuscript. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**CRB was supported by BMBF Grant 0313082H \"Hepatosys,\" and VS was supported by EU Grant 027269 \"Sealife.\" This work was supported by the Max Planck Society. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: CRB BHH. Performed the experiments: CRB. Analyzed the data: CRB BHH. Contributed reagents/materials/analysis tools: CRB VS RH MSM BHH. Wrote the paper: CRB BHH. Software design: CRB VS RH. Obtained permission to use High Performance Computing Facility: MSM. [^2]: ¤a Current address: Core Bioinformatics Group, Wellcome Trust Cancer Research UK Gurdon Institute, University of Cambridge, Cambridge, United Kingdom [^3]: ¤b Current address: Bioinformatics Laboratory, Max Planck Institute for Biology of Ageing, Cologne, North Rhine-Westphalia, Germany

PubMed Central

2024-06-05T04:04:19.795132

2011-3-10

PMC3053372

Introduction {#s1} ============ Pain is a complex and multi-factorial [@pone.0017724-Julius1] trait and influenced by various and heterogeneous factors such as gender [@pone.0017724-Derbyshire1], genetic [@pone.0017724-Kim1] or environmental causes [@pone.0017724-Nielsen1]. Individual differences in pain responses [@pone.0017724-Fillingim1] have been employed as a research tool of nociceptive or nocifensive mechanisms and are contemplated as a basis for personalized therapy approaches to pain. The multitude of factors modulating pain suggests a comparative assessment of their influences. However, an experiment has a statistically significant effect, but also the size of any observed effects. In practical situations, effect sizes are helpful for making decisions. Although various effects on pain have been reported with respect to their statistical significance, a standardized measure of effect size has been rarely added. Such a measure would ease comparison of the magnitude of the effects across studies, for example the effect of gender on experimental heat pain with the effect of a genetic variant on pressure pain or clinical pain estimates. Reporting effect sizes is considered good practice when presenting empirical research findings in many fields [@pone.0017724-Nakagawa1]. In the present analysis, the effect sizes of factors currently of interest as modulators of pain, i.e., common genetic variants reportedly modulating pain ([**Table 1**](#pone-0017724-t001){ref-type="table"}), gender [@pone.0017724-Derbyshire1], [@pone.0017724-Berkley1], [@pone.0017724-Riley1] and sensitization procedures by capsaicin [@pone.0017724-Petersen1] or menthol [@pone.0017724-Hatem1] are provided. ::: {#pone-0017724-t001 .table-wrap} 10.1371/journal.pone.0017724.t001 Table 1 ::: {.caption} ###### Effect sizes, expressed as percentage of the total variance explained by the genetic factors, on pain thresholds. ::: {#pone-0017724-t001-1} Factor Polymorphism (dbSNP database number) Ref. MAF \[%\] Effect sizes on pain thresholds (percentage explained variance of total variance), recessive hereditary model ------------------------------------------------------------------------------ --------------------------------------------------------------------------------------------------------- ------------------------------------------------------- --------------------------------------- --------------------------------------------------------------------------------------------------------------- -------- -------- ------ ---------- *OPRM1* (μ-opioid receptor) rs1799971 A\>G [@pone.0017724-Fillingim2], [@pone.0017724-Ltsch4] 9.2 0.35 0.76 0.08 1.74 1.43 *OPRD1* (δ-opioid receptor) rs1042114 T\>G [@pone.0017724-Kim2] 17.2 0.04 0.2 0.45 2.02 4.74 rs2234918 T\>C [@pone.0017724-Kim2] 44.4 0.25 0.79 0.04 1.14 0 *COMT* (Cathechol-O-methyl transferase) rs4646312 T\>C [@pone.0017724-Kim3] 36.8 1.08 0.77 0.47 0.44 0.04 rs6269 A\>G [@pone.0017724-Kim3] 37.6 0.77 *1.45* 0.01 0 0.21 rs4633 C\>T [@pone.0017724-Diatchenko1] 54 2.33 0.2 0.32 1.16 1.17 rs4680 G\>A [@pone.0017724-Zubieta1], [@pone.0017724-Diatchenko2] 53.2 0.99 0.41 0.29 0.62 1.17 rs6269G/rs4633C/4818G/rs4680G [@pone.0017724-Diatchenko1] 36.4 0.77 *1.45* 0.01 0 0.21 rs6269A/rs4633T/4818C/rs4680A 50.8 0.6 *1.05* 1.2 0.71 1.9 rs6269A/rs4633C/4818C/rs4680G 8.4 \- \- \- \- \- rs4646312T/rs165722T/rs6269A/rs4633T/rs4818C/rs4680A [@pone.0017724-Kim3] 49.6 1.08 *1.28* 0.55 0.18 1.41 rs4646312C/rs165722C/rs6269G/rs4633C/rs4818G/rs4680G 34 0.51 0.5 0.03 0.08 0.23 rs4646312T/rs165722C/rs6269A/rs4633C/rs4818C/rs4680G 7.6 \- \- \- \- \- *TRPV1* (Transient receptor potential cation channel, subfamily V, member 1) rs8065080 A\>G [@pone.0017724-Kim2] 36.8 0.27 0.18 0.01 0.04 0.11 *TRPA1* (Transient receptor potential cation channel, subfamily A, member 1) rs11988795 G\>A [@pone.0017724-Kim3] 32.8 1.02 0.06 0.26 0.01 0.26 rs13255063A/rs11988795G [@pone.0017724-Kim3] 38.8 0.1 0.06 *1.42* 0.26 0.01 rs13255063A/rs11988795A 32.8 1.02 0.06 0.26 0.01 0.26 rs13255063T/rs11988795G 28.4 3.49 0.92 *3.74* 0.25 **5.91** *FAAH* (Fatty acid amide hydrolase) rs932816 G\>A [@pone.0017724-Kim3] 23.6 0.03 0.06 0.06 1.26 0.56 rs4141964 T\>C [@pone.0017724-Kim3] 42.8 0.17 0.12 0.05 0.5 0.11 rs2295633 G\>A [@pone.0017724-Kim3] 41.6 0.47 0.32 0.02 0.09 0.01 rs932816G/rs4141964T 34.4 1.37 *1.31* 0.81 0.08 0.26 rs932816G/rs4141964C 42 0.04 0.31 0.03 0.01 0.11 rs932816A/rs4141964T 22.8 0.03 0.06 0.06 1.26 0.56 rs324419C/rs2295633G 58.4 0 0.13 0.01 1.2 0.58 rs324419C/rs2295633A 22.4 0.01 0.12 0.27 0.2 0.33 rs324419T/rs2295633A 19.2 1.3 0.14 0.99 0.72 0.99 *GCH1* (GTP cyclohydrolase 1) 1 particular haplotype of 3 SNPs associated to one of 15 SNPs [@pone.0017724-Tegeder2], [@pone.0017724-Tegeder3] 16.4 1.08 *1.6* 0.89 0.45 0.63 *MC1R* (Melanocortin-1 receptor) 2 variant alleles of 29insA, 451C\>T, 478C\>T, 479G\>A, 880G\>C ("red head fair skin" phenotype, n = 2) [@pone.0017724-Mogil3] 451T: 6.4%, 478T: 6%, others: 0--0.4% 0.01 *1.09* 0.47 0 0 \# MAF: Observed minor allelic frequencies. "Minor" refers to the allele reported to be minor in gene databases. When its reported allelic frequency is close to 50%, it can happen that the "minor" allele has a frequency \>50% in the actual cohort. We nevertheless preserved the denomination "minor" to be consistent with SNP databases. The reference and the observed allelic frequencies are given, and the recessive hereditary model was used, i.e., assigning heterozygous subjects to the group of homozygous mutated carriers. The effect sizes are given in italic letters when they were larger than those of gender, and in bold letters when exceeding, arbitrarily chosen, 5%. ::: Methods {#s2} ======= Subjects and design {#s2a} ------------------- The study was conducted following the Declaration of Helsinki on Biomedical Research Involving Human Subjects. The University of Frankfurt Medical Faculty Ethics Review Board approved the study protocol. Informed written consent was obtained from all subjects. Pain thresholds to various experimental stimuli had been determined during previous assessments [@pone.0017724-Neddermeyer1], [@pone.0017724-Flhr1] in a random sample of 125 unrelated healthy caucasian volunteers (69 men, 56 women, aged 18 to 46 years, mean 25±4.4 years). Exclusion criteria were drug intake dated back less than a week except for oral anticonceptionals, an actual clinical condition involving pain, and actual diseases according to questioning and medical examination. A training session was performed prior to the actual experiments, however, without application of sensitization procedures. Employing an open non-randomized design the actual measurements (for pain models, see the following section) took place in the order cold pain, menthol application, von Frey hair pain, cold/menthol pain, heat pain, capsaicin application, electrical pain, heat/capsaicin pain, pressure pain and von Frey hair/capsaicin pain, at intervals of 3--5 min between models. Assessment of pain {#s2b} ------------------ The study assessed pain thresholds to various stimuli defined as "the least experience of pain which a subject can recognize" (<http://www.iasp-pain.org>). Pain models were applied without knowledge of the genotypes. Five different stimuli were applied to include a broad variety of thermal, mechanical and electrical pain [@pone.0017724-Neddermeyer1], [@pone.0017724-Flhr1]. In brief, **heat** stimuli were applied using a 3×3 cm thermode (Thermal Sensory Analyzer, Medoc Advanced Medical Systems Ltd., Ramat Yishai, Israel) placed onto the skin of the left volar forearm. Its temperature was increased from 32°C by 0.3°C/s until the subject pressed a button at the first sensation of pain, which triggered cooling of the thermode by approximately 1.2°C/s. Heat stimuli were applied eight times at random intervals of 25--35 s. The median of the last five responses was defined as the heat pain threshold because in previous experiments a plateau was reached after the first three measurements. **Cold** stimuli were applied at inner side of the right forearm, similarly to heat pain thresholds. The temperature was decreased from 32°C to 0°C by 1°C/s. As previous experiments had shown that measurements are stable from the first application, five repetitions were used and the threshold was the median of these measurements. **Blunt pressure** was exerted perpendicularly onto the dorsal side of mid-phalanx of the right middle finger using a pressure algometer with a circular and flat probe of 1 cm diameter (JTECH Medical, Midvale, USA). The pressure was increased at a rate of approximately 9 N/cm^2^ per second until the subject reported pain. The procedure was repeated five times at intervals of 30 s. Mechanical pain threshold to blunt pressure was the median of the five measurements. **Punctate pressure** was exerted onto the left volar forearm using von Frey hairs (0.008, 0.02, 0.04, 0.07, 0.16, 0.4, 0.6, 1, 1.4, 2, 4, 6, 8, 10, 15, 26, 60, 100, 180, 300 g; North Coast Medical Inc., Morgan Hill, CA, USA). Von Frey hairs were applied at randomized order and the pain threshold was the (log-transformed) turning point at 50% probability of a logistic regression of the "pain/no-pain" answers. During the experiments, subjects had to keep their eyes closed to avoid recognition of the von Frey hairs\' strength. **Electrical** stimuli were applied using a constant current device (Neurometer® CPT, Neurotron Inc., Baltimore, MD). It delivered sine-wave stimuli at 5 Hz applied via two gold electrodes placed on the medial and lateral side of the mid-phalanx of the right middle finger. Their intensity was increased from 0 to 20 mA by 0.2 mA/s until the subjects interrupted the current by releasing a button. Measurements were repeated five times at intervals of 30 s and the median of these measurements was submitted to statistics as the electrical pain threshold. **Sensitization** was assessed with punctate mechanical and heat [@pone.0017724-Petersen1] stimuli and obtained using capsaicin cream (0.1 g, 0.1%, manufactured by the local pharmacy) applied onto a 3×3 cm skin area on the left volar forearm and covered with a plaster for 20 min. Sensitization to cold stimuli [@pone.0017724-Hatem1] was assessed using menthol solution (2 ml of a 40% menthol solution dissolved in ethanol) applied in a soaked plaster onto a 3×3 cm skin area on the right volar forearm for 20 min. Data analysis {#s2c} ------------- To obtain the genotypes for this assessment, those single nucleotide polymorphisms (SNP) or haplotypes reported until June 2008 to modulate experimental pain in healthy average people (n = 29, [Table 1](#pone-0017724-t001){ref-type="table"}) were diagnosed by means of validated Pyrosequencing^TM^ assays. Genotypes were submitted to further analysis after verifying that the distribution of homozygous and heterozygous carriers of variants was as expected from the Hardy-Weinberg [@pone.0017724-Hardy1] law (χ^2^ goodness-of-fit tests: p\>0.05). The physical strengths of the stimuli at which the subjects\' answer to the question "Does it hurt?" changed from "No" to "Yes" were the pain threshold to the respective stimuli and were analyzed for the effect sizes of the genetic and non-genetic factors. The **portion of the total variance** in a pain threshold **explained** by a particular factor was calculated for every factor *j* (genetic variants including SNPs and *in-silico* obtained haplotypes [@pone.0017724-Stephens1], gender or sensitization by capsaicin or menthol application) as , where *SS* denotes the sum of squared deviations from the mean of the respective pain scores *j*, and the error SS describes the SS being not due to the genetic or gender factor. As sensitization involved repeated measurements, the variance explained was assessed using a resampling procedure without replacement that provided 1000 new data sets containing either the non-sensitized or sensitized thresholds from a single person, which allowed using sensitization as an inter-individual factor as gender or genetics. In meta-analysis often performed to draw standardized information about effect sizes from heterogeneous data sets, the effect sizes are being quantified by calculating **Cohen\'s d** [@pone.0017724-Cohen1] as a widely used standardized effect size measure appropriate to use in the context of a t-test on means. Specifically, genotype effects can be assessed by means of t-tests, i.e., in the dominant hereditary model by comparing carriers with non-carriers of a variant, and in the recessive hereditary model by comparing homozygous carriers of a variant with pooled heterozygous and non-carriers. Standardized group differences in parameter means were calculated as , where *m~0,j~*, *m~1,j~* and *s~0,j~*, *s~1,j~* denote the means and standard deviations of the pain scores in the carriers or non-carriers of the compared property *j*. The result is a unit-free number of which, an absolute value of d = 0.2 is regarded as a small effect, 0.5 as a medium and \>0.8 as a large effect [@pone.0017724-Cohen1]. Results {#s3} ======= The original physical stimulus strengths at which the stimuli became painful are shown in [Figure 1](#pone-0017724-g001){ref-type="fig"} (left panels). The different **genotypes** explained 0--5.9% ([**Table 1**](#pone-0017724-t001){ref-type="table"} and \>[Table 2](#pone-0017724-t002){ref-type="table"}) of the variance in these pain thresholds to the different stimuli. For example, the *GCH1* haplotype explained 4% of the interindividual variance in pressure pain thresholds while the δ-opioid receptor variant rs1042114 explained 2.5% of the variance across subjects in von Frey hair thresholds. According to Cohen\'s d ([Table 3](#pone-0017724-t003){ref-type="table"} and [Table 4](#pone-0017724-t004){ref-type="table"}), the genotype effect sizes had to be considered as mostly small (range 0--1.78). A gene dose effect resulted in somewhat larger effect sizes in homozygous carriers. However, only the *TRPA1* rs13255063T/rs11988795G haplotype explained \>5% of the variance, namely of that in electrical pain threshold. Homozygous carriers had a higher pain threshold to electrical stimuli (3.8±2 mA versus 2.5±1.2 mA, p = 0.006). ::: {#pone-0017724-g001 .fig} 10.1371/journal.pone.0017724.g001 Figure 1 ::: {.caption} ###### Observed thresholds to different pain stimuli and sizes of modulatory effects. **Left** part: Single values of the measured pain thresholds to various stimuli are shown as dots, with statistical summaries in overlaid box plots. The boxes span the 25^th^ to 75^th^ percentiles, with the median crossing the box as a horizontal line, and the whiskers spanning values within 1.5 times the 25^th^ to 75^th^ percentiles. The subject\'s gender is indicated by different symbols and colors (men: red circles, women: blue crosses). At the **right** of each thresholds presentation, the effect sizes of the genetic variants obtained using the dominant hereditary model (blue filled circles), i.e., heterozygous and homozygous carriers of the variant alleles versus wild type subjects, and the recessive model (red empty circles), i.e., homozygous carriers of the variant versus the other subjects, are shown as correlation plots between the fraction of the total variance in the respective threshold explained by the respective factor and Cohen\'s d of that factor. An absolute value of d = 0.2 indicates a small effect, values around 0.5 a medium and above 0.8 a large effect [@pone.0017724-Cohen1]. In addition, the effects sizes of gender (green filled triangles) and sensitization (orange filled squares) by capsaicin (heat, von Frey hair punctate pressure) or menthol (cold) are shown. Note that the axis scaling is non-uniform among panels to enhance data visibility. At the bottom, the overall effect sizes (all Cohen\'s d per condition genetics, gender or sensitization) of all analyzed factors and stimuli are grouped for genetic, gender and sensitization influences on pain thresholds, showing decreasing sizes of effects in the order sensitization, gender and genetics. :::  ::: ::: {#pone-0017724-t002 .table-wrap} 10.1371/journal.pone.0017724.t002 Table 2 ::: {.caption} ###### Effect sizes, expressed as percentage of the total variance explained by the respective factor, on pain thresholds. ::: {#pone-0017724-t002-2} Factor Polymorphism (dbSNP database number) Effect sizes on pain thresholds (percentage explained variance of total variance), dominant hereditary model ------------------------------------------------------------------------------ --------------------------------------------------------------------------------------------------------- -------------------------------------------------------------------------------------------------------------- ------------ -------- ----------- ----------- *OPRM1* (μ-opioid receptor) rs1799971 A\>G 0.21 0.56 0.28 0.2 0.14 *OPRD1* (δ-opioid receptor) rs1042114 T\>G 2.47 0.01 1.17 0.12 0.58 rs2234918 T\>C 0.76 0.72 0.76 0.13 0.06 *COMT* (Cathechol-O-methyl transferase) rs4646312 T\>C 0.4 0.06 0.34 0.1 0.18 rs6269 A\>G 0 0.86 *3.5* 1.47 1.53 rs4633 C\>T 0.94 *1.23* 0.09 0.15 0.01 rs4680 G\>A 1.83 0.46 0.02 0.03 0.12 rs6269G/rs4633C/4818G/rs4680G 0.13 0.14 1.52 0.87 0.52 rs6269A/rs4633T/4818C/rs4680A 1.4 0.71 0.12 0.02 0.01 rs6269A/rs4633C/4818C/rs4680G 3.93 0 *1.47* 0 0.48 rs4646312T/rs165722T/rs6269A/rs4633T/rs4818C/rs4680A 1.51 0.65 0.01 0.02 0.03 rs4646312C/rs165722C/rs6269G/rs4633C/rs4818G/rs4680G 0.04 0 1.09 0.51 0.5 rs4646312T/rs165722C/rs6269A/rs4633C/rs4818C/rs4680G 4.47 0.12 1.3 0.03 0.57 *TRPV1* (Transient receptor potential cation channel, subfamily V, member 1) rs8065080 A\>G 0.03 0.85 0.29 0.63 1.37 *TRPA1* (Transient receptor potential cation channel, subfamily A, member 1) rs11988795 G\>A 0.01 0.12 0.25 0.25 1.06 rs13255063A/rs11988795G 0.1 0.25 0.22 0.31 0.93 rs13255063A/rs11988795A 0.01 0.12 0.25 0.25 1.06 rs13255063T/rs11988795G 0.84 0.12 0.2 0.01 0.1 *FAAH* (Fatty acid amide hydrolase) rs932816 G\>A 2.7 0.27 1.09 0.09 0.15 rs4141964 T\>C 0.02 0.27 0 1.06 0.62 rs2295633 G\>A 0 0.13 0.01 1.2 0.58 rs932816G/rs4141964T 0.51 0.7 0.48 0.28 0.01 rs932816G/rs4141964C 0.02 0.27 0 1.06 0.62 rs932816A/rs4141964T 3.38 0.47 1.03 0.05 0.15 rs324419C/rs2295633G 0.47 0.32 0.02 0.09 0.01 rs324419C/rs2295633A 0.08 0.04 *1.62* 0.34 0.39 rs324419T/rs2295633A 0.02 0.74 1.16 0.96 0.58 *GCH1* (GTP cyclohydrolase 1) 1 particular haplotype of 3 SNPs associated to one of 15 SNPs 0.38 0.3 0.35 3.97 0 *MC1R* (Melanocortin-1 receptor) 2 variant alleles of 29insA, 451C\>T, 478C\>T, 479G\>A, 880G\>C ("red head fair skin" phenotype, n = 2) **Gender** **5.87** 0.95 1.3 **14.75** **10.27** **Sensitization** ***7.46*** ***62.6*** *4.63* \# MAF: Observed minor allelic frequencies. "Minor" refers to the allele reported to be minor in gene databases. When its reported allelic frequency is close to 50%, it can happen that the "minor" allele has a frequency \>50% in the actual cohort. We nevertheless preserved the denomination "minor" to be consistent with SNP databases. In the case of the genetic factors, the reference and the observed allelic frequencies are given, and the dominant hereditary model was used, i.e., assigning heterozygous subjects to the group of wild-type carriers. The effect sizes are given in italic letters when they were larger than those of gender, and in bold letters when exceeding, arbitrarily chosen, 5%. ::: ::: {#pone-0017724-t003 .table-wrap} 10.1371/journal.pone.0017724.t003 Table 3 ::: {.caption} ###### Effect sizes, expressed as absolute values of Cohen\'s d [@pone.0017724-Cohen1], of the genetics factors on pain thresholds. ::: {#pone-0017724-t003-3} Factor Polymorphism (dbSNP database number) Effect sizes on pain thresholds (Cohen\'s d), recessive hereditary model ------------------------------------------------------------------------------ --------------------------------------------------------------------------------------------------------- -------------------------------------------------------------------------- ---------- ---------- ------------ ------------ *OPRM1* (μ-opioid receptor) rs1799971 A\>G \- \- \- \- \- *OPRD1* (δ-opioid receptor) rs1042114 T\>G 0.17 *0.35* *0.54* ***1.14*** ***1.78*** rs2234918 T\>C 0.13 *0.23* 0.05 0.28 0.00 *COMT* (Cathechol-O-methyl transferase) rs4646312 T\>C 0.30 *0.26* 0.20 0.19 0.06 rs6269 A\>G 0.26 *0.35* 0.03 0.01 0.13 rs4633 C\>T 0.34 0.10 0.12 0.24 0.24 rs4680 G\>A 0.22 0.14 0.12 0.18 0.24 rs6269G/rs4633C/4818G/rs4680G 0.26 *0.35* 0.03 0.01 0.13 rs6269A/rs4633T/4818C/rs4680A 0.18 *0.24* *0.25* 0.19 0.32 rs6269A/rs4633C/4818C/rs4680G \- \- \- \- \- rs4646312T/rs165722T/rs6269A/rs4633T/rs4818C/rs4680A 0.24 *0.27* 0.17 0.10 0.28 rs4646312C/rs165722C/rs6269G/rs4633C/rs4818G/rs4680G 0.23 *0.23* 0.06 0.09 0.15 rs4646312T/rs165722C/rs6269A/rs4633C/rs4818C/rs4680G \- \- \- \- \- *TRPV1* (Transient receptor potential cation channel, subfamily V, member 1) rs8065080 A\>G 0.16 0.13 0.03 0.06 0.10 *TRPA1* (Transient receptor potential cation channel, subfamily A, member 1) rs11988795 G\>A 0.36 0.09 0.18 0.03 0.18 rs13255063A/rs11988795G 0.10 0.07 *0.36* 0.15 0.03 rs13255063A/rs11988795A 0.36 0.09 0.18 0.03 0.18 rs13255063T/rs11988795G *0.74* *0.37* *0.76* 0.19 ***0.97*** *FAAH* (Fatty acid amide hydrolase) rs932816 G\>A 0.06 0.09 0.09 0.40 0.27 rs4141964 T\>C 0.10 0.08 0.05 0.17 0.08 rs2295633 G\>A 0.17 0.14 0.04 0.08 0.03 rs932816G/rs4141964T 0.37 *0.36* *0.28* 0.08 0.16 rs932816G/rs4141964C 0.05 0.14 0.04 0.02 0.08 rs932816A/rs4141964T 0.06 0.09 0.09 0.40 0.27 rs324419C/rs2295633G 0.00 0.07 0.02 0.23 0.16 rs324419C/rs2295633A 0.06 *0.28* *0.41* 0.35 0.46 rs324419T/rs2295633A *0.58* 0.19 *0.51* 0.44 0.51 *GCH1* (GTP cyclohydrolase 1) 1 particular haplotype of 3 SNPs associated to one of 15 SNPs *0.6* *0.72* *0.54* 0.38 0.45 *MC1R* (Melanocortin-1 receptor) 2 variant alleles of 29insA, 451C\>T, 478C\>T, 479G\>A, 880G\>C ("red head fair skin" phenotype, n = 2) 0.06 **0.84** **0.54** 0.05 0.03 The recessive hereditary model was used, i.e., assigning heterozygous subjects to the group of homozygous mutated carriers. The effect sizes are given in italic letters when they were larger than those of gender, and in bold letters when exceeding a value of 0.8 indicating a large effect. ::: ::: {#pone-0017724-t004 .table-wrap} 10.1371/journal.pone.0017724.t004 Table 4 ::: {.caption} ###### Effect sizes, expressed as absolute values of Cohen\'s d [@pone.0017724-Cohen1], of the respective factor on pain thresholds. ::: {#pone-0017724-t004-4} Factor Polymorphism (dbSNP database number) Effect sizes on pain thresholds (Cohen\'s d), dominant genetic model ------------------------------------------------------------------------------ --------------------------------------------------------------------------------------------------------- ---------------------------------------------------------------------- ------------ -------- ---------- ------ *OPRM1* (μ-opioid receptor) rs1799971 A\>G 0.12 *0.20* 0.14 0.12 0.10 *OPRD1* (δ-opioid receptor) rs1042114 T\>G 0.34 0.02 *0.23* 0.07 0.16 rs2234918 T\>C 0.19 0.19 0.19 0.08 0.06 *COMT* (Cathechol-O-methyl transferase) rs4646312 T\>C 0.13 0.05 0.12 0.06 0.09 rs6269 A\>G 0.01 0.19 *0.39* 0.25 0.26 rs4633 C\>T 0.24 *0.27* 0.07 0.09 0.02 rs4680 G\>A 0.34 0.17 0.04 0.04 0.08 rs6269G/rs4633C/4818G/rs4680G 0.07 0.08 *0.25* 0.19 0.15 rs6269A/rs4633T/4818C/rs4680A 0.28 *0.20* 0.08 0.03 0.02 rs6269A/rs4633C/4818C/rs4680G *0.54* 0.01 *0.33* 0.00 0.19 rs4646312T/rs165722T/rs6269A/rs4633T/rs4818C/rs4680A 0.29 0.19 0.02 0.03 0.04 rs4646312C/rs165722C/rs6269G/rs4633C/rs4818G/rs4680G 0.04 0.00 0.21 0.15 0.14 rs4646312T/rs165722C/rs6269A/rs4633C/rs4818C/rs4680G *0.60* 0.10 0.32 0.05 0.21 *TRPV1* (Transient receptor potential cation channel, subfamily V, member 1) rs8065080 A\>G 0.04 0.19 0.11 0.16 0.24 *TRPA1* (Transient receptor potential cation channel, subfamily A, member 1) rs11988795 G\>A 0.02 0.07 0.10 0.10 0.21 rs13255063A/rs11988795G 0.07 0.10 0.10 0.12 0.20 rs13255063A/rs11988795A 0.02 0.07 0.10 0.10 0.21 rs13255063T/rs11988795G 0.18 0.07 0.09 0.02 0.06 *FAAH* (Fatty acid amide hydrolase) rs932816 G\>A 0.34 0.11 0.22 0.06 0.08 rs4141964 T\>C 0.03 0.11 0.00 0.22 0.17 rs2295633 G\>A 0.0 0.07 0.02 0.23 0.16 rs932816G/rs4141964T 0.15 0.17 0.14 0.11 0.02 rs932816G/rs4141964C 0.03 0.11 0.00 0.22 0.17 rs932816A/rs4141964T 0.39 0.14 0.21 0.05 0.08 rs324419C/rs2295633G 0.17 0.14 0.04 0.08 0.03 rs324419C/rs2295633A 0.04 0.08 *0.28* 0.12 0.09 rs324419T/rs2295633A 0.02 0.14 0.26 0.21 0.2 *GCH1* (GTP cyclohydrolase 1) 1 particular haplotype of 3 SNPs associated to one of 15 SNPs 0.13 0.12 0.13 0.45 0.00 *MC1R* (Melanocortin-1 receptor) 2 variant alleles of 29insA, 451C\>T, 478C\>T, 479G\>A, 880G\>C ("red head fair skin" phenotype, n = 2) \- \- \- \- \- **Gender** 0.50 0.20 0.23 **0.84** 0.68 **Sensitization** *0.53* ***2.60*** *0.44* In the case of the genetic factors, the dominant hereditary model was used, i.e., assigning heterozygous subjects to the group of wild-type subjects. The effect sizes are given in italic letters when they were larger than those of gender, and in bold letters when exceeding a value of 0.8 indicating a large effect. ::: **Gender** produced explained 1--14.75% of the variance in the different pain thresholds (Cohen\'s d 0.2--0.84). The comparatively greatest fractions of the variance explained by gender were seen for blunt pressure and electric stimuli. The gender effect was directed toward higher pain sensitivity in women than in men. **Sensitization** by capsaicin increased the pain thresholds to heat and punctate pressure (Wilcoxon tests: p\<0.001) whereas menthol decreased the cold pain thresholds (Wilcoxon test: p\<0.001). Heat sensitization by capsaicin explained 63% of the total variance and produced the larges effects observed in this data according to Cohen\'s d = 2.6. Sensitizations by capsaicin or menthol of punctate or cold pain explained 7.5 or 4.6% of the variance in those thresholds and produced medium to small effect sizes (Cohen\'s d = 0.53 and 0.4, respectively). Discussion {#s4} ========== Pain thresholds were subject to various influences, which was most readily visible for capsaicin sensitization of heat pain perception and to a smaller extend of menthol sensitization for cold pain thresholds ([Figure 1](#pone-0017724-g001){ref-type="fig"} last two lines). The basis of this large effect on heat pain thresholds is the synergistic effects of the excitations of TPRV1 by both, heat (\>43°C) and capsaicin [@pone.0017724-Caterina1], [@pone.0017724-Davis1]. As TRPV1 receptor potential channels are also considered general nocisensors [@pone.0017724-Patapoutian1], the effect on punctate mechanical pain has a similar explanation although the sensitization had smaller effects than on heat pain. Analogously, the effect of menthol sensitization on cold pain thresholds can be explained by a concomitant excitation of TRPM8 by both, cold stimuli between 8 and 28°C and menthol [@pone.0017724-McKemy1], [@pone.0017724-Peier1], [@pone.0017724-Colburn1]. A part of the total variance in pain thresholds was accounted for by the subject\'s gender, exceeding 1/10 for blunt pressure and electrical pain stimuli. Gender effects on pain have been established for long and their present direction toward higher pain sensitivity in women than in men agrees with most studies (for reviews, see [@pone.0017724-Derbyshire1], [@pone.0017724-Berkley1], [@pone.0017724-Riley1]). Explanations use sex hormones [@pone.0017724-Kowalczyk1], [@pone.0017724-Rao1] or differences in the function of the endogenous opioid system [@pone.0017724-alAbsi1] such as a sexual dimorphism regarding opioid receptor function in rat brain structures mediating opioid analgesia [@pone.0017724-Tershner1]. Interaction of sex and genetics may follow from sex differences in the functioning of, e.g., μ- and δ-opioid receptors, COMT or FAAH [@pone.0017724-Snidvongs1]. Sex-differences in the response to exogenous opioids in rats were reported to depend on the genotype [@pone.0017724-Mogil1]. A sex by genotype interaction emerged for heat pain ratings with respect to the human *OPRM1* 118A\>G polymorphism [@pone.0017724-Fillingim2] and thermal pain sensitivity was also modulated by gender, ethnicity and psychological factors [@pone.0017724-Kim2]. Genetic factors contributed to the explanation of the overall variance in pain thresholds. The effect sizes agreed with few elsewhere reported effect sizes such as those of variants in *COMT* or *FAAH* explaining 5--8% of the variance in experimental pain measures [@pone.0017724-Kim3], or that of *OPRM1* rs1799971 of approximately Cohen\'s d = 0.3 for heat and ischemic pain thresholds or tolerance [@pone.0017724-Fillingim2]. The largest genetic effect size in the present data was seen for homozygous presence of the *TRPA1* rs13255063T/rs11988795G haplotype explaining \>5% of the variance in electrical pain thresholds. This genetic effect exceeded that of menthol sensitization on cold stimuli. The cold-sensitive TRPA1 receptor potential channel is mainly activated by noxious cold, chemical and endogenous irritants [@pone.0017724-Bandell1]. A decrease in cold pain withdrawal time associated with *TRPA1* rs1198795 had been observed in another study [@pone.0017724-Kim3]. A difficulty to explain the contrasting result arises from lack of shown molecular consequences of the genetic polymorphisms. As TRPA1 is a pain sensor [@pone.0017724-Patapoutian1], the results with cold pain [@pone.0017724-Kim3] point at an increased function associated with the rs1198795 variant. Increased function is conveyed by another TRPA1 mutation (N855S [@pone.0017724-Kremeyer1]) and associated with a rare autosomal-dominant familial syndrome characterized by episodes of debilitating upper body pain. However, the present observations of decreased pain sensitivity in carriers of a *TRPA1* haplotype, of which rs1198795 is a part, point at a decreased function of TRPA1 nocisensors associated with the frequent variants analyzed here. A few additional genetic variants modulating pain have not reached the present set of genotypes. This relates to potassium ion channels Kir3.2, for which the genetic variant *KCNJ6* rs2070995 increased opioid requirements [@pone.0017724-Nishizawa1], [@pone.0017724-Ltsch1], and Kv~9.1~ for which the genetic *KCNS1* variants rs734784 and rs13043825 were associated with greater pain [@pone.0017724-Costigan1]. There is no strong indication that their inclusion would have changed the picture of mostly small genetic effect sizes, and information so far only shows a modulation of clinical pain including neuropathic pain but not experimental pain. The latter applies also to interleukin related genetic modulations of pain [@pone.0017724-Solovieva1] and other genetically polymorphic nociceptive factors [@pone.0017724-Tegeder1]. Since pain is defined as "an unpleasant sensory and emotional experience ..." (International Association for the Study of Pain, <http://www.iasp-pain.org>), it cannot be measured directly. Correctly, the pain threshold is defined at a perceptional level as the least experience of pain which a subject can recognize. In contrast, the present pain threshold measures comprise the physical intensities of the stimuli at which the subjects indicated that they became painful. Therefore, the genetic factors, sex or sensitization have in fact modulated the lowest stimulus\' strength at which the subject indicated pain. The measurement of pain has been addressed since more than half a century [@pone.0017724-Beecher1]. Quantitative information about a subject\'s pain may be obtained with several other methods [@pone.0017724-Huskisson1]. For example, subjects may indicate the stimulus strength evoking unbearable pain (pain tolerance) or rate their pain on nominal or analog scales. In search for an objective quantification of pain, several surrogate measures have been established, such as pain-related evoked cortical electrical potentials, magnetic fields blood oxygen dependent signal, or muscle reflexes [@pone.0017724-Handwerker1]. The observed small effect sizes suggest that none of the tested common factors suffices as a basis for clinical decisions or prognostic judgments with respect to pain. This may be similar for experimentally induced and clinical pain as the genetic effect of some variants has been demonstrated in both. For example, the so-called "pain-protective" *GCH1* haplotype decreased pain in healthy volunteers following administration of mechanical, heat and ischemic pain [@pone.0017724-Tegeder2] or the same pain models as presently used [@pone.0017724-Tegeder3], and it was associated with lower clinical pain following surgical discectomy [@pone.0017724-Tegeder2] and delayed development of pain from the cancer diagnosis [@pone.0017724-Ltsch2]. The poor effect size of common genetic factors is reminiscent of other multigenetic traits such as body height or type 2 diabetes, for which genome wide association studies have mainly shown that the effects of single common genetic variants on the phenotype are small [@pone.0017724-Goldstein1]. This might a major reason why genetics-based pain management advices have not emerged in clinical practice [@pone.0017724-Ltsch3], [@pone.0017724-Mogil2], similar to gender differences that have also raised expectations for individualized therapies and have also not entered the clinical guidelines. Therefore, an individualized pain therapy based on genotyping information is not yet imminent. This study did not intend to reproduce genetic associations but to provide a basis for comparison of genetics\' effects on pain with other effects across different studies and pain measurements. Therefore, the report was limited to effect sizes, which have become a standard part of reporting [@pone.0017724-Thompson1]. Standardized effects sizes enhanced comparison across the pain stimuli and were therefore preferred. However, if the units of measurement are meaningful on a practical level, then reporting an unstandardized measure has been advised [@pone.0017724-Wilkinson1]. In the present data, the difference in the physical strength of each stimulus at which it evokes pain has a practical meaning, i.e., N/m^2^ for blunt pressure or mA of 5-Hz electrical sine waves. However, the comparison across stimuli is probably more meaningful when using standardized effect sizes. This preserves the relative order of factors. However, standardizing loses the direction of the effect. The "canned" effect sizes "small", "medium", or "large" should not replace a decision about how large a difference is that is based on understanding of the experimental system [@pone.0017724-Lenth1]. On several pain stimuli, heat sensitization by capsaicin, gender and genetic variants produced effects on pain in the mentioned order of effect sizes ([Figure 1](#pone-0017724-g001){ref-type="fig"} bottom). Reporting effect sizes is considered good practice when presenting empirical research findings in many fields and the present report may provide a basis for comparative discussions of factors influencing pain. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**The authors have no support or funding to report. [^1]: Conceived and designed the experiments: JL. Performed the experiments: AD KF TJM. Analyzed the data: NK GS JL. Wrote the paper: JL.

PubMed Central

2024-06-05T04:04:19.801972

2011-3-10

PMC3053373

Introduction {#s1} ============ The retinoblastoma tumor suppressor (Rb) plays crucial roles in development and homeostasis, and is commonly inactivated in human malignancies [@pone.0017682-Burkhart1], [@pone.0017682-Bremner1], [@pone.0017682-Jiang1]. Rb is a member of a family of proteins including p107 and p130 that exhibit similar or opposing functions in different tissues [@pone.0017682-Cobrinik1]. The Rb family is thought to control cell proliferation and survival by binding E2F family of transcription factors and repressing transcription by recruiting chromatin-modifying factors such as HDAC1 [@pone.0017682-Chen1]. The Rb family may also regulate differentiation by controlling expression of differentiation factors such as PPARγ and PGC-1α and by sequestering inhibitors of differentiation including ID2, HDAC1, EID1 and RBP2 [@pone.0017682-Calo1], [@pone.0017682-Scime1], [@pone.0017682-Lasorella1], [@pone.0017682-Puri1], [@pone.0017682-MacLellan1], [@pone.0017682-Benevolenskaya1]. Rb in particular was shown to potentiate the activity of lineage-specific transcription factors such as the myogenic factors MyoD and myogenin during skeletal myogenesis [@pone.0017682-Thomas1], [@pone.0017682-Chen2], [@pone.0017682-Skapek1], [@pone.0017682-Gu1], [@pone.0017682-Schneider1], [@pone.0017682-Novitch1]. These myogenic proteins bind promoters of muscle-specific genes like muscle creatine kinase (MCK) to activate the muscle differentiation program [@pone.0017682-Buckingham1], [@pone.0017682-Rudnicki1], [@pone.0017682-Olson1], [@pone.0017682-Black1]. Indeed, ectopic expression of MyoD in Rb-deficient fibroblasts fails to induce myogenesis [@pone.0017682-Sellers1], [@pone.0017682-Chen3]. In keeping with Rb\'s ability to potentiate myogenic conversion in fibroblasts, pRb is required for proper skeletal myogenesis *in vivo*. Rb-null embryos die at embryonic day (E) 13.5--14.5, exhibiting ectopic proliferation, massive apoptosis and incomplete differentiation in a number of tissues where Rb is normally highly expressed [@pone.0017682-Jiang2], [@pone.0017682-Jacks1], [@pone.0017682-Lee1], [@pone.0017682-Clarke1], [@pone.0017682-Lee2], [@pone.0017682-Mulligan1], [@pone.0017682-Macleod1]. The early embryonic death precluded studies of terminal skeletal myogenesis, which occurs after embryonic day (E) 14.5. An Rb mini-gene (mgRb), expressed exclusively in the placenta and the nervous system, but not in skeletal muscles, was used to extend the life-span of Rb^−/−^ embryos to birth [@pone.0017682-Jiang3], [@pone.0017682-Zacksenhaus1](Z. Jiang and EZ, unpublished). In mgRb:Rb^−/−^ embryos, myotubes are initially formed at E14.5--15.5, but thereafter degenerate in a process accompanied by massive myoblast apoptosis, endoreduplication within myotubes and normal expression of early myogenic differentiation markers (MHC, cardiac actin) but not late markers (MCK, MRF4). Similar muscle defects were subsequently reported after Rb-null embryos were partially rescued by other means [@pone.0017682-Lasorella1], [@pone.0017682-Takahashi1], [@pone.0017682-Wu1]. Moreover, a muscle-specific ablation of a floxed Rb allele (Rb^f^) via Myf5-Cre demonstrated impaired muscle differentiation both *in vitro* and *in vivo* [@pone.0017682-Huh1]. Finally, analysis of Rb mutant myoblasts (*in vitro*) revealed a similar pattern observed in Rb mutant fetuses (*in vivo*); that is, myoblasts initially fuse to form short myotubes but quickly degenerate and never twitch [@pone.0017682-Ciavarra1]. Although these observations implicate pRb in terminal myogenesis, a direct assignment of differentiation function to pRb was hampered by the fact that terminal differentiation is intimately coupled to myoblast survival. To overcome this obstacle, the ability of Rb mutant myoblasts to differentiate has been assessed in the presence of survival factors. Remarkably, expression of Bcl-2 rescued the Rb defect leading to long myotubes that twitched for weeks in culture [@pone.0017682-Ciavarra1]. Furthermore, differentiating Rb-deficient myotubes exhibited perinuclear mitochondrial aggregation and autophagy, not apoptosis, and inhibition of autophagy or exposure to hypoxia suppressed myotube degeneration [@pone.0017682-Ciavarra1]. Although differentiating Rb mutant myotubes initially failed to exit the cell cycle, the rescued myotubes eventually became stably post-mitotic despite absence of Rb. Together these results suggest that Rb is required to coordinate cell cycle exit with survival during the onset of differentiation, but not for actively stimulating the differentiation program. An important unresolved question is whether p107 and p130 compensate or exacerbate the differentiation defect of Rb-deficient myoblasts. In particular, we wish to know whether the initial cell fusion and myotube formation that occurs in the absence of pRb is p107 and/or p130-dependent. There are several notable differences among Rb family members, including their affinity to E2F members and other factors [@pone.0017682-Chen1], [@pone.0017682-Castano1], effects on cell cycle exit and senescence [@pone.0017682-Bruce1], , and roles during embryogenesis and cancer [@pone.0017682-Jiang2], [@pone.0017682-Cobrinik2], [@pone.0017682-Classon1], [@pone.0017682-Lee3], [@pone.0017682-LeCouter1], [@pone.0017682-LeCouter2], [@pone.0017682-Classon2], [@pone.0017682-Scime2], [@pone.0017682-Xu1]. During myogenesis, over-expression of p107 suppressed ectopic cell proliferation in Rb^−/−^ myotubes [@pone.0017682-Schneider1], whereas over-expression of p130, but not pRb, inhibited myogenic differentiation in Rb proficient mouse C2 myoblasts [@pone.0017682-Carnac1]. These findings raised the possibility that p107 might act in a manner redundant to Rb, whereas p130 might promote defects resulting from Rb deficiency during myogenic differentiation. To address the function of the Rb protein family during terminal differentiation, we analyzed double Rb^−/−^:p107^−/−^ and Rb^−/−^:p130^−/−^ as well as triple Rb^−/−^:p107^−/−^:p130^−/−^ mutant myoblasts following chronic or acute inactivation of Rb. We report that loss of either p107 or p130 enhanced the survival defect of Rb-deficient myoblasts, implying that both of these Rb-related proteins partially compensate for Rb loss. Nevertheless, myoblasts lacking Rb plus either p107 or p130 could differentiate and twitch, as long as autophagic cell death is suppressed or metabolism is shifted to glycolysis under hypoxia. In contrast, myoblasts lacking all three Rb family members do not efficiently fuse or survive, indicating that expression of at least one Rb family protein is essential for these processes. Results {#s2} ======= Combined mutation in p130 does not counteract Rb myogenic defects in mgRb:Rb^−/−^:p130^−/−^ double mutant fetuses {#s2a} ----------------------------------------------------------------------------------------------------------------- As noted, over-expression of p107 suppresses ectopic cell proliferation in Rb^−/−^ myotubes [@pone.0017682-Schneider1], whereas over-expression of p130 inhibits myogenic differentiation in Rb proficient mouse C2 myoblasts [@pone.0017682-Carnac1], suggesting that combined loss of Rb plus p107 or Rb plus p130 might worsen or ameliorate the Rb myogenic defect, respectively. To test the effects of p107 and p130 on differentiation of Rb-deficient myoblasts, we generated and intercrossed mgRb:Rb^+/−^:p107^+/−^ and mgRb:Rb^+/−^:p130^+/−^ double heterozygote mice. Consistent with previous studies on early death of Rb/p107 double mutant embryos [@pone.0017682-Lee3], [@pone.0017682-Berman1], we were unable recover live E14.5--E16.5 mgRb:Rb^−/−^:p107^−/−^ double mutant embryos after breeding mgRb:Rb^+/−^:p107^+/−^ mice (not shown). In contrast, E14.5--E16.5 mgRb:Rb^−/−^:p130^−/−^ fetuses were identified following mgRb:Rb^+/−^:p130^+/−^ interbreeding at the expected Mendelian frequency (6/53 = 11.32%; expected 12.5%; [**Table 1**](#pone-0017682-t001){ref-type="table"}). At E17.5, the frequency of mgRb:Rb^−/−^:p130^−/−^ double knockout (KO) embryos dropped to 1/59 (1.69%), whereas viable Rb^−/−^ single knockouts were present at 9/59 (24%), approximating the expected 25% frequency. ::: {#pone-0017682-t001 .table-wrap} 10.1371/journal.pone.0017682.t001 Table 1 ::: {.caption} ###### Frequency of *mgRb:Rb^−/−^*:*p130^−/−^* embryos recovered. ::: {#pone-0017682-t001-1} Gestation *mgRb:Rb^−/−^* [\*](#nt101){ref-type="table-fn"} *mgRb:Rb^−/−^:p130^−/−^* [\*](#nt101){ref-type="table-fn"} *mgRb:Rb^−/−^:p130^−/−^* [\*\*](#nt102){ref-type="table-fn"} ----------- -------------------------------------------------- ------------------------------------------------------------ -------------------------------------------------------------- E14.5 4/54 (7.41%) 6/54 (11.11%) Not determined E15.5 8/41 (19.51%) 6/41 (14.63%) 6/38 (15.79%) E16.5 11/53 (20.75%) 6/53 (11.32%) 2/20 (10.0%) E17.5 9/59 (24.09%) 1/59 (1.69%) 4/45 (8.89%) \*Mutant embryos obtained from *mgRb:Rb^+/−^:p130^+/−^* × *mgRb:Rb^+/−^:p130^−/−^* intercrosses. Total of 5 *in utero* deaths at E17.5 were not included. A frequency of 12.5% is expected in each indicated group. \*\*Mutant embryos obtained from *mgRb:Rb^+/−^:p130^−/−^* × *mgRb:Rb^+/−^:p130^−/−^* intercrosses. Total of 4 *in utero* deaths at E17.5 were not included. A frequency of 25% is expected in *mgRb:Rb^−/−^:p130^−/−^* group. ::: Compared to E16.5 mgRb:Rb^−/−^ single KO embryos, mgRb:Rb^−/−^:p130^−/−^ double knockout (DKO) fetuses displayed a more pronounced hunchback, suggesting reduced muscle toning ([**Fig. 1A**](#pone-0017682-g001){ref-type="fig"}, top panels). However, histological sections through epaxial and hypaxial skeletal muscles of mgRb:Rb^−/−^:p130^−/−^ fetuses revealed similar defects as in mgRb:Rb^−/−^ littermates, including reduced density and shortened myofibers, and enlarged nuclei within myotubes compared to control ([**Fig. 1A**](#pone-0017682-g001){ref-type="fig"}, arrowheads). Expression of myosin heavy chain (MHC), an early marker of differentiation, was similar in wild type and mutant embryos whereas expression of troponin T, a late marker of differentiation, was similarly reduced in single and double mutants relative to control ([**Fig. 1B**](#pone-0017682-g001){ref-type="fig"}). Terminal deoxynucleotidyl transferase biotin-dUTP nick end labeling (TUNEL, [@pone.0017682-Gavrieli1]) analysis revealed no obvious differences in apoptosis: 22.2±9.3% and 20.1±4.7% for single and DKO, respectively (not shown). Thus, despite the enhanced hunchback mgRb:Rb^−/−^:p130^−/−^ DKO embryos, we did not detect obvious changes in myoblast differentiation *in vivo*. ::: {#pone-0017682-g001 .fig} 10.1371/journal.pone.0017682.g001 Figure 1 ::: {.caption} ###### Analysis of myogenesis mgRb:Rb^−/−^:p130^−/−^ double mutant embryos. \(A) Appearance of ctrl, mgRb:Rb^−/−^ and mgRb:Rb^−/−^:p130^−/−^ embryos at E16.5 (a--c). Hematoxylin & eosin (H&E) histology staining reveals the cellular morphology at mid-sagittal section (d--f), epaxial muscles (g--i) and hypaxial muscles (j--l). Arrowheads point to enlarged nuclei within myofibers. (B) Confocal images showing fast troponin T (green) expression in skeletal muscle sections of control (ctrl) (a--b) mgRb:Rb^−/−^ (c--d) and mgRb:Rb^−/−^:p130^−/−^ (e--f) embryos. MHC (green) expression in skeletal muscle sections of ctrl (g) mgRb:Rb^−/−^ (h) and mgRb:Rb^−/−^:p130^−/−^ (i) embryos. DAPI was used to counter-stain nuclei (blue). (C) Top, western blot analysis for pRb and p107 in DM-2 ctrl, mgRb:Rb^−/−^ and mgRb:Rb^−/−^:p130^−/−^ myoblast cultures. Tubulin was used as loading control. Bottom, western blot analysis of p130 in DM-2 ctrl and mgRb:Rb^−/−^:p130^−/−^ cultures. Arrow indicates location of p130. Lower band represents a splice variant or cross-reactive protein. (D) Average number of myotubes counted on indicated days post-differentiation. Each time point represents an average ± s.d. of 6 fields at 200X (n = 4). :::  ::: Combined mutations in Rb and p130 accelerated myotube degeneration *in vitro* {#s2b} ----------------------------------------------------------------------------- To test for a cell-autonomous effect of combined mutations in Rb and p130, we derived primary myoblasts from E16.5 mgRb:Rb^−/−^:p130^−/−^ limb muscles. Absence of pRb and p130 was verified in mgRb:Rb^−/−^:p130^−/−^ cultures 2 days post-differentiation (DM-2) ([**Fig. 1C**](#pone-0017682-g001){ref-type="fig"}). We note that the p130^−/−^ lane lacked full-length p130, but contained a lower band, which could represent a cross reactive protein or a truncated/spliced p130 variant. Expression of p107, which was undetected in DM-2 control myotube cultures, was elevated in Rb KO myotubes and further increased in Rb/p130 DKO cultures ([**Fig. 1C**](#pone-0017682-g001){ref-type="fig"}). This observation is consistent with the presence of E2F binding sites in the p107 promoter and its regulation by pRb [@pone.0017682-Zhu1], [@pone.0017682-Burkhart2], and may partly compensate for the combined loss of pRb and p130 in these DKO cells. Under differentiation conditions, wild-type myoblasts fused to form long multinucleated myotubes that twitched for weeks in culture ([**Fig. 1D**](#pone-0017682-g001){ref-type="fig"}). In contrast, both mgRb:Rb^−/−^ and mgRb:Rb^−/−^:p130^−/−^ myoblasts initially fused to form short myotubes containing 3--6 nuclei, but underwent rapid degeneration beginning 2--3 days post-differentiation; by day 5--6 virtually all myotubes had degenerated. Reproducibly, the mgRb:Rb^−/−^:p130^−/−^ DKO myoblasts exhibited reduced myotube formation relative to mgRb:Rb^−/−^ cultures, but similar degeneration kinetics ([**Fig. 1D**](#pone-0017682-g001){ref-type="fig"}). Unlike wild-type myotubes, which are stably post-mitotic ([**Fig. 2A**](#pone-0017682-g002){ref-type="fig"}, top), nuclei in Rb-deficient myotubes incorporate BrdU when stimulated with mitogens, indicating a failure in establishing a permanent cell cycle exit ([**Fig. 2A**](#pone-0017682-g002){ref-type="fig"}, middle) [@pone.0017682-Schneider1], [@pone.0017682-Chen3], [@pone.0017682-Ciavarra1]. To test the cell cycle status of DKO myotubes, cultures were differentiated for 1 day and then re-stimulated with growth medium in the presence of BrdU. Like Rb^−/−^, the mgRb:Rb^−/−^:p130^−/−^ myotubes incorporated BrdU ([**Fig. 2A**](#pone-0017682-g002){ref-type="fig"}, bottom). We next tested whether mgRb:Rb^−/−^:p130^−/−^ myotube degeneration was associated with apoptosis. TUNEL-positive nuclei were detected in unfused myoblasts, but not within myotubes, and were more abundant in Rb and Rb/p130 DKO cultures than in control ([**Fig. 2B**](#pone-0017682-g002){ref-type="fig"}). Importantly, the level of apoptosis was slightly, but reproducibly, elevated in Rb/p130 DKO relative to Rb KO cultures ([**Fig. 2B**](#pone-0017682-g002){ref-type="fig"}, see below). ::: {#pone-0017682-g002 .fig} 10.1371/journal.pone.0017682.g002 Figure 2 ::: {.caption} ###### Differentiation of mgRb:Rb^−/−^:p130^−/−^ DKO myoblasts. \(A) Confocal microscopy analysis for BrdU incorporation in ctrl, mgRb:Rb^−/−^ and mgRb:Rb^−/−^:p130^−/−^ myotubes at DM-2. Myoblasts were differentiated for 1 day, then exposed to 20 µM BrdU for an additional 16 hr in the presence of growth medium (GM) and immuno-stained for MHC (red) and BrdU (green). Arrowheads label BrdU positive nuclei within myotubes. (B) MHC (red) and TUNEL (green) staining at DM-2. Arrowheads indicate TUNEL positive nuclei, which are invariably located outside myotubes. (C) Mitotracker® (red) staining at DM-2. Arrowheads point to large Mitotracker®-positive perinuclear aggregates. :::  ::: To determine whether combined mutations in Rb and p130 led to further disruption of the mitochondrial network seen in Rb deficient myotubes [@pone.0017682-Ciavarra1], we used MitoTracker®, a live-cell probe that accumulates in mitochondria with intact membrane potential. In DM-2 control myotubes, MitoTracker® staining revealed a uniform, net-like organization of mitochondria throughout the cytoplasm ([**Fig. 2C**](#pone-0017682-g002){ref-type="fig"}, top). However, both mgRb:Rb^−/−^ and mgRb:Rb^−/−^:p130^−/−^ myotubes exhibited relatively sparse cytosolic staining and strong perinuclear MitoTracker®-positive clusters in ∼70% of myotubes, indicative of autophagy ([**Fig. 2C**](#pone-0017682-g002){ref-type="fig"}). Importantly, there was no detectable difference in the level of perinuclear aggregation in single and DKO myoblasts. Together, these results indicate that loss of p130 does not ameliorate the Rb myogenic defect, but rather, exacerbates it by reducing myoblast survival prior to fusion. Efficient rescue of Rb/p130 DKO myoblast degeneration by Bcl-2, autophagy inhibitor and hypoxia {#s2c} ----------------------------------------------------------------------------------------------- We next asked whether inhibition of autophagy would rescue muscle degeneration in mgRb:Rb^−/−^:p130^−/−^ DKO cultures as it does for Rb-deficient myotubes [@pone.0017682-Ciavarra1]. Remarkably, adenovirus mediated transduction of Bcl-2, which inhibits both apoptosis and autophagy [@pone.0017682-Pattingre1], effectively rescued myogenic degeneration of mgRb:Rb^−/−^:p130^−/−^ DKO myotubes, leading to long twitching myotubes ([**Fig. 3A**](#pone-0017682-g003){ref-type="fig"}). This striking result demonstrates that when provided with a survival signal that counteracts the pro-apoptotic effect of Rb loss, myotube formation and maintenance does not require active participation of pRb and p130. We previously demonstrated that inhibition of autophagy by 3-methyladenine (3-MA), an antagonist of a class III phosphatidylinositol 3-kinase, Vps34, which is necessary for autophagic vesicle nucleation [@pone.0017682-Levine1], also rescued the degeneration of Rb-deficient myotubes [@pone.0017682-Ciavarra1]. A single dose of 3-MA administered just prior to induction of differentiation prevented degeneration of mgRb:Rb^−/−^:p130^−/−^ DKO myotubes ([**Fig. 3B**](#pone-0017682-g003){ref-type="fig"}). The 3-MA-rescued myotubes twitched for weeks in culture and were indistinguishable from control myotubes (**[Videos S1](#pone.0017682.s002){ref-type="supplementary-material"}** and **[S2](#pone.0017682.s003){ref-type="supplementary-material"}**). The pan-PPAR agonist bezafibrate [@pone.0017682-Wenz1], [@pone.0017682-Bastin1], which induces mitochondrial biogenesis, also rescued the collapse of mgRb:Rb^−/−^:p130^−/−^ DKO myotubes with similar efficiency as 3-MA ([**Fig. 3B**](#pone-0017682-g003){ref-type="fig"}). As control, ectopic expression of a constitutively active, phospho-mutant pRb (Ad.Rb^ΔK11^) [@pone.0017682-Jiang4] prevented degeneration of mgRb:Rb^−/−^:p130^−/−^ DKO myotubes ([**Fig. 3A**](#pone-0017682-g003){ref-type="fig"}, bottom panel). ::: {#pone-0017682-g003 .fig} 10.1371/journal.pone.0017682.g003 Figure 3 ::: {.caption} ###### Rescue of mgRb:Rb^−/−^:p130^−/−^ myogenic defect by autophagy inhibitors and hypoxia. \(A) Brightfield images of mgRb:Rb^−/−^:p130^−/−^ myoblasts transduced with Ad.GFP, Ad.Bcl-2 or Ad.RbΔK11 and then induced to differentiate for 14 days. Arrowheads point to myotubes. (B) Average number of mgRb:Rb^−/−^:p130^−/−^ myotubes following treatment with 3-MA, bezafibrate or DM as indicated. Counts are average ± s.d. of 6 fields at 200X (n = 3). (C) Immunostaining for MHC (green) in ctrl and mgRb:Rb^−/−^:p130^−/−^ cultures differentiated under normoxia or hypoxia. Note myotubes in mgRb:Rb^−/−^:p130^−/−^ cultures at DM-5 under hypoxia but not normoxia. Nuclei were counterstained with DAPI. :::  ::: In addition to autophagy antagonists, hypoxia (∼1% O~2~) rescued the myogenic defect induced by loss of Rb by redirecting metabolism from oxidative phosphorylation to glycolysis [@pone.0017682-Ciavarra1]. To test whether hypoxia could also rescue the myogenic defect in Rb/p130 DKO myoblasts, mgRb:Rb^−/−^:p130^−/−^ and control cultures were induced to differentiate under hypoxia. Remarkably, despite the enhanced cell death and reduced myotube formation in mgRb:Rb^−/−^:p130^−/−^ cultures relative to mgRb:Rb^−/−^, hypoxia efficiently maintained the survival of DKO myotubes, leading to long, twitching myotubes ([**Fig. 3C**](#pone-0017682-g003){ref-type="fig"}). These results indicate that combined loss of Rb plus p130 reduces myoblast survival and myotube formation relative to Rb loss alone. Yet, when cell death is inhibited, DKO myotubes can differentiate as efficiently as Rb mutant or control cultures, demonstrating that neither factor is required for stimulating or maintaining the differentiation program. Inactivation of Rb but not p107 plus p130 leads to ectopic DNA synthesis and degeneration of primary myotubes {#s2d} ------------------------------------------------------------------------------------------------------------- To characterize the functions of p107 and p130 during skeletal myogenesis in combination with Rb deletion, we isolated primary myoblasts from E16.5 Rb^f/f^:p107^−/−^:p130^−/−^ composite mutant embryos. In these mice, Rb exon 19 is flanked by loxP sites, allowing acute inactivation through Cre mediated excision [@pone.0017682-Vooijs1]. Transduction of Rb^f/f^ myoblasts with Ad.Cre at a high multiplicity of infection (1300) led to loss of pRb expression in all cells as determined by immunostaining and immunoblotting ([**Fig. 4A--B**](#pone-0017682-g004){ref-type="fig"}), and to myotube degeneration ([**Fig. 4C**](#pone-0017682-g004){ref-type="fig"}). Deletion of p107 and p130 was detected by PCR (not shown) and confirmed by immunoblotting ([**Fig. 4D**](#pone-0017682-g004){ref-type="fig"}). ::: {#pone-0017682-g004 .fig} 10.1371/journal.pone.0017682.g004 Figure 4 ::: {.caption} ###### BrdU incorporation analysis of Rb^Δf^ versus p107^−/−^:p130^−/−^ myotubes. \(A) Rb^f/f^ myoblasts were transduced with Ad.GFP or Ad.cre and 48 hr later were immunostained for pRb (red). Nuclei were counterstained with DAPI. (B) Rb^f/f^ myoblasts were transduced with Ad.EV or Ad.cre and immunoblotted for pRb 48 hr later. Tubulin served as loading control. (C) Rb^f/f^ myoblasts were transduced with Ad.cre, induced to differentiate for 2 or 6 days and immunostained for MHC (red). (D) Western blot analysis of pRb, p107 and p130 in skeletal muscle of E16.5 Rb^f/f^:p107^+/−^:p130^+/−^ (ctrl) and Rb^f/f^:p107^−/−^:p130^−/−^ fetuses. (E) Immunostaining for BrdU and MHC in Ad.EV and Ad.cre transduced Rb^f/f^ myoblasts at DM-2. Myoblasts were differentiated for 1 day, then exposed to 20 µM BrdU for an additional 16 hr in the presence of GM and stained for MHC (red) and BrdU (green). Arrowheads label BrdU positive nuclei within myotubes. (F) Immunostaining for BrdU and MHC in Rb^f/f^, Rb^f/f^:p107^−/−^, Rb^f/f^:p130^−/−^ and Rb^f/f^:p107^−/−^:p130^−/−^ myoblasts at DM-2. Myoblasts were differentiated for 1 day, then exposed to 20 µM BrdU for an additional 16 hr in the presence of GM and stained for MHC (red) and BrdU (green). Note absence of BrdU-positive nuclei in myotubes. (G) Quantification of BrdU incorporation in Ad.EV or Ad.cre transduced Rb^f/f^ and Rb^f/f^:p107^−/−^:p130^−/−^ myotubes at DM-2. :::  ::: We first tested whether combined mutations in p107 and p130 resulted in deregulated cell proliferation in response to a differentiation signal. Interestingly, RNAi inhibition of Rb family members in C2C12 myotubes suggests that knock-down of p107 plus p130 does not cause cell cycle re-entry [@pone.0017682-Blais1]. This could be due to incomplete knockdown of p107 and/or p130 by RNAi. In addition, C2C12 cells lack Arf [@pone.0017682-Pajcini1], and therefore the observed effect could be confounded by deregulation of the ARF/MDM2/p53 pathway. To address the consequences of complete KO of p107 plus p130 in cells with an intact ARF-p53 pathway, we induced primary Rb^f/f^:p107^−/−^:p130^−/−^ and control myoblasts to differentiate and then re-stimulated with growth medium in the presence of BrdU. Rb^Δf^ mutant myotubes incorporated BrdU ([**Fig. 4E, G**](#pone-0017682-g004){ref-type="fig"}). In contrast, Rb^f/f^:p107^−/−^:p130^−/−^ double KO myotubes (which retain the two Rb^f/f^ alleles) did not ([**Fig. 4F, G**](#pone-0017682-g004){ref-type="fig"}). Thus, Rb loss prevents cell cycle exit during terminal differentiation and this function is unique to Rb. Notably, although Rb^f/f^:p107^−/−^:p130^−/−^ double mutant myotubes survived and twitched like control Rb-proficient cultures, we consistently observed reduced myotube formation in p107/p130 DKO cultures relative to control (e.g. [**Fig. 5B**](#pone-0017682-g005){ref-type="fig"}). ::: {#pone-0017682-g005 .fig} 10.1371/journal.pone.0017682.g005 Figure 5 ::: {.caption} ###### Differentiation potential of double and triple KO myoblasts. \(A) Immunostaining for MHC (green) in Ad.EV and Ad.cre transduced Rb^f/f^, Rb^f/f^:p107^−/−^, Rb^f/f^:p130^−/−^ and Rb^f/f^:p107^−/−^:p130^−/−^ myoblast cultures at DM-2. Nuclei were counterstained with DAPI. (B) Quantification of percent multinucleated myotubes relative to total number of MHC-positive cells (myocytes plus myotubes) in Ad.EV and Ad.cre transduced Rb^f/f^, Rb^f/f^:p107^−/−^, Rb^f/f^:p130^−/−^ and Rb^f/f^:p107^−/−^:p130^−/−^ myoblasts at DM-2 under normoxia. Numbers within bars indicate % for the respective samples. :::  ::: Acute inactivation of Rb plus p107 or p130 leads to reduced myotube formation whereas TKO myoblasts form short bi-nuclear myocytes {#s2e} ---------------------------------------------------------------------------------------------------------------------------------- To investigate the effect of Rb plus p107 and/or p130 on myogenesis, proliferating Rb^f/f^, Rb^f/f^:p130^−/−^, Rb^f/f^:p107^−/−^ and Rb^f/f^:p130^−/−^:p107^−/−^ myoblasts were transduced with Ad.cre or control empty vector (Ad.EV) and induced to differentiate 48 hr later (permitting pre-existing pRb protein degradation). Both Rb^Δf^:p130^−/−^ and Rb^Δf^:p107^−/−^ myoblasts differentiated to form short myotubes by DM-2 with slightly less myotubes in the Rb/p107 than the Rb/p130 DKO cultures ([**Fig. 5A**](#pone-0017682-g005){ref-type="fig"}). In addition to multinucleated myotubes, Rb^Δf^:p107^−/−^, and to a lesser extent Rb^Δf^:p130^−/−^ and Rb^Δf^ myoblasts formed elongated MHC-positive myocytes containing a single nucleus. Quantification of multinucleated myotubes relative to total MHC-positive cells (i.e. myocytes plus myotubes) revealed that Rb^Δf^, Rb^Δf^:p107^−/−^ and Rb^Δf^:p130^−/−^ cultures contained 51%, 31% and 43% myotubes relative to total MHC-positive cells, respectively ([**Fig. 5B**](#pone-0017682-g005){ref-type="fig"}). Rb^Δf^:p130^−/−^:p107^−/−^ TKO myoblasts formed primarily elongated myocytes that typically contained one or two nuclei ([**Fig. 5A**](#pone-0017682-g005){ref-type="fig"}, h; [**Fig. 6A**](#pone-0017682-g006){ref-type="fig"}, bottom right); some rare myotubes containing three nuclei were also observed (**[Fig. S1](#pone.0017682.s001){ref-type="supplementary-material"}**). The percentage of short bi-nuclear myotubes relative to total MHC-positive cells in the TKO cultures (single-nucleus myocytes plus binuclear myocytes or myotubes) was ∼9% ([**Fig. 5B**](#pone-0017682-g005){ref-type="fig"}). ::: {#pone-0017682-g006 .fig} 10.1371/journal.pone.0017682.g006 Figure 6 ::: {.caption} ###### Increased apoptosis associated with differentiation of double and triple KO myoblasts. \(A) Mitotracker® staining of Ad.EV and Ad.cre transduced Rb^f/f^, Rb^f/f^:p107^−/−^, Rb^f/f^:p130^−/−^ and Rb^f/f^:p107^−/−^:p130^−/−^ myoblasts at DM-2. Arrowheads point to large perinuclear aggregates in Ad.cre transduced myotubes. (B) TUNEL staining (green) of Ad.EV and Ad.cre transduced Rb^f/f^ (a--b), Rb^f/f^:p107^−/−^ (c--d), Rb^f/f^:p130^−/−^ (e--f) and Rb^f/f^:p107^−/−^:p130^−/−^ (g--h) cultures at DM-2. Nuclei were counterstained with DAPI. Note that TUNEL positive nuclei are outside myotubes. (C) Percent increase in TUNEL-positive cells in Ad.cre relative to Ad.EV transduced cultures. Error bars represent s.d. \*-p\<0.05 and \*\*-p\<0.07 t-test comparisons relative to Rb^f/f^. :::  ::: Reduced myotube formation in DKO and TKO cultures correlates with increased apoptosis in differentiating myoblasts {#s2f} ------------------------------------------------------------------------------------------------------------------ The reduced number of myotubes in Rb/p107, Rb/p130 and Rb/p107/p130 mutant cultures relative to Rb KO and control cultures could be due to intrinsic defects in differentiation or to excessive myoblast cell death, which would diminish the available pool of competent neighboring myocytes for fusion. To distinguish between these possibilities, we first assessed the level of mitochondrial perinuclear aggregation in differentiating myotubes. Following acute inactivation of Rb, transiently formed Rb^Δf^ myotubes, like mgRb:Rb^−/−^ myotubes, exhibited abnormal perinuclear aggregation of mitochondria ([**Fig. 6A**](#pone-0017682-g006){ref-type="fig"}, bottom left panel). The level of perinuclear mitochondrial aggregation in Rb/p107 and Rb/p130 DKO myotubes was similar to that observed in Rb KO cultures. In contrast, the short bi-nuclear TKO myocytes did not exhibit mitochondrial aggregation in the perinuclear region ([**Fig. 6A**](#pone-0017682-g006){ref-type="fig"}, bottom right panel). TUNEL analysis revealed that differentiating TKO myoblasts underwent the highest level of apoptosis (119%) relative to the level of apoptosis in wild-type, followed by Rb/p107 (49%), Rb/p130 (24%) and Rb (19%) KO myoblasts ([**Fig. 6B--C**](#pone-0017682-g006){ref-type="fig"}). Thus, the reduced myotube formation (and increase in elongated myocytes) was directly proportional to the level of apoptosis in the various mutant cultures. Interestingly, it was reported that myoblasts seeded at low-density do not fuse under differentiation conditions but instead form elongated myocytes that undergo differentiation in the absence of fusion [@pone.0017682-Pajcini1]. Thus, reduced myotube formation in the DKO and TKO cultures likely reflect, at least in part, the increased apoptosis and reduced number of competent myocytes available for fusion. Hypoxia efficiently rescues myotubes formed in the absence of Rb and p130 or p107, but not in the absence of all three Rb family proteins {#s2g} ----------------------------------------------------------------------------------------------------------------------------------------- Hypoxia most effectively rescues the myogenic defect following acute inactivation of Rb [@pone.0017682-Ciavarra1]. To test whether hypoxia could prevent myotube degeneration following acute inactivation of Rb plus its relatives, Rb^f/f^, Rb^f/f^:p130^−/−^, Rb^f/f^:p107^−/−^ and Rb^f/f^:p130^−/−^:p107^−/−^ myoblasts were transduced with Ad.cre or Ad.EV and then either maintained under normoxia or shifted to hypoxic conditions. Under normoxia, no myotubes survived by DM-5 ([**Fig. 5A**](#pone-0017682-g005){ref-type="fig"}). However, under hypoxia, Rb^Δf^:p130^−/−^ and Rb^Δf^:p107^−/−^ myotubes survived and twitched, and appeared indistinguishable from Rb^Δf^ or control myotubes ([**Fig. 7A--B**](#pone-0017682-g007){ref-type="fig"}; **[videos S3](#pone.0017682.s004){ref-type="supplementary-material"}--[5](#pone.0017682.s006){ref-type="supplementary-material"}**). In striking contrast, elongated TKO myocytes/myotubes degenerated, forming ultra-thin myocytes/myotubes, very few of which nonetheless twitched (**[video S6](#pone.0017682.s007){ref-type="supplementary-material"}**). The ratio of myotubes to myocytes at DM-5 was similar, relative to DM-2 (compare [**Fig. 5B** to **F**](#pone-0017682-g005){ref-type="fig"} **ig. 7C**). To directly test this, we induced the various cultures to differentiate under hypoxia and then counted the number of myotubes at DM-2 and DM-6. As shown in [**Figure 7D**](#pone-0017682-g007){ref-type="fig"}, the ratio of myotubes at DM-2 and DM-6 was similar in the two DKO cultures indicating that once formed, myotubes can survive in hypoxia independently of pRb-p107 or pRb-p130 protein family. In contrast, very few TKO binuclear myocytes/myotubes survived at DM-6 under hypoxia, and they were ultra-thin and clearly abnormal ([**Fig. 8A**](#pone-0017682-g008){ref-type="fig"}). Thus, at least one pRb protein family is required for robust differentiation even under hypoxic conditions, which rescue Rb-deficient myotube degeneration. ::: {#pone-0017682-g007 .fig} 10.1371/journal.pone.0017682.g007 Figure 7 ::: {.caption} ###### Differentiation of double and triple KO myoblasts under hypoxia. \(A) Immunostaining for MHC (green) of Ad.EV (a,c,e,g) or Ad.cre (b,d,f,h) transduced Rb^f/f^, Rb^f/f^:p107^−/−^, Rb^f/f^:p130^−/−^ and Rb^f/f^:p107^−/−^:p130^−/−^ cultures at DM-5 in hypoxia. Inlets, DAPI staining for nuclei. (B) Quantification of myotube formation in Rb^f/f^, Rb^f/f^:p107^−/−^, Rb^f/f^:p130^−/−^ and Rb^f/f^:p107^−/−^:p130^−/−^ cultures transduced with Ad.EV or Ad.cre and induced to differentiate 48 hr later for 5 days. Counts represent the average number of myotubes at DM-5 of 6 representative fields (n = 4); error bars represent s.d. \*-p\<0.05 and \*\*-p\<0.07. (C) Quantification of percent multinucleated myotubes relative to total number of MHC-positive cells in Ad.EV or Ad.cre transduced Rb^f/f^, Rb^f/f^:p107^−/−^, Rb^f/f^:p130^−/−^ and Rb^f/f^:p107^−/−^:p130^−/−^ cultures at DM-5 under hypoxia. Numbers within bars indicate % for respective samples. (D) Quantification of myotube formation in Rb^f/f^, Rb^f/f^:p107^−/−^, Rb^f/f^:p130^−/−^ and Rb^f/f^:p107^−/−^:p130^−/−^ cultures transduced with Ad.EV or Ad.cre and induced to differentiate in hypoxia. Counts were conducted at DM-2 and DM-6. Error bars represent s.d. :::  ::: ::: {#pone-0017682-g008 .fig} 10.1371/journal.pone.0017682.g008 Figure 8 ::: {.caption} ###### Evidence that bi-nuclear TKO myotubes originate from nuclear duplication, not cell fusion. \(A) Bright-field images of Ad.EV and Ad.cre transduced Rb^f/f^:p107^−/−^:p130^−/−^ cultures at DM-6 in hypoxia. Arrowheads point to myotubes. Note the presence of a thin myotube in Ad.cre transduced culture. (B) Top row, low magnification image of Ad.EV or Ad.cre transduced Rb^f/f^:p107^−/−^:p130^−/−^ cultures immunostained for pRb (green) at DM-2 (400x). Bottom row, high magnification (630x) of Ad.EV or Ad.cre transduced Rb^f/f^:p107^−/−^:p130^−/−^ cultures induced to differentiate and then immunostained for pRb (green), demonstrating absence of detectable pRb within binuclear myocyte. Nuclei were counterstained with DAPI. Arrowheads point to nuclei in the myocyte. (C) Immunostaining for MHC (red) and BrdU (green) at DM-2 of Ad.EV and Ad.cre transduced Rb^f/f^:p107^−/−^:p130^−/−^ cultures, which were induced differentiate after equal mixing of BrdU^+^ labeled and BrdU^−^ myoblast populations. Note mixed BrdU^+^ and BrdU^−^ nuclei in control myotube (top panel) but only BrdU^−^:BrdU^−^ or BrdU^+^:BrdU^+^ nuclei in TKO myotubes (middle and bottom, respectively). :::  ::: Short TKO binuclear myocytes originate primarily from acytokinetic mitosis and not fusion {#s2h} ----------------------------------------------------------------------------------------- If the bi-nuclear TKO myocytes originate from cell fusion, this would suggest that the entire Rb family is dispensable for this process. To address this possibility, we first asked whether the bi-nuclear TKO myocytes represented rare Rb^f/f^:p130^−/−^:p107^−/−^ myoblasts in which the Rb^f/f^ allele had not been deleted. However, upon staining, none of over 80 short binuclear myocytes expressed pRb (n = 2) ([**Fig. 8B**](#pone-0017682-g008){ref-type="fig"}). Next, we asked whether the elongated binuclear myocytes originated from cell fusion of TKO myocytes or from acytokinetic mitosis (i.e. absence of cytokinesis at the end of mitosis), which occurs normally in cardiac myocytes and other tissues [@pone.0017682-Ullah1]. To distinguish between these possibilities, proliferating TKO or control myoblasts (two independent cultures for each) were labeled with BrdU for 24 hr. The labeled cells were then mixed with an equal number of unlabeled TKO or control myoblasts, respectively, and induced to differentiate. Control cultures contained three populations of myotubes (BrdU^+^/BrdU^+^; BrdU^+^/BrdU^−^; BrdU^−^/BrdU^−^), with an average of 35% (of \>200 myotubes) containing a mixture of BrdU^+^ and BrdU^−^ nuclei, indicative of cell fusion (n = 4) ([**Fig. 8C**](#pone-0017682-g008){ref-type="fig"}). In striking contrast, with the exception of one elongated bi-nuclear TKO myocyte with BrdU^+^/BrdU^−^ nuclei, more than 150 other bi-nuclear TKO myocytes contained either BrdU^−^/BrdU^−^ or BrdU^+^/BrdU^+^ nuclei, strongly suggesting that binuclear TKO myocytes originate primarily from acytokinetic mitosis, not from cell fusion. This is consistent with our observation that bi-nuclear TKO myocytes do not exhibit perinuclear aggregation of mitochondria as observed in Rb KO and DKO myotubes ([**Fig. 6A**](#pone-0017682-g006){ref-type="fig"}). Thus, these results suggest that efficient myocyte fusion requires the presence of at least one member of the Rb protein family. Discussion {#s3} ========== We report that mutant myoblasts lacking Rb and one of its relatives, p107 or p130, can undergo robust myogenic differentiation under conditions we previously established whereby the survival defect in Rb-deficient myotubes is rescued by autophagy antagonists or hypoxia. In contrast, under the same conditions, combined mutations in all Rb protein family members, pRb, p107 and p130, severely abrogate myogenic differentiation. Thus, myoblast fusion and myotube survival require at least one Rb family member. We discuss our findings in the context of recent published work by another group demonstrating that various tissue-types can differentiate in the absence of Rb family and propose a tissue specific active/default model for Rb during cell fate determination and differentiation. It is commonly thought that the tumor suppressor pRb plays at least two independent functions during differentiation: control of cell division and apoptosis through inhibition of E2F responding genes, and stimulation of the differentiation program through activation of differentiation factors. However, we recently showed that survival factors, autophagy inhibitors and hypoxia can prevent the degeneration of Rb-deficient muscles, leading to contracting myotubes [@pone.0017682-Ciavarra1]. While these observations challenge the notion that pRb is actively required to stimulate the differentiation program, the possibility that p107 and/or p130 partially compensate for pRb during differentiation was not ruled out. Specifically, p107 or p130 could account for the ability of Rb-deficient myoblasts to fuse to form short myotubes prior to degeneration. In addition, it was unresolved whether p130 compensates or counteracts pRb during differentiation [@pone.0017682-Carnac1]. Here, we addressed these issues by analyzing primary myoblasts with composite mutations in the Rb gene family. We showed that combined mutations in pRb plus p107 or pRb plus p130 increased apoptosis in myoblasts, and accordingly, reduced the number of DKO myotubes. p107 was clearly more critical than p130 in preventing apoptosis of Rb-deficient myoblasts during differentiation. Nevertheless, the myotubes that formed in the absence of Rb and either p107 or p130 survived and twitched under hypoxia or following treatment with autophagy-antagonists. Several studies suggest that the Rb family may affect differentiation and lineage commitment by transcriptionally regulating the expression of differentiation factors such as PPARγ and PGC-1α, by sequestering inhibitors of differentiation like ID2, HDAC, EID1 and RBP2 [@pone.0017682-Calo1], [@pone.0017682-Scime1], [@pone.0017682-Lasorella1], [@pone.0017682-Puri1], [@pone.0017682-MacLellan1], [@pone.0017682-Benevolenskaya1], or by binding and stimulating differentiation factors such as CBFA1 during osteoblast differentiation, C/EBPβ during adipocyte differentiation and MyoD and myogenin during myogenesis [@pone.0017682-Thomas1], [@pone.0017682-Chen2], [@pone.0017682-Skapek1], [@pone.0017682-Gu1], [@pone.0017682-Schneider1], [@pone.0017682-Novitch1]. A recent genome-wide mammalian protein-protein interaction analysis independently demonstrated that human RB1 and p130 but not p107 interact with MYOD1, RUNX2 and C/EBP [@pone.0017682-Ravasi1]. p107 does not seem to interact with MyoD or any other differentiation factor, which is consistent with its reduced expression during differentiation. Thus, the ability of Rb/p130 DKO myoblasts to fully differentiate when treated with autophagy inhibitors or hypoxia is remarkable and questions the notion that the Rb family is actively required for differentiation. Possibly, interaction of pRb and p130 with MyoD may be required for cell survival or cell cycle exit, but it does not seem essential for differentiation *per se*. While we observed a quantitative reduction in myotube formation between single KO and double KO myoblasts, there was a qualitative difference between DKO and TKO cultures. TKO myoblast cultures exhibited excessive myoblast death and primarily formed elongated, bi-nuclear myocytes and some rare *bona fide* myotubes. Using nuclear labeling and mixing experiments, we present evidence suggesting that the bi-nuclear myocytes originate from acytokinetic mitosis, not cell fusion. Under hypoxia, the short TKO myocytes/myotubes became abnormally thin, yet some rare myocytes/myotubes persisted and twitched. The appearance of rare twitching TKO myotubes with three nuclei suggests that some cell fusion can occur, albeit inefficiently, in the absence of all three Rb protein family. We conclude that the presence of a single member of the Rb protein family is required for efficient myocyte fusion, survival and differentiation even under hypoxia. However, we cannot rule out the possibility that under certain conditions, yet to be identified, TKO myoblasts might fuse to form normal-like twitching myotubes. After submission of this manuscript, the Sage group reported that Rb-family TKO embryos form various tissues containing multiple cell lineages. However, skeletal myotubes were completely absent in cross sections through back/axial muscles of TKO embryos [@pone.0017682-Wirt1]. Although more detailed analyses of the TKO muscle defect is needed, these observations are consistent with our *in vitro* results demonstrating an autonomous requirement for the Rb family for myogenesis, even under hypoxia. We propose the following active/default model for pRb. In this model, pRb is actively required for differentiation of certain tissues, whereas other tissues, including adipose and those that develop early in embryogenesis before Rb family gene expression is observed [@pone.0017682-Jiang2], can differentiate in the absence of Rb, i.e. as a default pathway. Indeed, it was recently shown that Rb status dictates fate choice between osteogenic and adipogenic differentiation by positively regulating the osteogenic factor RUNX2 and negatively regulating the adipogenic factor PPARγ [@pone.0017682-Calo1]. Likewise, Rb (and p107) is required for differentiation of adipocytes to white adipose tissue by suppressing PGC-1α expression, whereas the default differentiation in the absence of Rb or p107 is brown fat [@pone.0017682-Scime2]. In TKO hematopoietic stem cells, myeloid progenitors hyper-proliferate whereas lymphoid progenitors are ablated [@pone.0017682-Viatour1]. While Rb may be required solely for survival of muscle, bone, white fat and lymphoid cells, these studies suggest that it is actively engaged in sequestering inhibitors of differentiation or stimulating expression or activity of differentiation factors in stem/progenitors cells at the stage of bifurcation into different cell lineages. Whether Rb dictates cell fate choice during differentiation of multipotent stem cells in the somite or in Pax3^+^/Pax7^+^ muscle stem cells in the dermomyotome [@pone.0017682-Buckingham2] is yet to be determined. Such instructive functions by the Rb family during cell fate determination may no longer be required once a cell becomes committed to a specific lineage, and the major function of Rb in committed cells might be to allow proper cell cycle exit and survival; a function that can be bypassed by survival factors or hypoxia. The diminished ability of TKO myoblasts to differentiate and survive may be a consequence of complete deregulation of the E2F protein family, which may create conditions that are incompatible with differentiation, or reflect a requirement for Rb protein family in myoblast fusion and myotube survival, which cannot be rescued by hypoxia. Finally, Rb but not p107 or p130 is often lost in cancer. As a potential basis for this observation, it was suggested that Rb inactivation is uniquely required for cancer progression because only after its loss can tumor cells escape senescence under oncogenic stress [@pone.0017682-Chicas1]. Conversely, in response to p16^ink4a^ over-expression, mutations in p107 plus p130 allow cells to escape cell cycle inhibition as efficiently as mutations in Rb [@pone.0017682-Bruce1], suggesting that these factors have similar functions in cell cycle exit in response to CDK4/6 inhibition. Here, using KO myoblasts, we demonstrated that Rb, but not p107 plus p130, is uniquely required for cell cycle exit during terminal differentiation of primary myoblasts. Thus, the exclusive role of Rb but not its relatives in certain cancers may be due to its unique role in enforcing cell cycle exit during terminal differentiation. Likely, disruption of either function of pRb, senescence or cell cycle exit during differentiation, can lead to neoplastic transformation depending on the cellular and oncogenic context. Methods {#s4} ======= Mouse maintenance, genotyping & timed-pregnancy {#s4a} ----------------------------------------------- Experiments were performed in accordance with guidelines of the Canadian Council on Animal Care and approved by the TGRI-UHN Animal Care Committee, Ontario (Approval ID: AUP1050). Mice were genotyped using DNA extracted from tail biopsies and the following primers: mgRb:Rblox -- forward 5′-ATTTCAGAAGGTCTGCCAAC, reverse 5′-AGAGCAGGCCAAAAGCCAGGA; Rb mutant, AATTGCGGCCGCATCTGCATCTTTATCGC and GAAGAACGAGATCAGCAG; Rb wild-type AATTGCGGCCGCATCTGCATCTTTATCGC and CCCATGTTCGGTCCCTAG; Rb^floxed^ (Rb18 + Rb19E) GGCGTGTGCCATCAATG and CTCAAGAGCTCAGACTCATGG; p130 wild-type ACGGATGTCAGTGTCACG and TACATGGTTTCCTTCAGCGG; p130 mutant ACGGATGTCAGTGTCACG and GAAGAACGAGATCAGCAG; p107 wild-type TCGTGAGCGGATAGAAAG and GTGTCCAGCAGAAGTTA; p107 mutant TCGTGAGCGGATAGAAAG and CCGCTTCCATTGCTCAGCGG. For timed-pregnancies, mice were mated overnight and the day of vaginal plug observation was considered E0.5. Isolation of myoblasts and cell culture {#s4b} --------------------------------------- Skeletal muscles from limbs E16.5--E17.5 embryos obtained following timed-pregnancy were used to generate primary myoblast cultures. More than 700 embryos were used in this study. To maintain consistency between experiments, primary myoblasts were induced to differentiate at passage two. Muscle tissues were digested for 20 min at 37°C in 80 µl solution containing 1.5 U/ml collagenase IV (Sigma), 2.4 U/ml Dispase (Roche) and 5 mM CaCl~2~, gently triturated and plated onto 60 mm collagen-I coated culture dishes. Primary myoblasts were maintained in Growth Medium (GM) - HAM\'s-F10 (Lonza) supplemented with 20% FBS (PAA) and 2.5 ng/ml basic fibroblast growth factor (bFgf) (Sigma) - in a humidified incubator at 5% CO~2~ and 37°C. To induce differentiation, myoblasts were washed once in 1x Phosphate Buffered Saline (PBS) and shifted to Differentiation Medium (DM) - Dulbecco\'s modified Eagle\'s medium (DMEM, high-glucose and sodium pyruvate) (Sigma) supplemented with 3% Horse Serum (PAA) [@pone.0017682-Ho1]. For drug treatment, a single dose of 3-methyladenine (5 mM) was added upon differentiation. Bezafibrate (500 µM) was refreshed every other day. BrdU DNA synthesis assay {#s4c} ------------------------ Post-differentiation day 1 myotube cultures were re-stimulated in GM supplemented with 20 µM BrdU for 16 h before fixation with 3.7% formaldehyde (10 min). For BrdU-labeled cell mixing experiments, cultures were divided into two equal populations and transduced with Ad.cre or Ad.EV (see Ad.cre transduction methods). One of the two populations was fed 20 µM for 36 hr. BrdU was then removed, both populations trypsinized, mixed in equal proportions and plated in GM. Differentiation was induced 12--16 hr later. Cultures were permeabilized using 0.3% Triton X-100 for 10 min, treated with 2N HCl for 25 min, and neutralized with two washes of 0.5 M sodium borate, pH 8.5 for 5 min. After blocking in 1.0% BSA for 20 min, primary anti-MHC antibody; 1:50 (clone MY-32, Sigma) for 1 hr, cells were washed 3x, 3 min each with PBS. Secondary antibody: fluorescein-conjugated (Alexa Fluor 563 (Red) - Invitrogen). BrdU was detected using anti-BrdU antibody directly conjugated to FITC as per manufacturer\'s protocol (BD Biosciences). Images were captured using an Axioskop2 fluorescent microscope (Carl Zeiss Inc.). Immunofluorescence {#s4d} ------------------ 500,000 primary myoblasts were seeded on 22 mm round Collagen-I coated coverslips (BD Biosciences) and induced to differentiate. Cells were fixed in 3.7% formaldehyde, permeabilized in 0.3% Triton X-100 and blocked for 20 min in 1% BSA/PBS at room temperature. Primary antibodies: MHC, 1:50 (clone MY-32, Sigma), pRb, 1:100 (BD Biosciences), were incubated on samples for 1 hr at room temperature. Secondary antibodies, fluorescein-conjugated (Alexa Fluor 563, Alexa Fluor 488 - Invitrogen) were added for 45 min. Nuclei were counterstained with DAPI (Invitrogen) for 10 min and mounted in fluorescent mounting media (Dako). Confocal images of 0.5 µm sections were captured at room temperature using a 40x or 63x c-apochromat objective lens (water)/1.2NA using a Zeiss LSM510 META confocal microscope (Carl Zeiss Inc.) and Zeiss AIM 3.2 acquisition software. Adobe Photoshop CS2 was used to overlay images. Western Blot Analysis {#s4e} --------------------- Cells were lysed on ice in K4IP buffer (50 mM HEPES, pH 7.5, 0.1% Tween-20, 1 mM EDTA, 2.5 mM EGTA, 150 mM NaCl, 1.0 mM DTT, 10% Glycerol) containing protease inhibitors (Sigma). Antibodies were used for 3 hr at room temperature or overnight at 4°C: α/β-tubulin, 1:4000 (Cell Signaling), pRb, 1:1000 (BD Biosciences), p130 (Santa Cruz), p107 (Santa Cruz). Secondary antibodies were HRP-linked anti-IgG, 1:2000 (Cell Signaling) for 1.5 hr in blocking buffer and HRP activity was detected using SuperSignal West Dura chemiluminescent substrate (Pierce) and captured by X-ray film. Films were digitized using a Canon scanner. MitoTracker® Red CMXRos and TUNEL Assays {#s4f} ---------------------------------------- Mitochondria membrane potential was detected using MitoTracker® Red CMXRos according to manufacturer\'s protocol (Molecular Probes). TUNEL analysis on section was performed as described [@pone.0017682-Ho1]. For TUNEL analysis of tissue culture, differentiating myoblasts on collagen-I coated coverslips were fixed in 3.7% formaldehyde for 10 min, washed three times in PBS, permeabilized with 0.3% Triton X-100 solution and washed 3x in PBS. Subsequently, 30U Terminal Deoxynucleotidyl Transferase (TdT) (Fermentas) was added to 50 µl TUNEL-Label Solution (Roche). Nuclei were counterstained with DAPI (Invitrogen) for 10 min and mounted in fluorescent mounting media (Dako). Adenovirus-cre Transductions {#s4g} ---------------------------- Adenoviruses were amplified in 293T cells maintained in DMEM plus 10% FBS and Penicillin/Streptomycin (Sigma). For Adenovirus-cre (Ad.cre; Vector Biolabs) infection, Rb^f/f^ primary myoblasts were infected with multiplicity of infection (MOI) of 1300, which completely eliminated Rb expression. At lower MOI (\<1000), some pRb positive nuclei within myoblasts were detected. For transductions, 50,000 cells (96-well) or 450,000 cells (22 mm round coverslip) were seeded and 8 hr later transduced with Ad.cre or Ad.EV in 50 µl or 600 µl GM, respectively. After 18 hr, medium was re-freshed for additional 24 hr. Cells were rinsed once using 1x PBS and switched to DM. The following adenovirus vectors were kindly provided by Marco Crescenzi - Ad.Bcl-2 - Dept. of Environment and Primary Prevention, Higher Institute of Health, Viale Regina Elena 299, 00161 Roma, Italy; Ad.EV - Genzyme Corporation, 31 New York Ave, P.O. Box 9322, Framingham, MA 01701-9322; David S. Park -- Ad.Rb^ΔK11^ and Ad.p27 - Department of Cellular and Molecular Medicine, University of Ottawa, Ottawa, Ontario, Canada K1H 8M5. Histology and Immunofluorescence on embryo sections {#s4h} --------------------------------------------------- Briefly, embryos were fixed in 4% paraformaldehyde at 4°C overnight, dehydrated, and embedded in paraffin. Sections (8 µm) were cut using a Reichert Ultracut E microtome. For histology, sections were stained with hematoxylin and eosin. For immunofluorescence staining, samples were de-paraffinized, hydrated and subjected to antigen retrieval by boiling in 10 mmol/L sodium citrate (pH 6.0) for 10 min in microwave followed by 30 min gradual cooling at room temperature. Slides were incubated with primary troponin T, 1:200 (clone c-18, Santa Cruz) antibody or MHC, 1:50 (clone MY-32, Sigma) in a humidified chamber at 4°C overnight. Secondary antibody, fluorescein-conjugated (Alexa Fluor 488 - Invitrogen) was added for 1 hr. Nuclei were counterstained with DAPI (Invitrogen) for 10 min and mounted in fluorescent mounting media (Dako). Images were captured using an Axioskop2 fluorescent microscope (Carl Zeiss Inc.). Confocal images of 0.5 µm sections were captured at room temperature using a 40x or 63x c-apochromat objective lens (water)/1.2NA using a Zeiss LSM510 META confocal microscope (Carl Zeiss Inc.) and Zeiss AIM 3.2 acquisition software. Adobe Photoshop CS2 was used to overlay images. Brightfield Images and Videos {#s4i} ----------------------------- Brightfield images and videos were captured at room temperature using 20x or 40x air-objective lenses on a Nikon TE200 microscope (Nikon) fitted with a Hamamatsu CCD digital camera. Images were acquired using SimplePCI imaging software (Hamamatsu). Adobe Photoshop CS2 was used to enhance clarity and contrast using same parameters for control and experimental samples. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### **A rare TKO myotube containing 3 nuclei.** Immunostaining for pRb (green) of Ad.EV and Ad.cre transduced Rb^f/f^:p107^−/−^:p130^−/−^ cultures at DM-2. Top row, low exposure images demonstrating absence of detectable pRb in Ad.cre transduced culture. Bottom row, high exposure images to highlight a short myotube containing 3 nuclei, which is devoid of detectable nuclear pRb, in Ad.cre transduced Rb^f/f^:p107^−/−^:p130^−/−^ culture. Nuclei counterstained with DAPI. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Video S1 ::: {.caption} ###### **Twitching wild-type myotubes treated with 3-MA.** Control wild-type myoblasts were induced to differentiate and treated with 3-MA and video captured at DM-5. (SWF) ::: ::: {.caption} ###### Click here for additional data file. ::: Video S2 ::: {.caption} ###### **Twitching mgRb:Rb** ^−**/**−^ **:p130** ^−**/**−^ **myotubes treated with 3-MA.** mgRb:Rb^−/−^:p130^−/−^ myoblasts were induced to differentiate and treated with 3-MA and video captured at DM-5. (SWF) ::: ::: {.caption} ###### Click here for additional data file. ::: Video S3 ::: {.caption} ###### **Twitching Ad.cre transduced Rb^Δf^ myotubes under hypoxia.** Rb^f/f^ myoblasts were transduced with Ad.cre, induced to differentiate after 48 hr, transferred to hypoxia and video captured at DM-5. (SWF) ::: ::: {.caption} ###### Click here for additional data file. ::: Video S4 ::: {.caption} ###### **Twitching Ad.cre transduced Rb^Δf^:p107** ^−**/**−^ **myotubes under hypoxia.** Rb^f/f^:p107^−/−^ myoblasts were transduced with Ad.cre, induced to differentiate after 48 hr, transferred to hypoxia and video captured at DM-5. (SWF) ::: ::: {.caption} ###### Click here for additional data file. ::: Video S5 ::: {.caption} ###### **Twitching Ad.cre transduced Rb^Δf^:p130−/− myotubes under hypoxia.** Rb^f/f^:p130^−/−^ myoblasts were transduced with Ad.cre, induced to differentiate after 48 hr, transferred to hypoxia and video captured at DM-5. (SWF) ::: ::: {.caption} ###### Click here for additional data file. ::: Video S6 ::: {.caption} ###### **Twitching Ad.cre transduced Rb^Δf^:p107** ^−**/**−^ **:p130** ^−**/**−^ **myotube/myocyte under hypoxia.** Rb^f/f^:p107^−/−^:p130^−/−^ myoblasts were transduced with Ad.cre, induced to differentiate after 48 hr, transferred to hypoxia and video captured at DM-5. (SWF) ::: ::: {.caption} ###### Click here for additional data file. ::: We thank Zhe Jiang for advice, HuQin Li for TUNEL analysis on histological sections, Michael Ohh for access to hypoxia incubator, Marco Crescenzi, Karen A. Vincent and David S. Park for valuable adenovirus vectors. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This study was funded by the Canadian Institutes of Health Research (<http://www.cihr-irsc.gc.ca/e/193.html>), grant number: MOP-93674. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: GC ATH EZ. Performed the experiments: GC ATH. Analyzed the data: GC ATH EZ. Contributed reagents/materials/analysis tools: DC EZ. Wrote the paper: GC EZ. [^2]: ¤ Current address: Groupe Myologie, Paris VI Faculté de Médicine Site Pitié-Salpétrière, Paris, France

PubMed Central

2024-06-05T04:04:19.808280

2011-3-10

PMC3053374

Introduction {#s1} ============ Acetylcholine (ACh) is the major peripheral neurotransmitter controlling the parasympathetic and the sympathetic autonomic nervous system as well as the somatic motor system. Moreover, the cholinergic system is thought to play key roles in many functions in the CNS, including the control of locomotor activity, emotional behavior, and higher cognitive processes such as learning and memory [@pone.0017611-Everitt1]--[@pone.0017611-Taly1]. Changes in cholinergic neurotransmission are associated with a variety of important neurological disorders including Alzheimer\'s disease, schizophrenia, Parkinson\'s disease, epilepsy and attention-deficit hyperactivity disorder [@pone.0017611-Scarr1]. ACh changes cellular activity of target cells through metabotropic muscarinic receptors [@pone.0017611-Wess1], [@pone.0017611-Wess2] and ionotropic nicotinic receptors [@pone.0017611-Taly1], [@pone.0017611-Grutter1]. The brain expresses five different types of muscarinic receptors (M1--M5). The nicotinic receptors, which are formed by five identical or homologous subunits, are generated from twelve different subunits (nine α-subunits and three β-subunits) [@pone.0017611-Taly1]. The various pentameric nAChR subunit combinations have different pharmacological and kinetic properties, and are widely distributed in the brain. Similar complexity is observed for the different G-coupled muscarinic receptors. Knowledge of the interplay between different receptors is not fully understood, and because of this complexity, defining the actual contribution of brain ACh to specific behaviors has been challenging. There have been several attempts to generate animal models of cholinergic dysfunction by elimination of cholinergic neurons using electrolytic or excitotoxic methods, which are nonselective and destroy indistinctly both noncholinergic and cholinergic neurons, as well by the more selective strategy of cholinergic immunolesion, which preferentially destroy cholinergic neurons [@pone.0017611-Everitt1]. Although these studies have provided important information regarding the cholinergic system, they also have raised a number of inconsistent results concerning behavioral processes that are affected by altering cholinergic transmission [@pone.0017611-Everitt1]. The fact that some of these techniques may not be specific and can eliminate non-cholinergic neurons or that they may not eliminate all cholinergic neurons could explain some of the differences. In addition, other signalling molecules, such as neuropeptides, growth factors and co-transmitters, can be co-released by cholinergic neurons, further confounding the interpretation of neuronal degeneration-induced cholinergic deficiency. Furthermore, neuronal death causes inflammation which can also complicate interpretation of the experiments [@pone.0017611-Weisman1]--[@pone.0017611-Wilms1]. Therefore it is important to develop alternative, more consistent and targeted approaches to complement these previous studies and to investigate specific roles of ACh in brain functions. Using genetics to generate mouse models of cholinergic deficiency is equally challenging. ChAT KO mice die shortly after birth and adult heterozygous ChAT KO mice exhibit compensatory increases in choline uptake and show no behavioral phenotype [@pone.0017611-Misgeld1], [@pone.0017611-Brandon1]. We have recently generated novel mouse lines of cholinergic deficiency by targeting the vesicular acetylcholine transporter (VAChT knockdown - VAChT KD and VAChT knockout - VAChT ^del/del^). VAChT is essential for ACh release as mice null for VAChT expression do not survive [@pone.0017611-deCastro1]. In contrast, mice with reduction of VAChT expression by 40% (VAChT KD^HET^) and 70% (VAChT KD^HOM^) are viable [@pone.0017611-Prado1]. Analysis of ACh release in VAChT KD mice indicate that decreased expression of VAChT perturbs storage of ACh in vesicles. During stimulation, impaired ACh storage becomes more pronounced leading to significant decrease in ACh release [@pone.0017611-Prado1], [@pone.0017611-Lima1]. VAChT KD^HOM^ mice are myasthenic and present social and object recognition memory deficits [@pone.0017611-Prado1] and cardiac dysfunction [@pone.0017611-Lara1], indicating that perturbation of ACh storage affects several physiological functions [@pone.0017611-deCastro1], [@pone.0017611-Prado1], [@pone.0017611-Lara1]--[@pone.0017611-deCastro2]. All these phenotypes can be rescued by inhibition of cholinesterase, indicating that they are the result of decreased ACh release due to the exocytosis of partially-filled synaptic vesicles and are not the result of developmental changes [@pone.0017611-Prado1]--[@pone.0017611-Lara1]. The organization of the VAChT gene locus is complex. The entire VAChT open reading frame is encoded by one single exon that is contained inside the first intron of the ChAT gene [@pone.0017611-Eiden1]. This nested gene structure is frequently named cholinergic gene locus (CGL). Control of expression of VAChT and ChAT is poorly understood, and distinct cholinergic neurons show different requirements for regulatory regions within the cholinergic gene locus [@pone.0017611-Naciff1]--[@pone.0017611-Schutz2]. To further investigate the roles of the cholinergic system we have developed novel strains of VAChT targeted-mice. Our strategy was to generate a VAChT allele that is flanked by loxP sequences and carries a TK-Neo resistance cassette approximately 1.5kb downstream from the VAChT stop codon, in a ChAT intronic region. We show that interrupting the intron between ChAT exons N and M with a TK-Neo cassette maintains VAChT expression in the somatomotor subset of cholinergic neurons relatively intact, but causes a pronounced decrease in VAChT expression in other groups of cholinergic neurons in the CNS. As a consequence, these mice present preserved neuromuscular function, but altered brain cholinergic activity. We show that these new mutant mice are hyperactive when exposed to a new environment. Interestingly, hyperactivity is a behaviour trait found in several diseases such as Alzheimer\'s disease [@pone.0017611-Harper1]--[@pone.0017611-GilBea1], schizophrenia [@pone.0017611-Balla1], [@pone.0017611-Mattsson1] and Attention-deficit hyperactivity disorder [@pone.0017611-MehlerWex1], [@pone.0017611-Granon1]. Genetic removal of the TK-Neo resistance cassette rescues VAChT expression and the hyperactivity phenotype. These results suggest that release of ACh is normally required to "turn down" neuronal circuits controlling locomotion. Results {#s2} ======= Generation of VAChT-deficient mice {#s2a} ---------------------------------- We generated a new VAChT targeted mouse line by inserting a lox-P flanked TK-Neo cassette in the 3′ region of the VAChT gene, in the intron between exons N and M of the ChAT gene, and a third lox-P sequence 260 bp upstream from the VAChT translational initiation codon ([Figure 1](#pone-0017611-g001){ref-type="fig"}). Successful recombination of the mutated VAChT allele was confirmed by Southern-blot and PCR analyses ([Fig. 1C and 1D](#pone-0017611-g001){ref-type="fig"}). ::: {#pone-0017611-g001 .fig} 10.1371/journal.pone.0017611.g001 Figure 1 ::: {.caption} ###### Schematic Drawing of the Cholinergic Gene Locus and Generation of VAChT Deficient Mice. a\) Boxes represent the different exons of ChAT or VAChT. The position of the initiation codon (ATG) for VAChT and ChAT and the stop codon (Stop) of VAChT are indicated. Potential transcription initiation sites are indicated for VAChT (green arrowheads) and ChAT (orange arrowheads). Note that the VAChT gene is within the first intron of ChAT. b) Different VAChT alleles generated. P1, P2, P3 and P4 indicate the primers used for PCR genotyping and the fragment sizes generated. LoxP sequence, some restriction enzymatic sites and probe annealing are represented. c) Southern blot analysis of WT (lane 1), VAChT^FloxNeo/FloxNeo^(lane 2), VAChT^WT/FloxNeo^ (lane 3), VAChT^WT/Del^(lane 4) and VAChT^FloxNeo/Del^(lane 5). d) PCR analysis of VAChT^WT/FloxNeo^ (lanes 1 and 4), VAChT^WT/WT^ (lane 2), and VAChT^FloxNeo/FloxNeo^ (lanes 3 and 5). e) PCR analysis of VAChT^WT/del^ mice (lane 1), VAChT^FloxNeo/Del^ (lane 2), and VAChT^WT/FloxNeo^ mice (lane 3). f) PCR analysis of VAChT^Flox/Flox^ mice (lane1), VAChT^WTWT^ mice (lane2 and 4), and VAChT^WT/Flox^ (lane 3 and 5). :::  ::: We initially characterized this novel mouse line by evaluating VAChT expression expecting that the new location chosen for the insertion of the TK-Neo cassette would not alter VAChT gene expression. However, we found that VAChT^FloxNeo/FloxNeo^ mice showed a large decrease in VAChT expression in the striatum (76% decrease in VAChT mRNA- [Figure 2A](#pone-0017611-g002){ref-type="fig"}), but VAChT expression in the spinal cord was decreased only by 46% ([Figure 2B](#pone-0017611-g002){ref-type="fig"}). We have found in previous experiments that decreased expression of VAChT up to 50% in the spinal cord does not alter neuromuscular function [@pone.0017611-deCastro1], [@pone.0017611-Prado1]. As the VAChT^FloxNeo^ allele showed much pronounced decrease of VAChT expression in the brain compared to the spinal cord, it offered the chance to knock-down VAChT expression in the brain, but preserve peripheral cholinergic function. To examine this possibility, we crossed VAChT^FloxNeo/FloxNeo^ mice to heterozygous VAChT-null mice in order to generated VAChT^FloxNeo/del^ mice anticipating that this novel mouse line might present even more significant knockdown of VAChT in the brain, but relatively preserved peripheral function. Genotyping of these mice was obtained by PCR ([Fig. 1E](#pone-0017611-g001){ref-type="fig"}). ::: {#pone-0017611-g002 .fig} 10.1371/journal.pone.0017611.g002 Figure 2 ::: {.caption} ###### VAChT mRNA expression is changed in VAChT mutant mice. a\) VAChT, ChAT, CHT1 and AChase mRNA levels in striatum and b) spinal cord of WT and VAChT^FloxNeo/FloxNeo^ mice. c) VAChT, ChAT, CHT1 and AChase mRNA levels in striatum and d) spinal cord of VAChT^WT/WT^, VAChT^FloxNeo/Del^ and VAChT^WT/Del^ mice. mRNA expression levels were quantified by qPCR using actin to normalize the data. Graphs represent average of 4--6 different mice. (\*) and (\*\*) indicate p\<0.01 and p\<0.001 respectively. :::  ::: We examined VAChT expression in VAChT^FloxNeo/del^ mice compared to VAChT^wt/wt^ mice. The levels of mRNA for VAChT were decreased in the striatum of VAChT^FloxNeo/del^ mice even further (89% decrease - [Fig. 2C](#pone-0017611-g002){ref-type="fig"}). Similarly to VAChT^FloxNeo/FloxNeo^ mice, VAChT expression in the spinal cord of VAChT^FloxNeo/del^ mice was relatively preserved (57% decrease, [Figure 2D](#pone-0017611-g002){ref-type="fig"}). Confirming results obtained previously, levels of mRNA for VAChT^wt/del^ mice decreased 50% when compared to VAChT^wt/wt^ mice ([Figure 2C](#pone-0017611-g002){ref-type="fig"} and [@pone.0017611-deCastro1]). We also examined other components of cholinergic nerve terminals that can impact cholinergic tone. Of significant interest both VAChT^FloxNeo/del^ and VAChT^wt/del^ presented an increase in ChAT mRNA expression in the spinal cord, but not in the striatum ([Fig. 2C and 2D](#pone-0017611-g002){ref-type="fig"}) and this was compatible with previous findings for the *del* allele [@pone.0017611-deCastro1]. This increase in ChAT expression is likely related to the removal of the VAChT gene with a decrease in the distance between two ChAT promoters (see [Fig. 1](#pone-0017611-g001){ref-type="fig"}). In contrast, ChAT mRNA expression was not changed in VAChT^FloxNeo/FloxNeo^ mice ([Fig, 2A and 2B](#pone-0017611-g002){ref-type="fig"}). Also, CHT1 and AChase mRNA expression were not changed in any of the VAChT mutants in either the striatum or spinal cord ([Fig. 2A--D](#pone-0017611-g002){ref-type="fig"}). To investigate the expression of VAChT in distinct brain regions we used immunofluorescence. In brain sections VAChT expression was drastically reduced in the striatum, cortex and hippocampus of VAChT^FloxNeo/del^ and VAChT^FloxNeo/FloxNeo^ mice compared to VAChT^wt/del^ or VAChT^wt/wt^ mice ([Fig. 3](#pone-0017611-g003){ref-type="fig"} and [4](#pone-0017611-g004){ref-type="fig"}). In contrast CHT1 labelling was preserved. We also examined expression of VAChT in the facial motor nuclei ([Fig. 4](#pone-0017611-g004){ref-type="fig"}). Staining in the cell bodies was similar in all genotypes, although both VAChT^FloxNeo/del^ and VAChT^FloxNeo/FloxNeo^ mice had a decreased labelling in the punctated fluorescence for nerve terminals that contact these neurons. To further investigate if indeed VAChT expression was reduced in the brain, we used immunoblot analysis of striatum tissues. A decrease of 75 to 85% in the expression of VAChT in the striatum of VAChT mutants was observed (supplementary [Fig. S1](#pone.0017611.s001){ref-type="supplementary-material"}). ::: {#pone-0017611-g003 .fig} 10.1371/journal.pone.0017611.g003 Figure 3 ::: {.caption} ###### VAChT immunorreactivity is altered in VAChT mutant mice. a\) Representative optical sections from striatum stained with a VAChT antibody (green) or b) stained with CHT1 antibody (green). c) Representative optical sections from hippocampus stained with a VAChT antibody or d) CHT1 antibody (green). Dapi labelling (blue) was used to stain nuclei. Scale bar 50 µm. :::  ::: ::: {#pone-0017611-g004 .fig} 10.1371/journal.pone.0017611.g004 Figure 4 ::: {.caption} ###### VAChT immunorreactivity is altered in VAChT mutant mice. a\) Representative optical sections from cortex stained with a VAChT antibody (green) or b) CHT1 antibody (green). c) Representative optical sections from facial motor nuclei stained with a VAChT antibody or d) CHT1 antibody (green). Dapi labelling (blue) was used to stain nuclei. Scale bar 50 µm. :::  ::: To further explore VAChT expression in the periphery we stained VAChT in the NMJ of diaphragm. In contrast to the decreased VAChT expression in distinct brain regions, we found negligible differences for VAChT expression in nerve-endings at the diaphragm of VAChT^FloxNeo/del^ and VAChT^FloxNeo/FloxNeo^ mice when compared to the two control genotypes ([Fig. 5A](#pone-0017611-g005){ref-type="fig"}). Furthermore, analysis of nicotinic ACh receptor labelling using fluorescent bungarotoxin suggested normal nAChR distribution ([Fig. 5A](#pone-0017611-g005){ref-type="fig"}). ::: {#pone-0017611-g005 .fig} 10.1371/journal.pone.0017611.g005 Figure 5 ::: {.caption} ###### NMJ morphology and transmission in VAChT mutant mice. a\) Diaphragms were immunolabelled with VAChT antibody (green) and α-bungarotoxin (red) to label nicotinic receptors. Right columns show overlay pictures. Dapi (blue) was used to stain nuclei. Images are representative of 3 independent experiments. WT control mice (a.1--3), VAChT^FloxNeo/FloxNeo^ mice (a.4--6), WT control mice (a.7--9), VAChT^WT/Del^ mice (a.10--12), and VAChT^FloxNeo/Del^ mice (a.13--15). No alterations were observed between the genotypes. Scale bar 50µm. b) Quantal size of the four genotypes quantified by plotting the cumulative frequency of MEPP amplitudes. WT control (black line), VAChT^WT/Del^ mice (red line), VAChT^FloxNeo/FloxNeo^ mice (blue line) and VAChT^FloxNeo/Del^ mice (green line). c) Frequency of MEPPs at synapses for the four genotypes. (\*) indicates statistically significant difference from control wild-type mice (two-way ANOVA followed by Bonferroni post hoc; F(2,14) = 21,98, p\<0.005). :::  ::: To test if neuromuscular transmission was preserved in VAChT^FloxNeo/del^ and VAChT^FloxNeo/FloxNeo^ mice we recorded from the NMJ of the diaphragm. Both the amplitude and the frequency of miniature end-plate potentials (MEPPs) were increased for VAChT^FloxNeo/del^, VAChT^FloxNeo/FloxNeo^ and VAChT^wt/del^ mice when compared to VAChT^wt/wt^ ([Fig. 5B and C](#pone-0017611-g005){ref-type="fig"}). These results are compatible with our previous observations that close to 50% reduction of VAChT at neuromuscular junctions affects quantal release of ACh only mildly [@pone.0017611-Prado1]. These results suggest that quantal release in the two mutant mice with decreased expression of VAChT in the brain was well preserved at the NMJ. To examine whether VAChT^FloxNeo/del^ and VAChT^FloxNeo/FloxNeo^ mice had preserved muscular function we performed a series of neuromuscular tests. These mutant mice showed no difference in grip-force and wire-hang tasks suggesting preserved neuromuscular function ([Fig. 6A--D](#pone-0017611-g006){ref-type="fig"}). Moreover, because previous observations showed that VAChT KD^HOM^ mice had gait problems [@pone.0017611-Prado1], we also tested if VAChT^FloxNeo/del^, VAChT^FloxNeo/FloxNeo^ and VAChT^wt/del^ mice presented any gait abnormality. In agreement with the lack of neuromuscular phenotype, we found no gait deficiency in these mutant mice ([Fig. 6 E--F](#pone-0017611-g006){ref-type="fig"}). These results indicate that VAChT^FloxNeo/del^ and VAChT^FloxNeo/FloxNeo^ mice, unlike VAChT-KD^HOM^ mice, do not present a neuromuscular phenotype. ::: {#pone-0017611-g006 .fig} 10.1371/journal.pone.0017611.g006 Figure 6 ::: {.caption} ###### Neuromuscular function in VAChT mutant mice. a\) Time spent hanging upside down from a wire netting for WT and VAChT^FloxNeo/FloxNeo^ mice. No significant difference was observed \[T~(33)~ = 295, P = 0.728\]. b) Wire Hang for WT, VAChT^WT/Del^ and VAChT^FloxNeo/Del^ mice. No significant difference was observed \[Kruskal-Wallis, H~(2)~ = 2.604, P = 0.272\]. c) Grip force measured for WT and VAChT^FloxNeo/FloxNeo^ mice. There is no significant difference between the two genotypes \[T~(21)~ = 125, P = 0.689\]. d) Maximal force expressed in gram. No difference was observed between WT, VAChT^WT/Del^ and VAChT^FloxNeo/Del^ mice \[One way ANOVA, F~(2)~ = 0.600, P = 0. 507\]. e) Gait analysis for WT, VAChT^FloxNeo/FloxNeo^ mice. No significant difference between genotypes was revealed \[Student test, t~(13)~ = 0.263 P = 0. 797\] f) Gait analysis for WT, VAChT^WT/Del^ and VAChT^FloxNeo/Del^. No significant difference between genotypes was observed \[One way ANOVA, F~(2)~ = 0.699, P = 0. 559\]. :::  ::: We have previously demonstrated that decreased VAChT expression leads to proportional increase in the amount of total ACh in the brains of mutant mice, as ACh that is not released accumulates in nerve terminals [@pone.0017611-deCastro1], [@pone.0017611-Prado1]. Because we did not detect any alteration in either ChAT or CHT1 in the striatum, we measured the amount of ACh in the brains of mutant mice as an indirect assessment of ACh output. We determined the ACh content in the striatum of VAChT^FloxNeo/del^ mice (the line with largest decrease in VAChT expression) and VAChT^wt/del^ mice. VAChT^FloxNeo/del^ mice presented several-fold more ACh in the striatum than wild-type controls, whereas the increase in VAChT^wt/del^ mice was around two-fold ([Fig. 7](#pone-0017611-g007){ref-type="fig"}). These data show a gene-dosage effect in the ACh content in these VAChT mutant mice and corroborate the mRNA and protein findings that the decrease in VAChT expression is more accentuated in VAChT^FloxNeo/del^ when compared to VAChT^WT/del^ mice. ::: {#pone-0017611-g007 .fig} 10.1371/journal.pone.0017611.g007 Figure 7 ::: {.caption} ###### Acetylcholine content in the striatum VAChT mutant mice. Striatal tissue ACh levels for WT, VAChT^WT/Del^ and VAChT^FloxNeo/Del^ mice were assayed by chemiluminescent detection. Data represent 4--9 experiments (mean ± SEM). (One-way Anova with Bonferroni post hoc, F~(2,14)~ = 21,98, (\*)p\<0. 05\* for wild-type controls. :::  ::: VAChT^FloxNeo/del^ and VAChT^FloxNeo/FloxNeo^ mice are hyperactive {#s2b} ------------------------------------------------------------------ As these new VAChT mutant mice have preserved peripheral function, they became candidates to explore phenotypes that were previously difficult to study using VAChT KD^HOM^ mice due to their neuromuscular deficiency. To start assessing the consequences of decreased VAChT expression for central functions we examined locomotor activity in the open field, which has been shown to be altered by antagonists of muscarinic receptors as well as by genetic elimination of some nicotinic and muscarinic receptors [@pone.0017611-Granon1]--[@pone.0017611-Maubourguet1]. [Figure 8](#pone-0017611-g008){ref-type="fig"} indicates that VAChT^FloxNeo/del^ mice showed increased locomotion throughout the 2 hour monitoring period when compared to wild-type controls (Kruskal-Wallis test show difference between the genotypes (H~(3)~ = 31.680, p\<0.001). The average total distance traveled by VAChT^FloxNeo/del^ mice in 2 h was 2.1-fold higher than that of WT controls. An intermediate increase (1.4-fold) in locomotor activity was observed in VAChT^FloxNeo/FloxNeo^ mice when compared to wild-type controls ([Fig. 8A and B](#pone-0017611-g008){ref-type="fig"}). Activity levels of VAChT^wt/del^ mice showed a tendency to increase however it did not meet statistical significance, similar to previously reported observations [@pone.0017611-deCastro1]. These results suggest that decreased VAChT expression to the levels found in VAChT^FloxNeo/del^ mice causes abnormal motor activity. In addition, vertical exploration in the open field was increased in VAChT^FloxNeo/del^ mice as shown by the number of rearings ([Figure 8C and D](#pone-0017611-g008){ref-type="fig"}; Kruskal-Wallis test; (H~(3)~ = 13.764, p\<0.05; post-hoc Dunn reveal a significant higher rearing number of VAChT^FloxNeo/del^ compared to WT controls). ::: {#pone-0017611-g008 .fig} 10.1371/journal.pone.0017611.g008 Figure 8 ::: {.caption} ###### VAChT mutant mice are hyperactive. a\) Spontaneous horizontal activity during two hours in the open field for WT, VAChT^WT/Del^, VAChT^FloxNeo/FloxNeo^ and VAChT^FloxNeo/Del^ mice. b) Total spontaneous horizontal activity during the two hour was increased for VAChT^FloxNeo/FloxNeo^ and VAChT^FloxNeo/Del^ mice compared to WT/WT. But no difference between WT and VAChT^WT/Del^ was observed. c) Spontaneous vertical activity during two hours in the open field for WT, VAChT^WT/Del^, VAChT^FloxNeo/FloxNeo^ and VAChT^FloxNeo/Del^ mice. d) Total number of rearings during the two hour period. Rearings for VAChT^FloxNeo/Del^ were significantly higher when compared to WT (Kruskal-Wallis test; (H~(3)~ = 13.764, post-hoc Dunn p\<0.05). (\*) indicate p\<0.01. :::  ::: Lack of habituation does not seem to be the cause of the hyperactivity as all three mutants showed decreased motor activity across the 2-hour test session ([Figure 8A](#pone-0017611-g008){ref-type="fig"}). Moreover, VAChT^FloxNeo/FloxNeo^ and VAChT^FloxNeo/del^ mice were retested after 24 h and 48 h under the same conditions to investigate intersession habituation and both genotypes showed significant decrease in locomotor activity in the second and third days \[*between-sessions habituation* in the open field; [Figure 9A](#pone-0017611-g009){ref-type="fig"}, two-way repeated measures ANOVA- main effect of genotype,*F* ~(2,\ 86)~ = 15.825, *p*\<0.001, day *F* ~(2,\ 86)~ = 35.318, *p*\<0.001 and interaction genotype x day *F* ~(4,\ 86)~ = 2.505, *p*\<0.05\], further suggesting no impairment in habituation in the novel environment. It is important to note that even after the third day, both VAChT^FloxNeo/FloxNeo^ and VAChT^FloxNeo/del^ mice remained hyperactive when compared to WT (Tukey test respectively P\<0.01 and P\<0.001). ::: {#pone-0017611-g009 .fig} 10.1371/journal.pone.0017611.g009 Figure 9 ::: {.caption} ###### Habituation and anxiety are not changed in VAChT mutant mice. a\) Habituation to open field during 3 consecutive days for WT, VAChT^FloxNeo/FloxNeo^ and VAChT^FloxNeo/Del^ mice. Mice showed no impairment in habituation in the novel environment. Two-way repeated measures ANOVA- main effect of genotype,*F* ~(2,\ 86)~ = 15.825, *p*\<0.001, day *F* ~(2,\ 86)~ = 35.318, *p*\<0.001 and interaction genotype x day *F* ~(4,\ 86)~ = 2.505, *p*\<0.05\]. b) Time spend in the centre during the 2 hour in the open field for WT, VAChT^WT/Del^, VAChT^FloxNeo/FloxNeo^ and VAChT^FloxNeo/Del^ mice. VAChT^FloxNeo/del^ mice spent significant more time in the center of the open field apparatus (Kruskal-Wallis test, H~(3)~ = 11.537, p\<0.05; post-hoc Dunn\'s method, *p*\<0.05). c) Time spend in the open arm of elevated plus maze for WT, VAChT^FloxNeo/FloxNeo^ and VAChT^FloxNeo/Del^ mice was not significantly affected. One Way Analysis of Variance F~(2,58)~ = 1,603, NS). d) Number of entries in the open arm of elevated plus maze for WT, VAChT^FloxNeo/FloxNeo^ and VAChT^FloxNeo/Del^ mice was not significantly affected. One Way Analysis of Variance (F~(2,58)~ = 1,845, NS). (\*), (\*\*) and (\*\*\*) indicate p\<0.05, p\<0.01 and p\<0.001 respectively. :::  ::: We also tested for changes in anxiety level. The time spent in the center vs. the periphery of the open field was evaluated in the same open field trials used to quantify locomotor movement. As shown in [Figure 9B](#pone-0017611-g009){ref-type="fig"} VAChT^FloxNeo/del^ mice spent significantly more time in the center of the open field apparatus (Kruskal-Wallis test, H~(3)~ = 11.537, p\<0.05; post-hoc Dunn\'s method, *p*\<0.05) which could be an indication of reduced anxiety [@pone.0017611-Crawley1]. However, when we assessed the willingness of VAChT mutant mice to explore a novel unprotected environment (open arms) of the elevated plus maze, the time spent in the open arms ([Fig. 9C](#pone-0017611-g009){ref-type="fig"}; One Way Analysis of Variance F~(2,58)~ = 1,603, NS), and the number of entries in the open arms in the elevated plus maze test ([Fig. 9D](#pone-0017611-g009){ref-type="fig"}; One Way Analysis of Variance F~(2,58)~ = 1,845, NS) were not significantly affected in VAChT^FloxNeo/del^ mice. These data indicate that VAChT^FloxNeo/del^ mice do not show consistent changes in anxiety-related behaviors. Genetic rescue of VAChT-mutant mice hyperactive behavior {#s2c} -------------------------------------------------------- If decreased VAChT expression causes hyperactivity, it would be expected that correcting VAChT levels should allow for rescue of this phenotype. The VAChT^FloxNeo^ allele carries a TK-Neo cassette 3′ from the ORF of VAChT and this is likely the cause of decreased VAChT expression ([Fig. 1](#pone-0017611-g001){ref-type="fig"}). Cre excision of loxP flanked DNA sequences is a stochastic event [@pone.0017611-Rajewsky1], we therefore crossed VAChT^FloxNeo/wt^ mice to distinct Cre mice (see [Methods](#s4){ref-type="sec"}) to obtain an allele in which the TK-Neo cassette was deleted ([Fig. 1B](#pone-0017611-g001){ref-type="fig"}; VAChT^Flox^ allele). We screened the offspring from this cross by PCR to identify founder mice carrying only the floxed VAChT gene, with removal of the TK-Neo cassette. VAChT floxed founders (VAChT^Flox^) were crossed to C57BL/6J mice to confirm germ-line transmission and the progeny obtained were intercrossed to obtain VAChT^Flox/Flox^ mice and WT controls (PCR in [Fig. 1F](#pone-0017611-g001){ref-type="fig"}). We investigated VAChT expression at the mRNA and protein levels and found that VAChT^Flox/Flox^ mice have essentially the same level of expression for this transporter as VAChT^wt/wt^ mice in the striatum, cortex, spinal cord and hippocampus ([Fig. 10](#pone-0017611-g010){ref-type="fig"}). Moreover, ChAT and CHT1 expression were not changed in VAChT^Flox/Flox^ mice ([Fig. 10](#pone-0017611-g010){ref-type="fig"}). Accordingly VAChT^Flox/Flox^ mice showed no deficits in neuromuscular function in the grip-force ([Fig. 11A](#pone-0017611-g011){ref-type="fig"}) or wire-hang (not shown). Measures of anxiety in the elevated plus maze were identical to measures of WT controls ([Fig. 11B and C](#pone-0017611-g011){ref-type="fig"}). When we tested VAChT^Flox/Flox^ mice in the open-field we also found that locomotor activity was identical to that of VAChT^wt/wt^ mice and no habituation deficits were observed ([Fig. 11D--F](#pone-0017611-g011){ref-type="fig"}). These results strongly suggest the recovery of VAChT expression by removal of the TK-Neo cassette rescued the hyperactivity phenotype of VAChT^FloxNeo^. ::: {#pone-0017611-g010 .fig} 10.1371/journal.pone.0017611.g010 Figure 10 ::: {.caption} ###### Genetic rescue of VAChT-mutant mice. a\) VAChT, ChAT, and CHT1 mRNA levels were measured by qPCR in the striatum of WT mice (white bar), VAChT^WT/Flox^ (grey bar) and VAChT^Flox/Flox^ (black bar) mice. b) VAChT, ChAT, CHT1 mRNA levels in the spinal cord of WT mice (white bar), VAChT^WT/Flox^ (grey bar) and VAChT^Flox/Flox^ (black bar) mice. c) Representative immunoblot of control, VAChT^WT/Flox^ and VAChT^Flox/Flox^ mice in striatum. d) Quantification of protein levels. Actin immunoreactivity was used to correct for protein loading between experiments. Data are presented as a percentage of wild-type levels. e) Representative immunoblot of control, VAChT^WT/Flox^ and VAChT^Flox/Flox^ mice in spinal cord. f) Quantification of protein levels. Actin immunoreactivity was used to correct for protein loading between experiments. Data are presented as a percentage of wild-type levels. :::  ::: ::: {#pone-0017611-g011 .fig} 10.1371/journal.pone.0017611.g011 Figure 11 ::: {.caption} ###### Restoration of normal phenotype by removing of the Neo-cassette. a\) Spontaneous horizontal activity during two hours in the open field for VAChT^Flox/Flox^ mice. The total locomotion is similar in both genotype (t~(36)~ = −0.769 P = 0.447). b) Grip force for VAChT^Flox/Flox^ mice. (t~(12)~ =  −1.414 P = 0.183) c) Time spent in the open arm of elevated plus maze for VAChT^Flox/Flox^ mice. No difference in anxiety level was observed (t ~(20)~ = −0,670, P = 0,510). d) Number of entries in the open arm of elevated plus maze for WT and VAChT^Flox/Flox^ mice. e) Spontaneous horizontal activity during two hours in the open field for VAChT^Flox/Flox^ mice. f) Habituation to open field during 3 consecutive days. The ANOVA reveal no effect of genotype (F~(1,44)~ = 0.475, P = 0.498), a significant effect of the factor day (F~(2,44)~ = 16.733, P\<0.001) and no interaction genotype x day (F~(2,44)~ = 0.364, P = 0.697). Post-hoc showed difference between the Day1 and Day2, 3. (\*\*\*) indicate p\<0.001. :::  ::: Discussion {#s3} ========== The VAChT^FloxNeo^ allele shows differential regulation of VAChT expression {#s3a} --------------------------------------------------------------------------- The present experiments explore some of the remarkable features of the cholinergic gene locus to target VAChT and generate mice with decreased cholinergic function. We show that interference with the VAChT-ChAT locus, by insertion of a TK-Neo cassette in the intron between ChAT exons N and M, differentially affected the expression of VAChT in the brain and the spinal cord. Owing to the relative preservation of cholinergic function in the spinal cord and NMJ, we were able to show that one of the consequences of reduction of VAChT expression in the forebrain, and consequent reduction of ACh release, is hyperactivity. This phenotype shows a gene-dose effect with lesser expression of VAChT causing a more pronounced hyperactivity. VAChT has a unique genomic organization; its open reading frame is encoded within the first intron of the ChAT gene. This arrangement [@pone.0017611-Eiden1] is conserved in nematode [@pone.0017611-Alfonso1], [@pone.0017611-Alfonso2], Drosophila [@pone.0017611-Kitamoto1] and mammals [@pone.0017611-Erickson1]. Transcriptional control of the CGL is rather complex as multiple promoters and alternative splicing are used to generate different mRNA species from both VAChT and ChAT genes [@pone.0017611-Cervini1]. Transgenic mice containing different DNA segments of the CGL fused to reporter genes have been used to identify regulatory regions that are important for the expression of VAChT and ChAT *in vivo* [@pone.0017611-Naciff1], [@pone.0017611-Schutz1], [@pone.0017611-Schutz2], [@pone.0017611-Kitamoto1], [@pone.0017611-Kitamoto2]--[@pone.0017611-Yasuyama1]. These studies indicate that multiple regulatory elements are necessary to control expression in the CGL and suggest that regulation of the CGL is different in different types of cholinergic neurons. Moreover, this regulatory strategy seems to be conserved in insects and vertebrates [@pone.0017611-Schutz2], [@pone.0017611-Lee1]. A core promoter containing regulatory elements necessary to activate the CGL in cholinergic cells and to repress its activity in non-neuronal cells is present in the sequence spanning approximately 4 kb upstream of the R exon [@pone.0017611-Lonnerberg1], [@pone.0017611-Lonnerberg2]. Other regulatory elements have been described in the genomic region between exon-M and the first ChAT coding exon [@pone.0017611-Schutz2], however the complete set of regulatory sequences controlling the CGL remains to be determined. A cholinergic group-specific transcriptional activator has been identified in *Drosophila*. Mutant flies that lack expression of the transcription factor abnormal chemosensory jump6 (acj6) showed decreased ChAT in primary olfactory neurons, whereas expression in mechanosensory neurons was unaffected [@pone.0017611-Lee1]. Our results give further support to the subset-specific regulation of the CGL. Because the TK-Neo cassette used to generate the VAChT^FloxNeo^ allele was introduced 450 bp upstream from the beginning of M-exon, it is reasonable to suggest that its presence interfered with the function of additional regulatory elements. As sensorymotor cholinergic neurons rely mainly on the core promoter [@pone.0017611-Schutz1], [@pone.0017611-Schutz2], VAChT expression in these neurons may be relatively preserved while all the other groups of cholinergic neurons in the brain have pronounced decrease in VAChT expression. Interestingly, ChAT expression was not altered in the VAChT^FloxNeo^ allele. This might suggest that VAChT and ChAT rely on different regulatory elements. In contrast, mice that present the VAChT^del^ allele showed an increased expression of ChAT in the spinal cord but not in the striatum. These results agree with our previous experiments showing increased ChAT expression in the spinal cord of VAChT^del^ mice [@pone.0017611-deCastro1]; see also [Fig. 2](#pone-0017611-g002){ref-type="fig"} and [3](#pone-0017611-g003){ref-type="fig"}). This occurs likely due to the proximity of the M-promoter of ChAT to VAChT promoters after excision of the intervening DNA sequences flanked by loxP (see [Fig. 1](#pone-0017611-g001){ref-type="fig"}). Our experiments suggest that whereas ChAT expression in the spinal cord may be regulated by elements that were modified by the del allele, the missing genomic fragment does not seem to be necessary for regulation of ChAT expression in the striatum ([Fig. 2C and D](#pone-0017611-g002){ref-type="fig"}). Overall, our experiments examining VAChT and ChAT expression point to differential gene regulation between the striatum, and likely other forebrain regions, and the spinal cord. The significance for this differential regulation for normal cholinergic physiology is poorly understood, but likely plays an important role to maintain proper expression level of these two critical cholinergic genes in these distinct sets of neurons. Although unlikely, we cannot discard the possibility that expression of the neomycin resistance protein (aminoglycoside 3′-phosphotransferase) may partially contribute to the phenotypes observed. The fact that VAChT expression was rescued by the removal of the TK-Neo cassette shows unequivocally that the two loxP sequences that flank the VAChT gene do not alter transcription. VAChT^FloxNeo/del^ and VAChT^FloxNeo/FloxNeo^ mice are hyperactive {#s3b} ------------------------------------------------------------------ As decreased VAChT expression leads to proportional decrease in ACh release in the brain [@pone.0017611-deCastro1], [@pone.0017611-Prado1] the availability of mutant mouse lines displaying different levels of VAChT expression in the brain (VAChT^WT/del^ mice: 50% decrease; VAChT^FloxNeo/FloxNeo^ mice: 75% decrease; VAChT^FloxNeo/del^ mice: 85% decrease) provided us with unique tools to evaluate the consequences of reduced VAChT levels for brain functions. Also, understanding the consequences of decreased VAChT expression is made easier now by the new mouse lines that do not show confounding peripheral phenotypes. Because ACh is known to play a major role in the regulation of locomotor control [@pone.0017611-Di1], we used these mutants to investigate the role of VAChT in locomotor activity. We found that up to 50% decrease in VAChT expression in the brain does not change locomotor activity in mice, similar to previous experiments with VAChT KD^HET^ mice and VAChT^wt/del^ mice. However, our data show clearly that a more pronounced decrease in VAChT expression causes hyperactivity in a new environment. These results suggest that release of ACh is normally required to regulate neuronal circuits controlling locomotion. Injection of muscarinic antagonists in distinct brain regions cause pronounced augmentation in locomotor activity levels [@pone.0017611-Molinengo1], [@pone.0017611-Shannon1], [@pone.0017611-Ukai1] and a hyperactivity phenotype was observed in mouse strains lacking M1 and M4 muscarinic receptors [@pone.0017611-Gerber1], [@pone.0017611-Miyakawa1], [@pone.0017611-Gomeza1] as well as mouse strains null for the nicotinic β2 receptor [@pone.0017611-Granon1], [@pone.0017611-Granon2], [@pone.0017611-Maubourguet1]. Paradoxically, systemic injections of nicotinic agonists can cause an increase in locomotor activity [@pone.0017611-Reavill1]--[@pone.0017611-Panagis1]. However, this effect should be considered with caution, as the hyperactivity most probably results from desensitization of specific types of nicotinic receptors due to prolonged activation. Therefore, ACh may regulate locomotor circuitry in multiple and redundant ways. Our data provide additional support to the notion that insults that cause cholinergic presynaptic deficiency can also increase activity. Locomotor hyperactivity is a symptom present in many disorders including Attention Deficit Hyperactivity Disorder (ADHD), schizophrenia, Alzheimer\'s diseases and some forms of autism [@pone.0017611-Scarr1]. Interestingly, all these disorders have in common some degree of cholinergic deficit. VAChT^FloxNeo/del^ and VAChT^FloxNeo/FloxNeo^ mice are novel complementary models to understand the specific consequences of decreased cholinergic activity in the brain and should be useful to further investigate the role of ACh in distinct brain functions. Importantly, as VAChT^Flox/Flox^ mice have preserved VAChT expression and do not show any phenotype, they can be used in the future to generate novel lines with suppression of ACh release in specific brain regions. These conditional mutants will be valuable to investigate the role of specific groups of cholinergic neurons in distinct brain functions. Materials and Methods {#s4} ===================== Ethics Statement {#s4a} ---------------- The experimental procedures in this study were conducted in compliance with the Canadian Council of Animal Care (CCAC) guidelines for the care and use of animals. The protocol was approved by the University of Western Ontario Institutional Animal Care and Use Committee (protocol \# 2008-089). All efforts were made to minimize the suffering of animals. Generation of VAChT mutant mice {#s4b} ------------------------------- Construction of the gene-targeting vector was described previously [@pone.0017611-deCastro1]. In short, one LoxP sequence was placed 260 bp upstream from the VAChT translational initiation codon, and a second LoxP was added approximately 1.5 kb downstream from the stop codon. The Neomicin-resistance gene (TK-Neo cassette) was inserted immediately after the second LoxP and was followed by a third LoxP ([Figure 1](#pone-0017611-g001){ref-type="fig"}). The linearized targeting vector was electroporated into J1 embryonic stem cells derived from 129/terSv mice, and selected embryonic stem cell clones harbouring homologous recombination (determined by PCR and Southern blotting (not shown) were injected into C57BL/6J blastocysts to produce chimeric mice. Germ line transmission was achieved, and mice were bred to C57BL/6J mice to produce heterozygous mutant mice (VAChT^WT/FloxNeo^). Heterozygous mice were intercrossed to generate the homozygous (VAChT^FloxNeo/FloxNeo^) and wild-type controls (VAChT^WT/WT^) used in these experiments. VAChT^FloxNeo/del^ and VAChT^WT/del^ mice were generated by intercrossing VAChT^WT/FloxNeo^ to heterozygous VAChT KO mice (VAChT^wt/del^; [@pone.0017611-deCastro1]. Only male mice were used in this study. Animals were housed in groups of three to four per cage in a temperature-controlled room with a 12∶12 light-dark cycles in microisolator cages. Food and water were provided ad libitum. Mouse colonies were maintained at the University of Western Ontario, Canada, in accordance with Canadian Council of Animal Care (CCAC) guidelines for the care and use of animals. Genotyping, Southern blotting {#s4c} ----------------------------- Genotyping by PCR was performed using tail DNA as a template. The set of three primers used were P1 (5-GAGAGTACTTTGCCTGGGAG GA -3), P2 (5- GGCCACAGTAAGACCTCCCTTG -3), P3 (5- GCAAAGCTGCTATTGGCCGCTG -3) and P4 (5-TCATAGCCCCAAGTGGAGGGAGA-3). For Southern blot analysis, genomic DNA was digested with the enzymes *Bam*HI and *Sac*I. Digested DNA was subjected to electrophoresis in a 1.5% agarose gel and transferred onto a nylon membrane. After UV cross-linking, DNA on the membrane was hybridized to the NdeI/PmeI VAChT DNA fragment (see [Fig. 1](#pone-0017611-g001){ref-type="fig"} for the position of the probe fragment). Detection was done using the Alkphos direct labelling and detection system kit (GE Healthcare) according to the manufacturer\'s instructions. qPCR {#s4d} ---- For real-time quantitative PCR (qPCR), total RNA was extracted using the Aurum Total RNA for fatty and fibrous tissue kit from Biorad. Quantification and quality analysis of RNA in the extracted samples was done by microfluidic analysis (Agilent Technologies\' Bioanalyzer). First-strand cDNA was synthesized using the iSCRIPT cDNA SYNTHESIS KIT from Biorad. cDNA was subsequently subjected to qPCR on a CFX-96 Real Time System (Bio-Rad) using the iQ SYBR GREEN SUPERMIX (Bio-Rad). For each experiment, a non-template reaction was used as a negative control. In addition, the absence of DNA contaminants was assessed in reverse transcription-negative samples and by melting-curve analysis. Relative quantification of gene expression was done with the ΔΔ*CT* method using β-actin gene expression to normalize the data. Western blotting {#s4e} ---------------- Immunoblot analysis was carried out as described previously [@pone.0017611-deCastro1]. Antibodies used were anti-VAChT (rabbit polyclonal 1∶2000, Synaptic System, Germany), anti-CHT1 (rabbit polyclonal 1∶1000, kindly provided by R. Jane Rylett, University of Western Ontario, London, Canada), anti-CHAT (rabbit polyclonal 1∶1000, Chemicon) and anti-actin (Chemicon, CA). Images were acquired using the FluorChem Q System from Alpha Innotech and analysed using the AlphaVie software. Immunofluorescence analysis of brain slices were performed as described previously [@pone.0017611-deCastro1]. Images were acquired using an Axiovert 200 M using the ApoTome system or a LEICA SP5 confocal microscope as previously described [@pone.0017611-deCastro2]. Tissue ACh measurements {#s4f} ----------------------- Brains were dissected rapidly, homogenized in 5% TCA, and centrifuged (10,000×*g* for 10 min) at 4°C. Supernatants were frozen at −80°C until use. For ACh determinations, TCA was removed with ether, and a chemiluminescent assay was done with choline oxidase as described previously [@pone.0017611-Cervini1]. The data are presented as means and standard errors of the means (SEM). One-way analysis of variance (ANOVA), followed by Bonferroni\'s test, was used to analyze the differences in tissue ACh concentrations in VAChT^FloxNeo/del^, VAChT^FloxNeo/FloxNeo^,VAChT^WT/del^ and wild-type controls (VAChT^WT/WT^); a *P*\<0.05 was considered to be statistically significant. Electrophysiology {#s4g} ----------------- Recordings were performed on isolated hemi-diaphragm nerve-muscle preparations. Animals were euthanized and the diaphragm with attached rib bone was rapidly dissected and placed into Tyrode\'s solution containing NaCl (124 mM), KCl (5 mM), NaHCO~3~ (26 mM), NaH~2~PO~4~ (1.2 mM), MgCl~2~ (1.3 mM), CaCl~2~ (2.4 mM), glucose (10 mM). This solution was gassed with a mixture of 5%CO~2~/95%O~2~ and in this condition had a pH of 7.4. The diaphragm was bisected, and one half was transferred into a custom recording chamber in which the muscle was held in place with metal pins that passed through the surrounding tissue and were inserted into a Sylguard bed. During recording, the muscle was continuously perfused with gassed Tyrode solution containing 0.0003 mM tetrodotoxin to avoid spontaneous action potentials. Borosilicate (WPI) microelectrodes were fabricated on a Narashige PN-30 puller, and had resistances of 5--15 MOhm when filled with 3 M KCl. Fine branches of the motor nerve were visually identified under 100X magnification and the muscle fiber was impaled using a fine micromanipulator (WPI). Membrane potential and synaptic potentials were amplified 10X with an Axon Instruments Axoclamp 2A, and membrane potential was monitored throughout the experiment. To digitalize the miniature endplate potentials (MEPPs), the signal was high-pass filtered at 0.1 Hz to subtract the resting potential and amplified a further 200-1000X using a Cyberamp (Axon Instruments) amplifier. This signal was fed to a Lab Master A--D conversion board controlled by Strathclyde Electrophysiology Software (University of Strathclyde, Glasgow, Scotland). To measure quantal size, a software event detector was used to record 25 ms of data on either side of the MEPP. The threshold of the event detector was set just below the peak of the noise so as not to miss any small MEPPs. Under these conditions, approximately 15% of detected "MEPPs" were false positives and were manually detected and removed. To measure MEPP frequency, membrane potential was recorded without selection, and MEPPs were manually identified and counted. Grip force and wire-hang {#s4h} ------------------------ Mice were brought to the testing room and allowed to acclimatize for 10 minutes before initiating tests. A Grip Strength Meter from Columbus Instruments (Columbus, OH) was used to measure forelimb grip strength as an indicator of neuromuscular function as described previously [@pone.0017611-deCastro1], [@pone.0017611-Prado1]. Briefly, the grip strength meter was positioned horizontally and mice were held by the tail and lowered toward the apparatus. Mice were allowed to grasp the smooth, metal, triangular pull bar (forelimbs only) and were then pulled backward in the horizontal plane. The force applied to the bar at the moment the grasp was released was recorded as the peak tension (kg)`.` The test was repeated 10 consecutive times within the same session and the highest value from the 10 trials was recorded as the grip strength for that animal. Mice were not trained prior to testing and each mouse was tested once (10 trials equal one test session). For wire-hanging experiments the laterals of a cage top were covered with tape to prevent the mice to reach the borders [@pone.0017611-Sango1]. The mouse was gently put on the cage top, which was then briefly shaken to induce the mouse to grasp the wire in the top. The cage top was then inverted and suspended approximately 40 cm above an empty cage. Time spent hanging upside down was determined with a cut-off time of 60 sec. Gait analysis {#s4i} ------------- Mice were subjected to gait assessment [@pone.0017611-Neumann1] using a CatWalk automated gait analysis system (Noldus Information Technology). The apparatus is made of a 1.3 m long glass plate with dim fluorescent light beamed into the glass from the side. The reflexion of the paw in contact to the glass was recorded by a video camera. Mice were placed in the walkway and allowed free exploration for 1 min before recording the first run. A minimum of 3 correct runs (the mouse cross the walkway with no interruption or hesitation) for each mouse was recorded. Runs were analysed using the Noldus software and only the runs where it was possible to discern all steps were used for the analysis. We only used the mean stride length of hind paw as data, stride length is the distance between two successive prints of the same paw. Locomotor activity and habituation {#s4j} ---------------------------------- Locomotor activity was automatically recorded (AccuScan Instrument, Inc. Columbus, OH). The open field arena was a 20 cm×20 cm platform surrounded by 30 cm high walls. Mice were acclimated to the testing room for 20 minutes prior to beginning the test, and had not experienced a cage change for at least 24 hours. Mice were placed in the center of the apparatus and allowed to freely explore the arena. Horizontal locomotion and rearings were recorded and used as measures of locomotion and exploration, respectively [@pone.0017611-Vianna1]. Locomotor activity was measured at 5 min intervals and cumulative counts (120 min) were taken for data analysis as described elsewhere [@pone.0017611-Gainetdinov1]. For the intersession habituation, mice were exposed for 120 min to the same open field during 3 consecutive days. Measurements of total activity were obtained and one-way ANOVA and Tukey\'s Multiple Comparison Test was used to test for statistical significance. Activity was measured by the Versamax software. Elevated Plus-maze {#s4k} ------------------ Animals were placed in the center of the elevated plus maze (Med Associate Inc.) and activity was recorded for five minutes with a webcam connected to a computer. Total amount of time spent in the open and in the open sections of the maze was calculated with the Any-maze software (Stoelting Co., USA); an animal was considered to be completely within a section of the maze when its center of gravity was in this section. The result was expressed as the percentage of time spent in the open arm. Statistical Analysis {#s4l} -------------------- Data were statistically analyzed by a two-tailed Student\'s *t* test or by two-way or repeated measure ANOVA. If data were not normal, we used the adequate non-parametric test. The specific statistical analyses used are noted in the text and legends. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### **Protein expression is changed in VAChT mutant mice.** a) Western blot analysis of VAChT in the striatum of VAChT^FloxNeo/FloxNeo^ mice compared to WT control and b) quantification of protein levels. c) Western blot analysis of VAChT in the striatum of VAChT^FloxNeo/Del^ mice, VAChT^WT/Del^ and VAChT^WT/WT^. d) quantification of protein levels. Actin immunoreactivity was used to correct for protein loading between experiments. Data are presented as a percentage of wild-type levels. Graphs represent average of 4--6 different mice. (\*) indicates statistically different from WT/WT control (Student test, p\<0.05), (\*\*) indicates statistically different from VAChT^WT/Del^ (Student test, p\<0.01). (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: We appreciate for technical assistance and outstanding care for mouse colonies from Jue Fan and Sanda Raulic. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was supported by the Canadian Institutes of Health Research (<http://www.cihr-irsc.gc.ca/e/193.html>) (CIHR; MOP-89919; for V.F.P. & M.A.M.P), Canadian Foundation for Innovation (CFI), the Ontario Research Fund (ORF) and the University of Western Ontario (V.F.P. & M.A.M.P.). M.G.C. received support from the National Institutes of Health (NIH). X. De J. received a PhD fellowship from CAPES (Brazil). C.K. was funded by CNPq and FAPEMIG (Brazil). No additional external funding was received for this study. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: VFP MAMP CK. Performed the experiments: CM-S XDJ MSG RDFL MSS. Analyzed the data: VFP MAMP CK CM-S XDJ MSG. Contributed reagents/materials/analysis tools: VFP MAMP MGC MVG CK. Wrote the paper: VFP. Revised manuscript: MAMP MGC MVG CK. [^2]: ¤a Current address: Department of Physiology, Universidade Federal do Espírito Santo, Vitoria, Espírito Santo, Brazil [^3]: ¤b Current address: Department of Psychiatry, University of California San Francisco, San Francisco, California, United States of America

PubMed Central

2024-06-05T04:04:19.812424

2011-3-10

PMC3053375

Introduction {#s1} ============ In developing countries, micronutrient deficiencies such as folic acid and vitamin B~12~ are common and associated with poor pregnancy outcomes, having long term health effects. The peri or post-conceptional period represents a sensitive window during which suboptimal maternal micronutrients may affect feto-placental development [@pone.0017706-Owens1]. In view of this, maternal folic acid supplementation is in operation for the last few decades in India. This is regardless of the fact that there is still a poor knowledge about potential adverse effects on a population mainly consuming a vegetarian diet that may lack vitamin B~12~ [@pone.0017706-Yajnik1]. Although early results from the fortification policy indicate beneficial effects in terms of reduction in plasma homocysteine [@pone.0017706-Choumenkovitch1]--[@pone.0017706-Jacques1], recent reports suggest adverse effects in humans [@pone.0017706-Windham1], [@pone.0017706-Nelen1]. During growth high folic acid administration has been shown to alter dietary protein metabolism and to decrease fetal size, as compared to rats fed a control diet [@pone.0017706-Achn1]. It has also been reported recently, that high folate intakes in vitamin B~12~ deficient mothers are shown to increase the risk of type 2 diabetes in the offspring suggesting that defects in one-carbon metabolism might be at the heart of intrauterine programming of adult disease [@pone.0017706-Yajnik2]. Our earlier studies in the rat model that maternal folic acid supplementation at marginal protein levels reduces the levels of brain essential polyunsaturated fatty acid levels especially omega 3 fatty acids in the offspring [@pone.0017706-Rao1], [@pone.0017706-Pita1]. Ongoing studies in our lab have also highlighted the importance of docosahexaenoic acid (an omega 3 fatty acid) during pregnancy [@pone.0017706-Kilari1], [@pone.0017706-Dangat1]. Docosahexaenoic acid (DHA) is an indispensable component of all cell membranes and is incorporated in high concentrations in the membrane phospholipids of brain and retina [@pone.0017706-Fliesler1] and its availability during the perinatal period is shown to be associated with long term cognitive and visual development [@pone.0017706-Helland1], [@pone.0017706-Whalley1]. It is well established that folate and vitamin B~12~ are the major determinants of one carbon metabolism in which S-adenosyl methionine (SAM) a methyl group donor is formed [@pone.0017706-Selhub1]. Dietary folate is converted in the body to 5 methyl tetrahydrofolate (5-MTHF) by the enzyme methylene tetrahydrofolate reductase (MTHFR). The transfer of methyl group from 5-MTHF to homocysteine requires vitamin B~12~ and results in the synthesis of methionine. Methionine is the precursor for SAM. Methyl groups from SAM are transferred by phosphatidyl ethanolamine--N--methyltransferase (PEMT) to DHA and to DNA and histones by the respective methyltransferases. Phosphatidylcholine (PC) is critical for the delivery of important polyunsaturated fatty acids (PUFA) such as docosahexaenoic acid from the liver to the plasma and distribution to peripheral tissues. [Fig. 1](#pone-0017706-g001){ref-type="fig"} shows the interactions of folic acid, vitamin B~12~ and DHA. We have recently described that when DHA levels are low, there will be less methyl group requirement for conversion of PE-DHA to PC-DHA and may result in excess methyl group availability for other trans-methylation reactions such as DNA and histone methylation leading to altered chromatin remodeling and gene expression [@pone.0017706-Kale1]. We therefore hypothesize that omega 3 fatty acids play a key role in one carbon metabolism affecting global methylation levels. ::: {#pone-0017706-g001 .fig} 10.1371/journal.pone.0017706.g001 Figure 1 ::: {.caption} ###### One-Carbon Cycle: Interactions of folic acid, vitamin B~12~ and DHA. THF- tetrahydrofolate; 5, 10-MTHF- 5, 10-methylenetetrahydrofolate; 5-MTHF- 5-methyltetrahydrofolate; MTHFR- methylenetetrahydrofolate reductase; MS- methionone synthase; SAH- S-adenosylhomocysteine; SAM- S-adenosylmethionine; DHA-docosahexanoic acid; PE-DHA- phosphatidylethanolamine-DHA; PC-DHA-phosphatidylcholine-DHA; PEMT- phosphatidylethanolamine-N-methyltransferase; DNMT- DNA methyltransferase; HMT- Histone methyltransferase; ↓- reduced; ↑↑-increased. :::  ::: Placenta is an organ whose proper development and function are crucial to the health, growth, and survival of the developing fetus. A number of studies are now making important links between alterations to appropriate epigenetic regulation in the placenta and diseases of gestation and early life [@pone.0017706-Kim1]. Examining epigenetic alterations in the placenta will prove especially important in the search for biomarkers of exposure, pathology, and disease risk and can provide critical insights into the biology of development and pathogenesis of disease [@pone.0017706-Maccani1]. The present study therefore for the first time, examines the effect of normal and excess folic acid in the absence and presence of vitamin B~12~ deficiency on global methylation patterns in the placenta. Further, the effect of maternal omega 3 fatty acid supplementation on the above vitamin B~12~ deficient diets was also examined. Materials and Methods {#s2} ===================== This study was carried out in accordance with the CPCSEA guidelines (Committee for the purpose of control and supervision of experimental animals) Govt of India. This study was approved by the Bharati Vidyapeeth Animal Ethical Committee (IAEC/CPCSEA/258). The institute is recognized to undertake experiments on animals as per the CPCSEA, Govt of India. Animals {#s2a} ------- Wistar albino rats (60F, 20M) of average weight 150 g were obtained from National Toxicology Center animal house. Instead of using them directly for the experimental protocol it was thought appropriate to use their progeny. They were maintained at 22°C on a controlled 12-hr light and 12 hr dark cycle with appropriate ventilation system. Animals were marked with picric acid as H (head), Back (B), Tail (T) etc for identification. Breeding {#s2b} -------- These pups were then put for breeding at 3 months of age. Males were housed individually prior to mating to acquire cage dominance. Virgin female rats were allowed to breed (sex ratio 1:3). On the following morning the vaginal smears were taken to confirm mating. Vaginal smears were taken on a clean microscope slide using a cotton bud dipped in saline. The slides were examined under a microscope at 10× magnification. The sperm positive smear was considered a result of successful mating and considered day 0 of gestation. The pregnant dams were housed individually (in polypropylene cages of 29×22×14 cm dimensions containing rice husk as bedding material). Animals receiving Vitamin B~12~ deficient diets were kept in special cages to prevent coprophagy. Out of 60 females, 47 females became pregnant and were divided randomly into 6 dietary groups. All dams were delivered by C section on day 20 of gestation. Diets {#s2c} ----- The composition of the control and the treatment diets ([Table 1](#pone-0017706-t002){ref-type="table"}) was as per AIN 93 purified diets for laboratory rodents [@pone.0017706-Reeves1]. Protein level in the control and treatment diets was 18%. Total of six isocaloric treatment diets were formulated and has been described by us recently [@pone.0017706-Dangat2], [@pone.0017706-Roy1]. Briefly four diets were formulated for examining the effects of 2 different levels of folic acid (i.e 2 and 8 mg folic acid/kg diet) during pregnancy both in the presence and absence of vitamin B~12~. In addition, 2 more diets were formulated to examine the effects of omega 3 fatty acid (DHA+EPA (Eicosapentaenoic acid)) supplementation on both the vitamin B~12~ deficient groups. Vitamin B~12~ deficiency was obtained exclusively through dietary means. Vitamin free casein was used for all treatment diets. Thus there were a total of 6 groups: Control- normal folate, normal B12, NFBD- normal folate, B12 deficient, NFBDO-normal folate, B12 deficient, omega 3 supplemented, EFB- excess folate, normal B12, EFBD-excess folate, B12 deficient, EFBDO- excess folate, B12 deficient, omega 3 supplemented. ::: {#pone-0017706-t001 .table-wrap} 10.1371/journal.pone.0017706.t001 Table 1 ::: {.caption} ###### Composition of the diets. ::: {#pone-0017706-t001-1} S.No Diets Control (g/kg) NFBD (g/kg) EFB (g/kg) EFBD (g/kg) NFBDO (g/kg) EFBDO (g/kg) ------ -------------------------------------------------- ---------------- ------------- ------------ ------------- -------------- -------------- 1. Corn Starch 398 398 398 398 398 398 2. Casein 200 200 200 200 200 200 3. Dextrinized Starch 132 132 132 132 132 132 4. Sucrose 100 100 100 100 100 100 5. Soya Bean Oil 70 70 70 70 25 25 6 Fish oil 0 0 0 0 45 45 7. Fiber 50 50 50 50 50 50 8. Mineral mixture[\*](#nt101){ref-type="table-fn"} 35 35 35 35 35 35 9 Vitamin mixture[†](#nt102){ref-type="table-fn"} 10 10 10 10 10 10 Folic acid 0.002 0.002 0.008 0.008 0.002 0.008 Vitamin B~12~ 0.025 0 0.025 0 0 0 10 Cystine 3 3 3 3 3 3 11. Choline Bitartarate 2.5 2.5 2.5 2.5 2.5 2.5 12. Tertiary Butyl Hydroquinone 0.014 0.014 0.014 0.014 0.014 0.014 Total Energy (kJ) 1.57 1.57 1.57 1.57 1.57 1.57 \***Mineral mixture (g/kg mixture):**Calcium carbonate, 357; Potassium Phosphate, 196; Potassium Citrate, 70.78; Sodium Chloride, 78; Potassium Sulphate, 46.6; Magnesium Oxide, 24; Ferric Citrate, 6.06; Zinc Carbonate, 1.65; Manganous Carbonate, 0.63; Cupric Carbonate, 0.3; Potassium Iodate, 0.01; Sodium Selenate, 0.01; Ammonium Paramolybdate, 0.007; Sodium Metasilicate, 1.45; Chromium Potassium Sulphate, 0.275; Lithium Chloride, 0.01; Boric Acid, 0.08; Sodium Fluoride, 0.06; Nickel Carbonate, 0.03; Ammonium Vanadate, 0.006; Sucrose, 221.02. † **Vitamin mixture (g/kg mixture):**Nicotinic Acid, 3; Calcium Pantothenate, 1.6; Pyridoxine-HCl, 0.7; Thiamin --HCl, 0.6; Riboflavin, 0.6; D-Biotin, 0.02; Vitamin B~12~ (in 0.1% Mannitol), 2.5;Vitamin E, 15;Vitamin A, 0.8; Vitamin D-3, 0.25;Vitamin K, 0.075;Folic acid, 0.2 (control) and Sucrose 974.655, was used to make total weight of the vitamin mixture to 1 kg. Control: Normal folate, normal B12, NFBD: normal folate, B12 deficient, NFBDO: normal folate, B12 deficient, omega 3 supplemented, EFB: Excess folate, normal B12, EFBD: Excess folate, B12 deficient, EFBDO: Excess folate, B12 deficient, omega 3 supplemented. ::: The lowest level, i.e. 2 mg/kg represents the normal level of folic acid used in the control diet as per the current AIN 93 guidelines while 8 mg/kg is roughly 4 times the requirement of a normal rat. This is in accordance with the fact that folic acid requirement for Indian pregnant woman is set at 400 µg/d, which is 4 times the requirement of a non-pregnant woman. The level of omega 3 fatty acid supplementation was chosen to have an omega 6/omega 3 ratio of 1:1 which is considered to be the ideal ratio [@pone.0017706-Simopoulos1]. Observations recorded {#s2d} --------------------- During pregnancy, dam weights were recorded at day 0, 7, 14 & 20 to obtain weight gains. On day 20 of gestation the litter weight and size was recorded in each group. Maternal plasma folic acid, vitamin B~12~ and homocysteine levels {#s2e} ----------------------------------------------------------------- Plasma vitamin B~12~ and plasma folic acid were measured using a radioimmunoassay kit (Diagnostic Products Corporation, USA) [@pone.0017706-Lee1] and plasma total homocysteine was determined using the IMx System (Abbott Laboratories, IL, USA) [@pone.0017706-Zighetti1]. Tissue collection and processing {#s2f} -------------------------------- Dams were dissected at day 20 of gestation and placental tissues were collected. Fetal membranes were trimmed off and the placenta was weighed. Placentas were snap frozen and stored at −80°C until assayed. Placental fatty acid levels {#s2g} --------------------------- The procedure for fatty acid analysis used in our study was revised from the original method of Manku et al. that has been reported by us earlier in a separate study [@pone.0017706-Manku1], [@pone.0017706-Kulkarni1]. Briefly, placental tissue was homogenized with chilled PBS and centrifuged at 10000 rpm at 4°C for 20 min. Supernantant and cell membrane fractions were separated. Transesterification of cell membrane phospholipid fraction was carried out using hydrochloric acid-methanol. These were separated using a Perkin Elmer gas chromatograph (SP 2330, 30 m capillary Supelco column. Helium was used as carrier gas at 1 mL/min. Oven temperature was held at 150°C for 10 min, programmed to rise from 150 to 220°C at 10°C/min, and at 220°C for 10 min. The detector temperature was 275°C and the injector temperature was 240°C. Retention times and peak areas were automatically computed. The column was calibrated by injecting the standard fatty acid mixture in approximately equal proportion. The data was recorded and the peaks were identified as per the retention time of the standard fatty acids (Sigma) run under the identical conditions. Fatty acids were expressed as g/100 g fatty acid. Total of 15 fatty acids were detected. Saturated fatty acids include myristic acid, palmitic acid and stearic acids, while total monounsaturated fatty acids include myristoleic, palmitoleic, oleic acid and nervonic acids. The omega 3 fatty acids included alpha linolenic acid, eicosapentaenoic acid and docosahexaenoic acid while total omega 6 fatty acids included linoleic acid, gamma linolenic acid, di-homo-gammalinolenic acid, docosapentaenoic acid and arachidonic acid. Placental global methylation patterns {#s2h} ------------------------------------- Genomic DNA extraction from placental tissues was carried out with the Qiagen Blood and Tissue kit. Global DNA methylation was measured using the Methylamp™ Global DNA Methylation Quantification Kit (Epigentek Group Inc., New York, NY, U.S.A.) as we have described recently [@pone.0017706-Kulkarni2]. The kit yields accurate measures of methylcytosine content as a percentage of total cytosine content. The methodology for estimation of global methylation levels used in this study takes into account methylation of all CpG\'s irrespective of their position in the genome (promoter and non-promoter CpG). This is in concurrence with studies indicating that CpG methylation in intragenic and intergenic regions are also critical to gene expression [@pone.0017706-Fazzari1]. Statistical Analysis {#s2i} -------------------- Litter means were used as the unit of analysis. Values are mean ± SD. The data were analyzed using SPSS/PC+ package (Version 11.0, Chicago IL). The treatment groups were compared with the control group by ANOVA and the post-hoc least significant difference test. Results {#s3} ======= Feed intake {#s3a} ----------- Feed intake during pregnancy was between 15--19 g/day. There was no effect of different levels of folic acid both in the presence and absence of vitamin B~12~ on feed intake. In contrast, feed intake in both the omega 3 fatty acid supplemented groups was lower (p\<0.01) than control. Further feed intake in omega 3 fatty acid supplemented groups were also lower than those in their respective B~12~ deficient groups (NFBD Vs NFBDO, p\<0.05) and (EFBD Vs EFBDO, p\<0.05) and has been reported by us recently [@pone.0017706-Roy1]. Reproductive performance {#s3b} ------------------------ There was no effect of folic acid supplementation in the presence of vitamin B~12~ (EFB) on weight gain of dams as compared to control during pregnancy. There was also no difference in weight gain in the dams fed omega 3 fatty acids as compared to control or any of the other treatment groups. The pup weight between groups was comparable and has been reported by us recently[@pone.0017706-Roy1]. Maternal plasma folic acid, vitamin B~12~, homocysteine and fatty acid levels {#s3c} ----------------------------------------------------------------------------- As expected, folic acid supplementation (EFB and EFBD) increased (p\<0.05) plasma folic acid as compared to controls ([Table 2](#pone-0017706-t002){ref-type="table"}). Similarly animals fed a vitamin B~12~ deficient diet had lower (p\<0.05) plasma vitamin B~12~ levels as compared to control. Homocysteine concentrations were comparable between groups. ([Table 2](#pone-0017706-t002){ref-type="table"}) ::: {#pone-0017706-t002 .table-wrap} 10.1371/journal.pone.0017706.t002 Table 2 ::: {.caption} ###### Dam plasma folate, vitamin B~12~ and homocysteine levels. ::: {#pone-0017706-t002-2} Control (n = 8) NFBD (n = 8) EFB (n = 7) EFBD (n = 8) NFBDO (n = 7) EFBDO (n = 6) ----------------------------- ----------------- -------------- ------------------------------------------- -------------- ------------------------------------------- --------------- ------------------------------------------------------------------------------ ------- -------------------------------------------- ------- ---------------------------------------------------------------------------------------------------------------- ------- **Folic acid (ng/ml)** 26.00 14.48 23.50 12.32 69.86 [\*\*](#nt104){ref-type="table-fn"} 8.47 73.88 [\*\*](#nt104){ref-type="table-fn"} 3.36 34.43 16.05 71.00 [\*\*](#nt104){ref-type="table-fn"} 4.43 **Vitamin B~12~ (pg/ml)** 287.63 56.33 192.29[\*\*](#nt104){ref-type="table-fn"} 28.53 274.00 70.58 188.25 [\*\*](#nt104){ref-type="table-fn"} [‡‡](#nt106){ref-type="table-fn"} 33.01 182.00 [\*\*](#nt104){ref-type="table-fn"} 17.52 181.00 [\*\*](#nt104){ref-type="table-fn"} [‡‡](#nt108){ref-type="table-fn"} [§§](#nt109){ref-type="table-fn"} 25.63 **Homocysteine (µmoles/L)** 7.67 1.24 7.53 1.42 6.89 1.48 7.24 1.35 8.97 1.34 7.58 ^\#^ 2.21 \*\**P*\<0.01 when compared to control; † *P*\<0.05, †† *P*\<0.01 when compared to NFBD; ‡ *P*\<0.05, ‡‡ *P*\<0.01 when compared to EFB; §§ *P*\<0.01 when compared to EFBD. AA: Arachidonic acid; Control: Normal folate, normal B~12~, NFBD: normal folate, B~12~ deficient, NFBDO: normal folate, B~12~ deficient, omega 3 supplemented, EFB: Excess folate, normal B~12~, EFBD: Excess folate, B~12~ deficient, EFBDO: Excess folate, B~12~ deficient, omega 3 supplemented. ::: Placental fatty acid levels {#s3d} --------------------------- DHA levels were significantly (p\<0.05) reduced in both the NFBD and EFBD groups as compared to control ([Table 3](#pone-0017706-t003){ref-type="table"}). In contrast, supplementation with omega 3 fatty acids improved (p\<0.01) DHA and omega 3 fatty acid levels but reduced arachidonic acid and omega 6 fatty acid (p\<0.05) levels in NFBDO as well as EFBDO groups. MUFA (mono-unsaturated fatty acid) levels in the NFBDO group and EFBDO were reduced as compared to NFBD and EFBD groups respectively (p\<0.01 for both). ::: {#pone-0017706-t003 .table-wrap} 10.1371/journal.pone.0017706.t003 Table 3 ::: {.caption} ###### Placental fatty acid levels in different treatment groups. ::: {#pone-0017706-t003-3} Control(n = 15) NFBD(n = 16) EFB (n = 16) EFBD (n = 15) NFBDO (n = 16) EFBDO (n = 14) ----------------------------- ----------------- -------------- ------------------------------------------ --------------- ----------------------------------------- ---------------- --------------------------------------------------------------------------- ------ ----------------------------------------------------------------------------- ------ ----------------------------------------------------------------------------------------------------------------------------------------------- ------ **Alpha linolenic acid** 1.86 4.72 0.62 0.19 0.40 [\*](#nt111){ref-type="table-fn"} 0.26 2.13 [‡‡](#nt116){ref-type="table-fn"} 4.15 1.62 3.75 1.73 4.47 **Linoleic acid** 12.77 1.39 12.01 1.30 12.19 1.73 13.25 [††](#nt114){ref-type="table-fn"} [‡](#nt115){ref-type="table-fn"} 1.20 11.51[\*\*](#nt112){ref-type="table-fn"} 1.80 10.89[\*\*](#nt112){ref-type="table-fn"} [†](#nt113){ref-type="table-fn"} [‡‡](#nt116){ref-type="table-fn"} [§§](#nt117){ref-type="table-fn"} 0.97 **Docosahexaenoic acid** 3.73 0.88 2.94 [\*](#nt111){ref-type="table-fn"} 0.99 3.48 0.64 2.76 [\*\*](#nt112){ref-type="table-fn"} [‡](#nt115){ref-type="table-fn"} 0.57 6.75 [\*\*](#nt112){ref-type="table-fn"} [††](#nt114){ref-type="table-fn"} 2.17 7.72 [\*\*](#nt112){ref-type="table-fn"} [‡‡](#nt116){ref-type="table-fn"} [§§](#nt117){ref-type="table-fn"} 2.07 **Arachidonic acid** 17.98 2.07 17.29 2.22 17.71 1.76 17.33 2.10 12.85 [\*\*](#nt112){ref-type="table-fn"} [††](#nt114){ref-type="table-fn"} 3.35 12.49 [\*\*](#nt112){ref-type="table-fn"} [‡‡](#nt116){ref-type="table-fn"} [§§](#nt117){ref-type="table-fn"} 1.58 **Omega 3 fatty acids** 5.68 4.53 3.68 [\*\*](#nt112){ref-type="table-fn"} 1.09 4.15 0.71 4.98 3.95 9.93 [\*\*](#nt112){ref-type="table-fn"} [††](#nt114){ref-type="table-fn"} 3.28 11.68 [\*\*](#nt112){ref-type="table-fn"} [‡‡](#nt116){ref-type="table-fn"} [§§](#nt117){ref-type="table-fn"} 3.76 **Omega 6 fatty acids** 31.86 3.32 30.49 2.90 31.17 2.88 31.60 2.85 27.35 [\*\*](#nt112){ref-type="table-fn"} [††](#nt114){ref-type="table-fn"} 4.49 26.97 [\*\*](#nt112){ref-type="table-fn"} [‡‡](#nt116){ref-type="table-fn"} [§§](#nt117){ref-type="table-fn"} 1.82 Monounsaturated fatty acids 13.85 1.80 15.08 1.78 13.25 3.18 15.01 [‡](#nt115){ref-type="table-fn"} 3.05 12.38 [††](#nt114){ref-type="table-fn"} 3.49 12.34 [§§](#nt117){ref-type="table-fn"} 2.72 **Saturated fatty acids** 40.61 4.82 43.58 2.39 36.80 [\*](#nt111){ref-type="table-fn"} 6.51 41.59 [‡](#nt115){ref-type="table-fn"} 4.71 44.27 [\*](#nt111){ref-type="table-fn"} 4.61 42.97 [‡](#nt115){ref-type="table-fn"} 4.40 \**P*\<0.05, \*\**P*\<0.01 when compared to control; † *P*\<0.05, †† *P*\<0.01 when compared to NFBD; ‡ *P*\<0.05, ‡‡ *P*\<0.01 when compared to EFB; §§ *P*\<0.01 when compared to EFBD. Control: Normal folate, normal B~12~, NFBD: normal folate, B~12~ deficient, NFBDO: normal folate, B~12~ deficient, omega 3 supplemented, EFB: Excess folate, normal B~12~, EFBD: Excess folate, B~12~ deficient, EFBDO: Excess folate, B~12~ deficient, omega 3 supplemented. ::: Placental global methylation levels {#s3e} ----------------------------------- Global DNA methylation levels in placental tissue were reduced in the EFBD group as compared to control and NFBD group (p\<0.05). In contrast, in the EFBDO group DNA methylation levels were higher (p\<0.05) as compared to the EFBD and were comparable to control ([Fig. 2](#pone-0017706-g002){ref-type="fig"}). ::: {#pone-0017706-g002 .fig} 10.1371/journal.pone.0017706.g002 Figure 2 ::: {.caption} ###### Percent global DNA methylation in Wistar rat placenta. \*p\<0.05 when compared to control (Normal folate, normal B~12~); ^†^ p\<0.05 when compared to NFBD (Normal folate, B~12~ deficient); ^‡^p\<0.05 when compared to EFBD (Excess folate, B~12~ deficient), NFBDO: normal folate, B~12~ deficient, omega 3 supplemented, EFB: Excess folate, normal B~12~ ; EFBDO: Excess folate, B~12~ deficient, omega 3 supplemented. :::  ::: Discussion {#s4} ========== This is the first report that has examined the effect of two levels (normal and excess) of folic acid both in presence and absence of vitamin B~12~ deficiency on global DNA methylation levels in the placenta. Our results indicate 1) altered levels of maternal micronutrients did not influence homocysteine concentrations 2) placental DHA levels were reduced in the vitamin B~12~ deficient groups 3) excess maternal folic acid supplementation in the absence of vitamin B~12~ results in reduced global DNA methylation levels 4) when omega 3 fatty acids were supplemented to the diet with excess maternal folic acid and vitamin B~12~ deficiency, DNA methylation levels revert back to the levels observed in the control group. Our findings for the first time suggest that maternal vitamin B~12~ deficiency, both at normal and excess folic acid levels reduces placental DHA concentrations although mechanisms need to be understood. We and others have previously reported that folic acid alters DHA levels in animals [@pone.0017706-Helland1], [@pone.0017706-Whalley1]. It may be possible that vitamin B~12~ deficiency either results in a virtual deficiency of folic acid thereby affecting methyl group supply or PEMT may be epigenetically altered leading to reduced expression affecting the conversion of PE-DHA (phosphatidyl ethanolamine-DHA) to PC-DHA (phosphatidyl choline-DHA) resulting in lower DHA levels in placenta. The PC/PE ratio also modulates the activity of Delta-5 and Delta-6 desaturases involved in omega 3 and omega 6 PUFA synthesis [@pone.0017706-Smith1]. Reports suggest that an imbalance between folate and vitamin B~12~ during pregnancy could influence imprinting in the embryo, perhaps by an effect on DNA methylation since folates are co-factors and co-substrates for biological methylation and nucleic acid synthesis and also function as regulatory molecules [@pone.0017706-Selley1]. DNA methylation patterns which are largely established in-utero, induce stable changes in gene expression that may be sustained throughout the life span of an individual [@pone.0017706-Waterland1]. Associations of homocysteine with global methylation patterns are not well established. High concentrations of homocysteine have been reported to be associated with reduced DNA methylation potential by some [@pone.0017706-Yi1], [@pone.0017706-Castro1], while others have reported increased DNA methylation [@pone.0017706-Bnsch1], [@pone.0017706-Sharma1]. On the other hand, recently Bromberg et al have reported no association between homocysteine and DNA methylation [@pone.0017706-Bromberg1]. Our findings also suggest that homocysteine concentrations may not be the only determinant of global methylation levels in the placenta. Low folic acid status is often associated with impaired DNA methylation [@pone.0017706-Friso1], affecting gene expression in complex ways [@pone.0017706-Kim2]--[@pone.0017706-McCabe1], however it is not known whether excess folic acid might have any adverse effects on these functions. In a recent study, Min et al.[@pone.0017706-Min1] have also shown that folic acid supplementation in vitamin B~12~ deficient rat did not alter hepatic SAM and SAH (S-adenosyl homocysteine) concentrations and DNA methylation. In our study, at normal folic acid there was no change in placental global DNA methylation levels. However, in contrast at excess folic acid levels in the absence of vitamin B~12~ it was lower as compared to control. It has been reported that effect of folate status on DNA methylation in animals and humans is tissue-, site-, and gene-specific [@pone.0017706-Kim2], [@pone.0017706-McCabe1]. Our results suggest that it may be the ratio of folic acid and vitamin B~12~ that may play an important role in determining global DNA methylation. Evidence from in vivo studies has not clearly established a link between vitamin B~12~ and DNA methylation. However it has been demonstrated in the animal model that a B~12~ deficient diet, disturbs normal homeostasis of one-carbon metabolism in the colonic mucosa and results in diminished genomic DNA methylation and increased uracil misincorporation in DNA [@pone.0017706-Friso2]. Further, it has recently been shown that gene expression patterns change under B~12~ deficient conditions and are recovered by dietary methionine supplementation to B~12~ deficient rats [@pone.0017706-Uekawa1]. For the first time this study has shown that supplementation with omega 3 fatty acids in excess folic acid and vitamin B~12~ deficient group increased placental global DNA methylation to control levels. During early development, there are two waves of demethylation which are followed by a gradual increase in de novo methylation in the embryonic and extraembryonic (which includes the placenta) tissues [@pone.0017706-Geiman1]. Deficiencies of vitamin B~12~ or other abnormalities within the one carbon cycle have been implicated in the development of such placental diseases [@pone.0017706-Ray1]. Our findings suggest that placental maturation and development may be hampered due to vitamin B~12~ deficiency and absence of DHA leading to lower total global methylation (hypomethylation). DHA is reported to play an important role in development and maturation of vital tissues such as brain and placenta [@pone.0017706-deUrquiza1], [@pone.0017706-DuttaRoy1]. Our findings indicate that DHA supplementation restores (increases) the global methylation levels to control levels suggesting that that omega 3 fatty acids especially DHA plays an important role in determining methylation levels in the placenta. Further, our results are in line with previous studies in the rat model which have demonstrated that supplementation with omega-3 fatty acids during pregnancy [@pone.0017706-Grenier1] or post natal life [@pone.0017706-Wyrwoll1] could prevent or limit adverse outcomes of fetal programming. Further studies on gene-specific methylation involved in placental growth and pathology will throw light on the mechanisms to explain the current data. The methodology for estimation of global methylation levels used in this study takes into account methylation of all CpG\'s irrespective of their position in the genome (promoter and non-promoter CpG). This is in concurrence with studies indicating that CpG methylation in intragenic and intergenic regions are also critical to gene expression [@pone.0017706-Fazzari1]. Although this study has not examined methylation at gene specific level, our lab has initiated studies to examine the epigenetic changes occurring at genes associated with one carbon metabolism. Understanding these mechanisms may help in elucidating pathways associated with adverse pregnancy outcomes. In conclusion, changes in maternal micronutrients such as folate, vitamin B~12~ and omega-3 fatty acids could alter the availability of these key metabolites of one carbon cycle in the fetus, providing a direct link between maternal nutrition and placental gene methylation. We acknowledge the help of Mr. Vinayak Dhavale, who took care of the animals. We also thank Dr (Mrs) Vijaya Pandit for her suggestions in carrying out the animal protocol. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was partially funded by Department of Biotechnology, Government of India. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: PCG A. Kale SJ. Performed the experiments: A. Kulkarni KD PS A. Kale. Analyzed the data: PCG A. Kale SJ. Contributed reagents/materials/analysis tools: A. Kulkarni KD PS A. Kale. Wrote the paper: PCG SJ.

PubMed Central

2024-06-05T04:04:19.816492

2011-3-10

PMC3053376

Introduction {#s1} ============ Invasive species offer useful models for studying rapid range expansion in novel environments, which can imply pre-adaptation, phenotypic plasticity or adaptation. Evolutionary aspects of biological invasions have long been neglected, with most past focus being on ecological aspects, but recently, interest in the evolution of invasive species has grown (e.g. [@pone.0017658-Barrett1]--[@pone.0017658-Prentis1]). However, knowledge of population history and historical relationships is a prerequisite for examining the evolution of phenotypic traits that may be subject to selection in the new environment. In particular, one needs to identify the most likely source populations/regions for the invasion and determine whether there were single or multiple introduction events. Have invasive populations undergone a genetic bottleneck? What are the pathways of introduction? What is the extent of contemporary gene flow? Such information is also crucial for understanding the success of invasive species, documenting their colonization modes, and designing measures to limit their expansion (e.g. biological control). Species introductions are sampling events and should therefore generate genetic bottlenecks. In agreement with this prediction, loss of variation is a frequent, although not ubiquitous, feature of introductions [@pone.0017658-Dlugosch1]. Multiple introductions can occur, either at the population level (i.e. one population resulting from introductions from several native populations) or at the regional level (i.e. each invasive population being founded from a different single source population). Such multiple introductions can lead to levels of genetic diversity as high in introduced as in native populations (or regions, respectively), and has been shown in numerous case studies (e.g. [@pone.0017658-Hufbauer1]--[@pone.0017658-Prentis2]). In the introduced range, founder effects tend to increase among-population differentiation and separate introductions may establish differentiated gene pools in different sites through founder effects, subsequent drift and/or responses to selection [@pone.0017658-Dlugosh1], [@pone.0017658-Rosenthal1]. On the contrary, repeated introductions into some sites from different sources may convert among-population variation in the native range into within-population variation in the introduced one, thereby decreasing among-population differentiation (e.g. [@pone.0017658-Kolbe1]--[@pone.0017658-Valliant1]). Recent range expansion and gene flow can also homogenize allele frequencies. Indeed empirical studies of invasive plants reveal that genetic differentiation among invading populations is often diminished relative to differentiation in the native range [@pone.0017658-Bossdorf1]. One of the most problematic invasive plants in Europe and Asia is *Ambrosia artemisiifolia* L. (Asteraceae; also called common ragweed). This wind-pollinated monoecious annual is a common native of North America, and has been introduced to South America, Europe, Asia and Australia, where it has become invasive [@pone.0017658-Bassett1], [@pone.0017658-Taramarcaz1]. It is a successful pioneer and grows abundantly in disturbed habitats, including cultivated fields, roadsides and railways, river banks, construction sites and waste places, on a variety of soil types. *A. artemisiifolia* causes large economic losses by reducing crop yields in agricultural fields (e.g. soybean, sunflower), represents a significant challenge to the management of natural resources [@pone.0017658-Protopopova1], and its massive production of pollen often causes serious allergic problems for humans [@pone.0017658-Bohren1]. Although *A. artemisiifolia* is a self-incompatible annual species ([@pone.0017658-Friedmann1]; but see [Discussion](#s4){ref-type="sec"}) with no vegetative propagation [@pone.0017658-Bassett1], three main characteristics may explain its success as an invader [@pone.0017658-Fumanal1]: enormous production of wind-borne pollen assuring pollination success even of isolated individuals, high fecundity (large plants can produce up to 62000 seeds [@pone.0017658-Bassett1]) and long-term seed dormancy (at least 20 years [@pone.0017658-Lewis1]). Highly infested countries include France (especially the Rhone valley), Italy (especially the Po valley), Hungary and Russia (North Caucasus, Krasnodar territory [@pone.0017658-Reznik1]). In Hungary, about 80% of arable land is colonised and 20% of the population suffers pollen allergies [@pone.0017658-Bohren1], [@pone.0017658-Kiss1]. Ragweed is also found in South America, China (mostly in the Eastern part of the country [@pone.0017658-Chen1]), Australia (mostly along the Eastern coast, across New South Wales and Queensland [@pone.0017658-Bass1], [@pone.0017658-McFadyen1]), Japan and Korea [@pone.0017658-Shin1]. In almost all cases, the species was detected as early as 1900--1950 (or even before), but its explosive spread occurred after 1950 (e.g. [@pone.0017658-Reznik1], [@pone.0017658-Chen1]--[@pone.0017658-McFadyen1], [@pone.0017658-Chauvel1]). Both herbarium records [@pone.0017658-Chauvel1] and recent molecular studies [@pone.0017658-Chun1], [@pone.0017658-Genton1] based on nuclear microsatellite markers suggested multiple independent introductions from North America into France. The main mechanism of ragweed dispersal is probably contamination of crop seed lots (e.g. cereals, sunflower [@pone.0017658-Chauvel1]). Global trade (together with inter-continental travel) has indeed been shown to have a major role in the increasing numbers of biological invasions over the last decades [@pone.0017658-McNeely1], [@pone.0017658-Tatem1]. Other origins are also suspected, such as contaminated bird food (in urban areas) and forage, ship ballast, and military movements [@pone.0017658-Taramarcaz1], [@pone.0017658-Fumanal1], [@pone.0017658-Chauvel1]. Once the species is established, the achenes of *A. artemisiifolia* are mostly dispersed by human activities (achenes do not possess any obvious morphological dispersal mechanism). In several instances, the massive spread of common ragweed has been correlated with major socio-economic transitions that increased the area of disturbed or fallow land, such as during the communist economy (1948--1989) in Eastern Europe, when many sites e.g. extensive border areas and military zones were left uncultivated [@pone.0017658-Kiss1], [@pone.0017658-Pysek1] and the political transitions to young democracies in Eastern Europe, with the closure and cessation of cultivation of many agricultural co-operatives [@pone.0017658-Kiss1], [@pone.0017658-Trk1]. The extensive waste lands generated by the war in former Yugoslavia [@pone.0017658-Taramarcaz1] also probably favoured ragweed expansion. Finally, European common agriculture policies may contribute to some extent to colonization by ragweed when arable land in low-productivity areas is abandoned, creating new suitable habitats for weed expansion [@pone.0017658-Taramarcaz1], [@pone.0017658-Trk1]. To gain insights into the historical relationships among *A. artemisiifolia* populations from the native and introduced ranges, and to shed light on the colonization history of this worldwide invader, we investigated the neutral genetic structure of this species, expanding previous samplings [@pone.0017658-Genton1], [@pone.0017658-Gladieux1] to include additional invaded regions and native populations. In the invasive range, Genton et al. [@pone.0017658-Genton1] previously surveyed French populations and Gladieux et al. [@pone.0017658-Gladieux1] later studied six additional populations from Eastern Europe. We sampled Europe as continuously as possible, and also studied several populations from South America, Asia and Australia. Moreover, in the native range, earlier studies [@pone.0017658-Genton1], [@pone.0017658-Gladieux1] focused on eastern North America whereas we included populations from western North America, therefore covering a larger geographic area and adding potential, hitherto unexplored, source populations. In France, Genton et al. [@pone.0017658-Genton1] found high within-population diversity, low among-population differentiation and no pattern of isolation by distance, indicating that introduced populations probably resulted from a mixture of different native populations. They also observed a cline in diversity away from the putative initial area of introduction, suggesting that range expansion occurred through sequential bottlenecks from the original populations, and not from subsequent new introductions. Gladieux et al. [@pone.0017658-Gladieux1] suggested that Eastern European populations did not originate from the earlier established French populations but rather represented multiple independent introductions from other sources, or introductions from an unidentified highly diverse native population. At the population level, previous studies reported high levels of heterozygote deficiency relative to Hardy-Weinberg equilibrium and null alleles were invoked to explain this result [@pone.0017658-Chun1], [@pone.0017658-Genton1], [@pone.0017658-Gladieux1]. We addressed the following questions: (i) Can we identify the sources of the different invasive populations in the world, especially of the previously unanalyzed populations in Australia, China and South America? Can we confirm that Eastern European populations originated from other sources than Western European populations? And if so, from where did they originate? (ii) Were all invasive populations founded by multiple introductions? (iii) How have the introductions affected the amount and structure of genetic variation in Europe (compared to the patterns observed in the native range)? (iv) In Europe, did the expansion proceed in a stepwise manner, each population being colonized by a neighboring population, or as a result of long-distance dispersal within the continent, or did populations result from independent colonization events from the native range? (v) Do populations exhibit significant heterozygote deficiencies? And if so, can we suggest plausible explanatory mechanism(s)? We used both nuclear and chloroplast microsatellite markers, that differ in their mode of inheritance (biparental vs. maternal only) and mutation rate (higher at nuclear markers [@pone.0017658-Jakobsson1]) and therefore give complementary insights into the invasion history and population dynamics of the species. Materials and Methods {#s2} ===================== Plant material {#s2a} -------------- Leaf material was collected from 32 natural populations: eight from North America, 19 from Europe (including Ukraine and Russia), one from Argentina, two from China and two from Australia ([Table 1](#pone.0017658.t001){ref-type="table"}, [Fig. 1](#pone.0017658.g001){ref-type="fig"}) during summer 2007 and 2008. Twenty plants, located 2--5 m apart from each other, were sampled from each site (except Bronx, Argentina and UKR where only 11, 16 and 18 samples were available, respectively). Most populations covered areas of at least one thousand square meters, and counted several hundred to several thousand plants. Leaf samples were dried in silica gel and stored at room temperature until DNA extraction. Moreover, material from herbarium specimens collected in Japan (seven samples; Takamatsu n°145, 398, 523, 1266, 2733, 3656, 3897) and Korea (two samples; Shin n°19892, 19613) and respectively stored at the Mie University Mycological Herbarium (Tsu, Japan) and Mycological Herbarium of the Korea University (Seoul, South Korea) was included in the analyses, leading to a total of 634 individuals. ::: {#pone.0017658.g001 .fig} 10.1371/journal.pone.0017658.g001 Figure 1 ::: {.caption} ###### Map and nDNA genetic composition (based on Structure) of the studied populations. For each population, the pie represents the membership coefficients to the five clusters inferred in Structure. :::  ::: ::: {#pone.0017658.t001 .table-wrap} 10.1371/journal.pone.0017658.t001 Table 1 ::: {.caption} ###### Geographic and genetic characteristics of the studied populations. ::: {#pone.0017658.t001-1} Population name Country Latitude Longitude Nb samples (nDNA) AR *H* ~e~ *F* ~IS~ Nb samples (cpDNA) Nb hapl Nb private hapl Mean nb of pairw diff among indiv --------------------------------------------------------------------- ----------- ------------- ----------------- ------------------- ----------- ----------- ----------- -------------------- --------- ----------------- ----------------------------------- Utah USA 41°44′44″ N 111°54′41″ W 20 6.00 0.687 0.291 20 3 1 1.326 Montana USA 46°59′24″ N 104°10′48″ W 20 6.27 0.771 0.373 20 3 0 3.200 Minnesota USA 45°52′52″ N 95°22′31″ W 20 7.23 0.813 0.351 20 6 0 1.837 Missouri USA 38°41′05″ N 90°18′45″ W 20 7.40 0.802 0.256 18 4 0 1.843 S. Car. USA 33°59′18″ N 81°01′34″ W 20 6.35 0.760 0.350 19 4 1 0.807 Ontario Canada 44°23′04″ N 78°23′00″ W 20 7.13 0.814 0.261 20 8 2 2.732 Quebec Canada 46°49′27″ N 71°13′38″ W 20 7.47 0.792 0.295 19 2 0 0.199 Bronx USA 40°51′28″ N 73°54′36″ W 11 5.71 0.724 *0,193* 11 1 0 0.000 B Belgium 50°55′25″ N 03°12′48″ E 20 5.24 0.740 0.350 19 6 0 3.813 F France 45°43′57″ N 04°59′49″ E 20 6.97 0.798 0.262 18 6 1 2.320 D1 Germany 47°49′14″ N 12°49′36″ E 20 5.59 0.725 0.188 19 5 1 1.532 IT9 Italy 45°21′36″ N 07°50′06″ E 20 5.77 0.742 0.356 19 4 0 0.901 IT4 Italy 45°28′30″ N 11°43′48″ E 20 5.08 0.719 *0,090* 20 1 0 0.000 IT1 Italy 45°54′11″ N 13°31′16″ E 20 6.42 0.784 0.245 20 3 1 3.705 SLO1 Slovenia 45°54′36″ N 15°27′00″ E 20 7.42 0.829 0.331 20 7 0 2.811 HU2 Hungary 46°37′48″ N 17°17′24″ E 20 7.04 0.794 0.400 18 5 1 2.373 HU6 Hungary 47°20′51″ N 19°26′56″ E 20 7.37 0.825 0.179 20 6 0 3.226 HU3 Hungary 47°07′27″ N 21°45′33″ E 20 6.85 0.773 0.225 19 2 0 0.561 SE3 Serbia 44°58′44″ N 19°36′57″ E 20 6.29 0.772 0.343 18 4 0 2.477 PO1 Poland 50°01′40″ N 20°00′33″ E 20 5.62 0.694 0.327 19 4 0 2.550 PO2 Poland 51°10′00″ N 23°47′60″ E 20 6.16 0.782 0.446 17 5 0 2.956 RO1 Romania 46°32′34″ N 24°34′02″ E 20 4.92 0.709 0.244 19 4 0 1.509 UKR Ukraine 50°43′00″ N 30°51′00″ E 18 4.40 0.634 0.400 18 1 0 0.000 UKR1 Ukraine 48°05′55″ N 37°49′52″ E 20 6.66 0.766 0.190 20 5 0 3.026 UKR2 Ukraine 49°01′38″ N 37°32′10″ E 20 6.64 0.798 0.359 19 4 0 2.807 RU4 Russia 46°20′17″ N 42°07′50″ E 20 5.89 0.760 0.360 20 3 0 1.037 RU5 Russia 50°58′26″ N 39°30′46″ E 20 6.83 0.769 0.271 20 4 0 2.905 Argentina Argentina 26°48′43″ N 65°18′04″ W 16 5.62 0.784 0.411 16 3 2 3.950 Wuhan China 29°09′54″ N 113°12′32″ E 20 5.94 0.727 0.331 19 3 1 1.930 Beijing China 41°36′10″ N 123°48′41″ E 20 4.80 0.734 0.422 20 2 0 1.437 Japan Japan 34°59′43″ N 135°51′23 E 7 \- \- \- 5 3 1 \- Korea Korea 37°33′59″ N 126°58′40\'\' E 2 \- \- \- 2 2 1 \- Austral1 Australia 27°14′26″ S 152°25′19″ E 20 5.43 0.771 0.522 19 3 1 0.561 Austral2 Australia 28°23′37″ S 153°24′13″ E 20 4.65 0.670 0.274 20 1 0 0.000 Mean\_North America (± S. D. among populations) 6.70±0.69 0.770±0.045 0.296±0.060 3.87±2.23 1.49±1.14 Mean\_Europe (± S. D. among populations) 6.17±0.87 0.759±0.048 0.293±0.093 4.16±1.64 2.13±1.18 Mean\_non European invasive populations (± S. D. among populations) 5.29±0.55 0.737±0.045 0.392±0.095 2.40±0.89 1.58±1.52 Populations are grouped into three spatial groups (North America, Europe and non-European invasive populations) and roughly ordered from West to East. Nb samples, number of samples; AR, allelic richness (based on the minimal sample size of 8 individuals); *H* ~e~, expected heterozygosity; *F* ~IS~ estimates in italics were not significant; Nb hapl, number of haplotypes; Mean nb of pairw diff among indiv, mean number of pairwise differences among individuals. ::: Microsatellite procedure {#s2b} ------------------------ DNA was extracted using the DNeasy 96 Plant Kit (QIAGEN). We used a total of nine nuclear and four chloroplast microsatellite markers: three nuclear microsatellite markers (Amb12, Amb30 and Amb82) developed by Genton et al. [@pone.0017658-Genton1], six nuclear markers (Ambart04, Ambart06, Ambart09, Ambart13, Ambart21, Ambart27) described by the Molecular Ecology Resources Primer Development Consortium [@pone.0017658-Molecular1], one universal chloroplast locus (NTCP9 [@pone.0017658-Bryan1]) and three chloroplast markers located in the trnC-ycf6 and rps16 regions, for which we developed primers (c6T448\_F: GATTGG ATA GCC GGC AGA TA; c6T448\_R: TTCCTT TTT CTT GGC CTT CA; s16T148\_F: AGCCGT TCC AAC AAA TGA AA; s16T148\_R: AAA CGA TGT GGT ARA AAG CAA C; s16T690\_F: ACT CAT AGT CCT TTT TAT TTA GCT TCC; s16T690\_R: TTT GAG AAT TAT TGA ACT TGA GTT ATG). We checked by direct sequencing that all differences between cpSSR size variants were due to variable numbers of mononucleotide repeats. Multiplex PCRs were performed, amplifying several loci simultaneously. The 16 µl reaction mix contained 1 µl DNA template, 1X Taq Buffer, 2 mM MgCl2, 0.2 mM of each dNTP, varying concentrations of primers (see below; one primer per pair was fluorescently labelled), and 0.4 U Taq polymerase per primer pair included in the reaction. Primer concentrations were experimentally determined so that the intensity of all microsatellites was high enough to prevent allelic drop-out and allow unambiguous genotyping. For multiplex1, primer concentrations were 0.30 µM for Amb82, Ambart04 and Ambart13, and 0.08 µM for c6T448 and s16T690. For multiplex2A, primer concentrations were 0.60, 0.16 and 0.06 µM for Amb12, Ambart27 and s16T148, respectively. And for multiplex2B, primer concentrations were 0.10, 0.30, 0.30 and 0.20 µM for Ambart06, Ambart09, Ambart21 and NTCP9, respectively. The reaction profile was the following: 40 cycles of denaturation at 95°C for 30 s, hybridization at 50°C (for multiplex1) or 52°C (for multiplex2A and multiplex2B, respectively) for 30 s, and elongation at 65°C for 4 min, followed by a final elongation step of 10 min at 72°C. Locus Amb30 was amplified separately using 2.5 mM MgCl2, 0.2 mM of each dNTP, 0.2 µM of each primer and 0.5 U Taq polymerase and the following reaction profile: 40 cycles of denaturation at 95°C for 30 s, hybridization at 50°C for 30 s, and elongation at 72°C for 30 s, terminated by an elongation step of 10 min at 72°C. The PCR product was then mixed with multiplex 2A in a 1∶1 ratio. Finally, the internal size standard LIZ500 was added to all samples prior to loading on an automated sequencer. This final step was performed by a private genotyping company (Genoscreen, Lille, France). Microsatellite profiles were manually genotyped using GeneScan 3.7 and Genotyper 3.7. Reproducibility was checked by performing the amplification and genotyping steps on 30 samples twice, leading to 30×9 = 270 sample × locus duplicates. Statistical analyses {#s2c} -------------------- For both nuclear and chloroplast loci, samples from Japan and Korea were discarded from population-level computations because they were not grouped into discrete natural populations and not in sufficient number to allow reliable statistical inferences. ### Nuclear microsatellites {#s2c1} Within each population, linkage disequilibrium was tested between loci based on random permutations of genotypes performed with the software FSTAT [@pone.0017658-Goudet1] and followed by a Bonferroni correction for multiple tests. Genetic diversity was estimated as allelic richness (mean number of alleles per locus based on the minimal sample size [@pone.0017658-ElMousadik1]) and expected heterozygosity using FSTAT. Genetic structure was quantified by within-population *F* ~IS~ and among-population *F* ~ST~ indices using FSTAT. The statistical significance of *F* ~IS~ was assessed by 5760 random permutations of alleles in each population at each locus, followed by a Bonferroni correction for multiple tests. To detect signs of recent bottlenecks, we examined deviations in heterozygosity from mutation--drift equilibrium in each population with the software Bottleneck [@pone.0017658-Cornuet1]. The loss of rare alleles in recently bottlenecked populations leads to an excess of heterozygosity relative to the expected heterozygosity with the same number of alleles at mutation--drift equilibrium [@pone.0017658-Cornuet1]. We assumed that microsatellite loci follow a two-phase mutation model (intermediate between the IAM and SMM models) with 70% single-step mutations and 30% multiple-step mutations. We used the implemented Wilcoxon test, which is considered the most powerful and robust among the tests proposed in Bottleneck, and we corrected the results by a Bonferroni procedure. Among-population differentiation was quantified with *F* ~ST~ indices both at the global scale and among all pairs of populations. We computed the 95% confidence interval of the global *F* ~ST~ by bootstrapping over loci. The overall differentiation of each population was estimated as the mean pairwise *F* ~ST~ between each population and all others. Exact tests of population differentiation were also performed among all pairs of populations using Genepop [@pone.0017658-Raymond1], [@pone.0017658-Rousset1]. Pairwise differences in expected heterozygosity, allelic richness, *F* ~IS~ and *F* ~ST~ among North America, Europe (including Ukraine and Russia) and non-European invasive populations (Argentina, Beijing, Wuhan, Austral1 and Austral2) were assessed using permutation tests in FSTAT (for *F* ~ST~ indices, we only compared North America and Europe because they cover similar geographic areas; non-European invasive populations were much more distant from each other, which would induce a bias). To identify the potential sources of invasive populations, we attempted to assign all sampled individuals from invasive populations to their most probable source population among the sampled North American populations. We adopted the method of Rannala & Mountain [@pone.0017658-Rannala1], which uses Bayesian criteria for likelihood estimation. The probabilities of assignment were calculated following Paetkau et al. [@pone.0017658-Paetkau1] based on 10,000 simulated individuals. These calculations were performed using the GeneClass 2.0.h software [@pone.0017658-Piry1]. Based on the matrix of pairwise *F* ~ST~ indices, the genetic similarity of populations was summarized using a Principal Coordinate Analysis, performed in NTSYS ([@pone.0017658-Rohlf1]; the analysis included double-centring the matrix and computing eigen-vectors using the Dcenter and Eigen modules, respectively). A hierarchical analysis of molecular variance (AMOVA) was conducted to partition the total genetic variance in among-region, among-population within region, and among-individual within population components using Arlequin [@pone.0017658-Excoffier1]. For this analysis, we considered two regions: North America and Europe. We tested the pattern of isolation by distance within these two regions by performing Mantel tests with 10000 random permutations to compare the genetic and geographic distance matrices. We used several Bayesian algorithms implemented in Structure [@pone.0017658-Pritchard1], [@pone.0017658-Falush1], Instruct [@pone.0017658-Gao1] and Structurama [@pone.0017658-Huelsenbeck1], to cluster individuals into genetically distinct groups. - Structure uses Markov chain Monte Carlo (MCMC) algorithms to group individuals in clusters (where the numbers of clusters must be set a priori) that deviate neither from Hardy--Weinberg nor linkage equilibrium within each cluster. It also calculates the posterior probability of the data given the inferred clustering. Structure was run 20 times for each K-value from one to seven to check the consistency of the results across runs. Each run comprised a burn-in period of 200000 iterations followed by 10^6^ iterations. We adopted the admixture model, the correlated allele frequencies model, and we used sampling locations as prior information to assist the clustering (LOCPRIOR option). Hubisz et al. [@pone.0017658-Hubisz1] showed that this option improves the performance of the clustering when the signal of structure is weak, but does not tend to find structure when none is present. We plotted the relationship between the K-value and (i) the probability of the data lnP(D) and (ii) as recommended by Evanno et al. [@pone.0017658-Evanno1], the ad hoc statistic ΔK which corresponds to the change of lnP(D) between consecutive K-values. We identified the most relevant number of clusters (K) as the one that maximized lnP(D) and/or ΔK, following Evanno et al. [@pone.0017658-Evanno1]. For each K-value, the similarity among runs (in terms of individual assignment to the K clusters) was estimated with Structure-sum-2009 [@pone.0017658-Ehrich1] and the most likely inferred clustering was graphically displayed with Distruct [@pone.0017658-Rosenberg1]. Structure was also run in a similar fashion within North America and within Europe, to compare how the genetic diversity was geographically structured in the two ranges. - Unlike Structure, which requires running the program several times under different K-values and then determining the best value *post-hoc*, Structurama employs a prior distribution of K to determine the most appropriate K-value. At each run, it also outputs posterior probabilities of each possible K-value and the mean partition, i.e. a partitioning of individuals among clusters that minimizes the squared distance to the sampled partitions across generations of the MCMC [@pone.0017658-Huelsenbeck1]. The program was run three times for each of three prior models (i.e. nine runs in total). The number of clusters and the alpha parameter were considered random variables, with the alpha parameter following a gamma probability distribution. The shape and scale (a, b) of this distribution were consecutively set to (1, 2), (2, 2) and (3, 2), respectively, corresponding to prior K-values of 4.1±2.8, 6.9±5.2 and 9.4±7.3, respectively. Each run comprised 20000 generations that were discarded as burn-in and 180000 generations that were sampled every 50 generations. - The Bayesian approach of Instruct is very similar to that of Structure, but Instruct allows inbreeding and estimates inbreeding coefficients (that are similar to within-population *F* ~IS~ indices) within the inferred clusters [@pone.0017658-Gao1]. The approach of Instruct may be biologically more suited to *A. artemisiifolia* since we detected significant departures from Hardy-Weinberg equilibrium in almost all populations (see [Results](#s3){ref-type="sec"}). We conducted five runs per K-value spanning from one to 10, with each run comprising 100000 iterations burn-in followed by 500000 iterations that were sampled every 50 generations (thinning). ### Chloroplast microsatellites {#s2c2} Because there is no recombination within the cpDNA molecule, alleles found at all cpSSR loci were combined to compose a unique chloroplast haplotype for each individual. Individuals with missing data (n = 24) were discarded from the inference of multilocus haplotypes and from the statistics based onto these haplotypes. First, we considered all multilocus haplotypes to draw a median-joining network based on the number of mutations among all pairs of haplotypes using the software NETWORK [@pone.0017658-Bandelt1]. We adopted a two-step procedure to reduce the potential impact of homoplasy: based on an initial network, the loci were inversely weighted by the number of mutations occurring at each of them, in a second run, as recommended by Bandelt et al. [@pone.0017658-Bandelt1], [@pone.0017658-Bandelt2]. Within populations, we computed the number of haplotypes, number of private haplotypes (found in only one population) and mean number of pairwise differences among individuals using the software Arlequin. For this, we coded cpSSR data in a binary way, representing for each locus the number of repeats of the largest variant with '1's and replacing the absent repeats of shorter variants with '0's. Permutation tests (in FSTAT) and non parametric Mann-Whitney tests were performed to detect any significant difference in number of haplotypes and mean number of pairwise differences among individuals, respectively, between North America, Europe and non-European invasive populations The program SpaGeDi [@pone.0017658-Hardy1] was used to compute global and pairwise *F* ~ST~ and *N* ~ST~ indices of among-population differentiation based on unordered and ordered haplotypes, respectively. The input dataset contained, for each individual, the multilocus haplotype displayed. For the estimation of *N* ~ST~ *\'*s, the distance between haplotypes was calculated as the sum of their absolute length differences across the four loci. We performed 10000 permutations of rows and columns of the distance matrix between haplotypes to test whether *N* ~ST~ \> *F* ~ST~. Such a significant relationship suggests that distinct haplotypes are more related within populations than among them, i.e. that genetic structure displays a significant geographic trend [@pone.0017658-Pons1]. We conducted a Principal Coordinate Analysis (using NTSYS) based on *N* ~ST~ indices, and an AMOVA of haplotype frequencies, implemented in Arlequin, to assess the proportion of genetic variance found at the region (North America vs. Europe), population and individual levels. We also performed Mantel tests (using FSTAT) within North America and Europe, based on *N* ~ST~ indices~.~ Results {#s3} ======= Reproducibility was high, with 97.7% of all sample × locus duplicates carrying the same genotype. We did not find any evidence of linkage disequilibrium between pairs of nuclear microsatellite markers. Genetic diversity {#s3a} ----------------- We detected a mean (± S. D. among loci) of 19.7 ± 8.2 alleles per nuclear microsatellite locus (spanning from 6 to 29 alleles per locus; [Table 1](#pone.0017658.t001){ref-type="table"}) and a mean (± S. D. among loci) of 4.8±1.5 alleles per chloroplast microsatellite locus (spanning from 3 to 7 alleles per locus; [Table 1](#pone.0017658.t001){ref-type="table"}). At nuclear loci, expected heterozygosity was quite similar across populations whereas mean allelic richness was more variable ([Table 1](#pone.0017658.t001){ref-type="table"}). At the regional level, mean allelic richness (± S. D. among loci) was 15.19±6.19, 14.92±6.42 and 13.35±5.85 alleles per locus in North America, Europe and non-European invasive populations, respectively, based on a minimum sample size of 53 individuals. All invasive populations (except B, PO1 and Wuhan) displayed at least one allele that was absent from North American populations. These alleles were most often found in several populations, but usually at very low frequency (\<0.1). Only IT9, UKR, and Austral2 exhibited alleles (one allele each) that were absent from America and present at frequencies higher than 0.2 within populations. Only two populations, PO2 and Beijing, showed significant excess of heterozygosity (after the Bonferroni correction, P = 0.031 for both populations), which suggests a recent bottleneck. CpDNA microsatellites allowed the definition of 33 multilocus chloroplast haplotypes. ([Table S1](#pone.0017658.s002){ref-type="supplementary-material"}). Fourteen haplotypes were private to one population, but only four of them were found in at least two individuals: haplotype K was observed in the population from Utah (in 11 individuals), haplotype D in IT1 (in 11 individuals) and haplotypes L and W were only observed in the population from Argentina (in 9 and 5 individuals, respectively; [Table S1](#pone.0017658.s002){ref-type="supplementary-material"}). In total, we observed 19 haplotypes in North America, 23 haplotypes in Europe (15 of which were shared with North America) and 14 haplotypes in the non-European invasive populations (8 of which were shared with North America and Europe; [Table S1](#pone.0017658.s002){ref-type="supplementary-material"}). Using the rarefaction method of El Mousadik & Petit [@pone.0017658-ElMousadik1] to account for different sample sizes across regions, estimates of haplotype richness were 18.6, 20.3 and 14.0 haplotypes for North America, Europe and non-European invasive populations, respectively, based on a minimum sample size of 101 individuals. Permutation tests showed that North American and European populations did not differ statistically in terms of diversity (allelic richness, expected heterozygosity, number of haplotypes per population, mean number of pairwise differences among individuals; [Table 1](#pone.0017658.t001){ref-type="table"}; all P\>0.1). In contrast, the group of non-European invasive populations (Argentina, Wuhan, Beijing, Austral1, Austral2) was less diverse than North America and Europe in terms of allelic richness ([Table 1](#pone.0017658.t001){ref-type="table"}; P = 0.004 and 0.044, respectively). The group of non-European invasive populations was not significantly different from North America but marginally less diverse than Europe for the number of haplotypes ([Table 1](#pone.0017658.t001){ref-type="table"}; P = 0.147 and P = 0.051, respectively). North American, European and non-European invasive populations were not significantly different in terms of expected heterozygosity and mean number of pairwise differences ([Table 1](#pone.0017658.t001){ref-type="table"}; all P\>0.1). Most populations that were characterized by low estimates of nuclear allelic richness also displayed few chloroplast haplotypes and/or low mean number of pairwise differences among individuals ([Fig. 2](#pone.0017658.f002){ref-type="fig"}; the correlation was significant, P = 0. 041) e.g. UKR, Austral2, IT4, and Bronx. ::: {#pone.0017658.f002 .fig} Figure 2 ::: {.caption} ###### Figure 2. Within-population genetic diversity. Relationship between nDNA allelic richness and cpDNA mean pairwise number of differences between individuals (the number of haplotypes in the population is indicated within brackets). The correlation was significant (P = 0.041). :::  ::: Genetic Structure - Bayesian Clustering (nuclear Microsatellites Only) {#s3b} ---------------------------------------------------------------------- Out of the nine runs of Structurama, seven indicated K = 6 as the most relevant number of clusters (i.e. within each run, K = 6 was associated to the highest probability, averaging P = 0.44±0.02 across runs) and two indicated K = 7 (with P = 0.40±0.01). However, six of the nine mean partitions that were inferred counted five clusters. In the three remaining runs, the sixth cluster was a subdivision of a pre-existing cluster and was present in very low proportions in several populations. Using Structure and Instruct, the probability of the data lnP(D) steadily increased, and the change of probability between consecutive K-values (ΔK) steadily decreased when assuming increasing K values ([Fig. 3A](#pone.0017658.f003){ref-type="fig"}). However, we observed that from K = 6 upwards, additional clusters did not individualize additional populations but were rather represented in moderate proportions in many populations, therefore probably not revealing a genuine population genetic structure. Therefore, we did not run Structure assuming higher K-values. ::: {#pone.0017658.f003 .fig} Figure 3 ::: {.caption} ###### Figure 3. Bayesian analysis performed in Structure. A\) on the overall nDNA dataset, relationship between K, lnP(D) and ΔK. B) On the overall nDNA data set, cluster partitioning of the populations at consecutive K-values from K = 2 to K = 5. Each vertical line represents one individual and the colors represent the membership coefficients to the K clusters. The clustering solutions inferred by Instruct and Structurama were highly similar. Colors are the same as in [Fig. 1. C](#pone.0017658.g001){ref-type="fig"}) On the North American (K = 3) and European (K = 7) datasets. :::  ::: At each K-value, most runs of Structure were consistent in terms of individual assignment to the K clusters (similarity ≥ 0.75). At K = 2, an East-West cline of cluster assignment was observed in North America, and the predominant cluster in western North America (in green; [Fig. 3B](#pone.0017658.f003){ref-type="fig"}) was also found in Eastern Europe, Beijing and the two Australian populations. At K = 3, the third cluster (in blue) was mostly represented in Australia, Italy (IT9) and Argentina while the second cluster (in green) was mostly observed in western North America and Eastern Europe. At K = 4, the Australian populations (in yellow) were again separated and, at K = 5, populations from Romania and Beijing were grouped together into a new cluster (orange). As before, we also observed some genetic similarity between western North America and Eastern Europe (Ukraine and Russia) on the one hand, and eastern North America and Western and Central Europe (IT9 to PO1) on the other hand. Instruct allowed estimation of inbreeding levels spanning from 0.28 to 0.40 within each of the five inferred clusters. Within North America, the highest lnP(D) and ΔK values were obtained at K = 3. Utah and Montana were individualized in one cluster each, while Missouri, South Carolina, Ontario, Quebec and Bronx were predominantly assigned to the third cluster. Minnesota was intermediate, with approximately equal contributions of the three clusters ([Fig. 3C](#pone.0017658.f003){ref-type="fig"}). At K = 4 and K = 5, South Carolina and Bronx clustered separately. Within Europe, lnP(D) increased with increasing K values and ΔK displayed two peaks, for K = 2 and K = 7. At K = 7, populations IT9, IT4, SE3, RO1 and UKR segregated in specific clusters, indicating their strong genetic divergence ([Fig. 3C](#pone.0017658.f003){ref-type="fig"}). Some other populations were predominantly assigned to the same clusters: B and D1 on the one hand, and F, IT1 and PO1 and the other hand. Finally, some populations in Central (SLO1, HU2, HU6, HU3) and Eastern Europe (PO2, UKR1, UKR2, RU4 and RU5) appeared highly admixed. Genetic Structure -- F-Statistics {#s3c} --------------------------------- Multilocus fixation indices *F* ~IS~ were significantly positive in all populations except Bronx and IT4 ([Table 1](#pone.0017658.t001){ref-type="table"}). Nineteen populations displayed at least two significant monolocus tests, with up to four significant tests in Montana, Minnesota, Wuhan and Austral1. *F* ~IS~ estimates were not significantly different between North America and Europe (P = 0.839), but were (marginally) significantly higher in non-European invasive populations than in North America and Europe ([Table 1](#pone.0017658.t001){ref-type="table"}; P = 0.085 and P = 0.037, respectively). At nuclear loci, the overall *F* ~STn~ estimate was 0.073 (95% C.I.: 0.065--0.083). Analyses of Molecular Variance (AMOVAs) showed that the split between North America and Europe did not explain a significant proportion of the observed genetic variance either at nuclear or at chloroplast markers. Furthermore, though not significantly so, European populations showed somewhat greater among-population differentiation than did North American populations (*F* ~STn~ = 0.065±0.006 and 0.054±0.012, respectively). At chloroplast loci, *F* ~STcp~ = 0.411 and *N* ~ST~ = 0.440 (not significantly different). In North America, *F* ~STcp~ = 0.373 and was significantly lower than *N* ~ST~ = 0.518 (P = 0.007). In contrast, in Europe, *F* ~STcp~ = 0.389 and was not significantly different from *N* ~ST~ = 0.384. These results showed a significant influence of the spatial component on the genetic structure in North America, but not in Europe. Most exact tests of population differentiation were significant but some populations, in Central (SLO1, HU2, HU6, HU3) and Eastern Europe (PO2, UKR1, UKR2, RU4 and RU5), appeared clearly less differentiated from North American populations than the others ([Table S2](#pone.0017658.s003){ref-type="supplementary-material"}). We computed mean pairwise *F* ~ST~ values among these two groups of low-differentiated European populations, to which we added clearly differentiated and geographically concomitant populations, and western and eastern North American populations, respectively. Standard deviations were large and differences were therefore not significant, but we observed the same pattern as in Structure: Central European populations were closer to eastern than to western North American populations while Eastern European populations were slightly closer to western North American than to eastern North American populations ([Fig. 4](#pone.0017658.f004){ref-type="fig"}). Populations B, F, D1 and PO2 were not included in these calculations because we observed incongruent results between Structure and *F* ~ST~ estimates, and RO1 was also excluded because of its strong divergence. ::: {#pone.0017658.f004 .fig} Figure 4 ::: {.caption} ###### Figure 4. Mean pairwise *F* ~ST~ indices (± S. D.) estimated at nDNA loci between populations from different regions. Regions include Central Europe (IT9 to PO1; 9 pops.), Eastern Europe (UKR to RU5; 5 pops.), western North America (Utah to Minnesota; 3 pops.) and eastern North America (Missouri to Bronx; 5 pops.). :::  ::: As for nDNA markers, there was no obvious geographical structure of the cpDNA genetic diversity at first sight. However, the close relationships of Utah with Eastern Europe, suggested by Bayesian clustering and *F* ~ST~ indices based on nuclear DNA, was confirmed: Utah displays three cpDNA haplotypes, one of which was private and the other two that were otherwise mostly found in RU4 (haplotype H) and UKR (haplotype F; [Fig. 5](#pone.0017658.f005){ref-type="fig"}, [Table S1](#pone.0017658.s002){ref-type="supplementary-material"}). This was less clear for the Montana and Minnesota populations. Nevertheless, populations from western North America and Eastern Europe were predominantly represented in the right part of the network (which includes 53.3 and 54.6% of the samples of these two regions, respectively; [Fig. 5](#pone.0017658.f005){ref-type="fig"}), while populations from eastern North America and Western Europe were predominantly represented in the left part of the network (which includes 71.4 and 51.5% of the samples of these two regions \[57.9% if excluding B, F, D1, PO2, RO1\], respectively; [Fig. 5](#pone.0017658.f005){ref-type="fig"}). ::: {#pone.0017658.f005 .fig} Figure 5 ::: {.caption} ###### Figure 5. Median-joining network of cpDNA haplotypes. Ten haplotypes counting only one individual each ([Table S1](#pone.0017658.s002){ref-type="supplementary-material"}) were discarded, so that the network includes 23 haplotypes and 600 individuals. The size of each pie is proportional to the frequency of the corresponding haplotype. The colors indicate the geographical origin of the populations displaying each haplotype. Light green: western North America; light blue: eastern North America; dark green: Eastern Europe; dark blue: Western Europe. Purple: Argentina; White: Asia (China, Japan, Korea); Yellow: Australia. Black dots stand for unsampled haplotypes and each segment joining haplotypes represent one mutation. The two ellipses indicate the two areas of the network discussed in the text. :::  ::: The nuclear- and chloroplast-based differentiation indices (*F* ~STn~ and *N* ~ST~, respectively) were significantly correlated, both when considering all pairs of populations (496 values, P\<0.001) and when considering the mean pairwise differentiation indices for each population (32 values, P = 0.002; [Fig. 6](#pone.0017658.f006){ref-type="fig"}). ::: {#pone.0017658.f006 .fig} Figure 6 ::: {.caption} ###### Figure 6. Relationship between mean pairwise differentiation indices at nDNA (*F* ~ST~) and at cpDNA (*N* ~ST~) loci. The correlation was significant (P = 0.002). :::  ::: Within invasive populations, individuals were assigned to at least three (and up to seven) different source populations in North America ([Table S3](#pone.0017658.s004){ref-type="supplementary-material"}). The probabilities of assignment were less than 0.5 in 71% of the cases but, when only considering individuals with assignment probabilities above 0.5, samples from Western and Central Europe (populations IT9 to PO1, n = 47) were mostly assigned to populations from eastern North America (S. Car to Bronx; 68%) whereas individuals from Eastern Europe (UKR to RU4, n = 36) were mostly assigned to populations from western North-America (Utah, Montana and Minnesota; 58%). Principal Coordinate Analyses allowed us to graphically represent the main patterns of genetic relationships, which were congruent with the results of the Bayesian clustering, haplotype network and differentiation indices: for both nuclear and chloroplast markers, the divergence of Austral1-Austral2 and Utah-UKR appeared clearly ([Fig. 7A and 7B](#pone.0017658.f007){ref-type="fig"}). Beijing and RO1 also appeared much differentiated for nuclear (but not for chloroplast) loci whereas Quebec and Bronx appeared more divergent at chloroplast than at nuclear loci ([Fig. 7A and 7B](#pone.0017658.f007){ref-type="fig"}). ::: {#pone.0017658.f007 .fig} Figure 7 ::: {.caption} ###### Figure 7. Principal Coordinate Analyses. A\) At nDNA loci. B) At cpDNA loci. Dots: North America; open squares: Europe; crosses: non-European invasive populations. The percentages of variance explained by each axis are indicated within brackets. :::  ::: Mantel tests revealed significant isolation by distance patterns in North America for both nuclear and chloroplast loci (P = 0.002 and P = 0.049, respectively; [Fig. S1A](#pone.0017658.s001){ref-type="supplementary-material"}) but not in Europe (P = 0.581 and P = 0.094, respectively; [Fig. S1B](#pone.0017658.s001){ref-type="supplementary-material"}). Discussion {#s4} ========== Our results showed that most invasive populations were as diverse as the native populations. In Europe, Western and Central European populations were genetically more related to eastern North America, while Eastern European populations were closer to western North America. There was also a stark contrast between genetic structure in the native range, which displayed a clear geographic cline from East to West, and in Europe, where we detected no pattern of isolation by distance and only a weak influence of geography on the genetic structure. Our North American Sampling Does Not Encompass All Sources Of The Worldwide Invasion {#s4a} ------------------------------------------------------------------------------------ The observation of some private alleles/haplotypes in invasive populations, the fact that some clusters inferred in Structure were virtually not represented in North America, and the low assignment probabilities of most invasive individuals to the North American sampled populations altogether suggested that our North American sampling does not encompass all sources of the worldwide (or even European) *A. artemisiifolia* invasion. An alternative explanation could be that some alleles/haplotypes were rare in the native area and that their frequency increased during introduction and subsequent invasion, but this scenario seems less parsimonious than the existence of unsampled source populations. A third possible explanation involves the *in situ* emergence of novel alleles/haplotypes following introduction, but this hypothesis appears even more unlikely: the worldwide expansion of common ragweed started in the mid- or late-XIXth century, i.e. at most 120--150 generations ago (since the plant is annual). The time scale of this study is thus much more restricted than in traditional phylogeographic studies, and the evolution of new alleles appears very improbable given the mutation rates at nuclear and chloroplast microsatellite loci (of the order of 10^−4^ and 10^−5^--10^−6^ mutation per locus per generation, respectively [@pone.0017658-Jakobsson1],[@pone.0017658-Thuillet1]). This is even less probable when the private alleles/haplotypes diverge by more than one mutation from other alleles/haplotypes, since this would involve multiple mutation events. Multiple Introductions In Europe, Originating From (at Least) Two Distinct Regions In North America {#s4b} --------------------------------------------------------------------------------------------------- We observed no significant loss of genetic diversity between North America and Europe, and European populations did not appear to have undergone recent bottlenecks (except population PO2). Furthermore, European populations were genetically differentiated (*F* ~ST~ = 0.065 and most exact tests of differentiation were significant). Because introductions almost always involve sampling and founder events, we consider it improbable that populations in the introduced range could have arisen from single population introductions and still retain this high amount of genetic diversity. This leaves us with two alternatives: i) populations were founded by multiple colonisations from different populations in the native range, as suggested by the assignment test and previous findings of extremely high allelic diversity in the introduced range [@pone.0017658-Genton1] or ii) introduced populations arose from independent introductions from single source populations and subsequent gene flow has restored diversity to similar levels as that found in the native range. We note, however, that the genetic differentiation of European populations suggests low gene flow. Furthermore, for restoring genetic diversity, gene flow must have involved (human-mediated) long distance dispersal and not natural processes of pollen and seed exchange between neighbouring populations, since this would have left a trace of isolation by distance, for which we found no evidence among European populations. Therefore, we favour the scenario involving multiple introductions at both the population and the regional scales. This study thus adds to the pre-existing body of evidence that multiple introductions seem to be a common feature of biological invasions [@pone.0017658-Dlugosh1],[@pone.0017658-Bossdorf1],[@pone.0017658-Wilson1]. Overall, Western and Central European populations seemed more related to eastern North American populations whereas Eastern European (Ukrainian and Russian) populations were genetically more similar to western North American populations. This clarifies the pattern observed by Gladieux et al. [@pone.0017658-Gladieux1], who found that eastern North American and French populations were clearly differentiated from Eastern European populations. However, they could not interpret this result further since they did not include populations from Western North America. The same kind of geographic pattern was observed in the invasive grass *Bromus tectorum* [@pone.0017658-Valliant1]: the authors detected some genetic similarity between eastern Canada and Germany-Czech Republic on the one hand, and between western Canada and Hungary-Slovakia on the other. These results together strengthen the idea that differences in commercial exchange between different regions of North America and Europe have influenced sources of invasive populations. We also observed that the two groups of populations in Slovenia-Hungary and Ukraine-Russia were less differentiated, more diverse and more admixed than most other European populations. This may indicate more frequent colonization events than in other regions, and/or higher ongoing gene flow among populations. Interestingly, these populations are located in some of the most heavily infested countries in Europe, i.e. Hungary and Russia. The French population was also sampled in a region where common ragweed is a very aggressive invader, and was also found to have low differentiation and high diversity. Whether higher genetic diversity increases invasion success or whether areas where an invasion is particularly successful (containing high number of populations) leads to high genetic diversity remains uncertain, but would deserve further investigations: genetic diversity has long been considered a prerequisite for invasion success because of the assumed correlation between variation at neutral markers and adaptive potential (e.g. [@pone.0017658-Kolbe1],[@pone.0017658-Lavergne1]), but this now appears controversial (see e.g. [@pone.0017658-Zayed1]). Fewer Introduction Events In Non-European Invasive Populations {#s4c} -------------------------------------------------------------- In contrast to Europe, other invasive populations displayed reduced genetic diversity and a trend towards increased within-population *F* ~IS~ indices compared to the native area. This suggests that introduction events may have been less frequent and involved a lower number of differentiated source populations and/or individuals (i.e. lower propagule pressure), possibly leading to lower population sizes and more genetic drift (but only the Beijing population seems to have undergone a recent bottleneck). In turn, this could cause increased rates of selfing and/or inbreeding (i.e. crosses between related plants), explaining the slightly higher *F* ~IS~ estimates (see below for a discussion on *F* ~IS~ estimates). Such a pattern would be consistent with greater commercial isolation from North America and less military exchange with North America than was the case for Europe during the two World Wars, and/or better quarantine procedure (e.g. in Australia, where the species is quite restricted). The two Australian populations were genetically similar and strongly differentiated from all other populations. They most likely originated from an unsampled source, either through a single introduction event followed by dispersal within Australia (probably from Austral1 to Austral2 since Austral1 exhibits a slightly higher genetic diversity), or through two independent introduction events from the same source population(s). A similar pattern was observed for the Beijing and Romanian populations, which were closely related but highly differentiated from all others. The genetic similarity between Romania and Beijing populations may be explained either by commercial trade with the same (unsampled) region(s) in North America, or by a secondary introduction from Romania to China (since we observed lower diversity in Beijing). The other non-European invasive populations, Wuhan and Argentina, were less differentiated from North America and Europe. These populations were also slightly more diverse than Australian and Beijing populations, indicating that they probably experienced more introduction events or less strong founder effects. Genetic Differentiation Is Influenced By Geography In North America, But Not In Europe {#s4d} -------------------------------------------------------------------------------------- We observed similar levels of population differentiation in North America and Europe (*F* ~STn~ = 0.054 vs. 0.065, respectively) and the slightly stronger genetic structure in Europe may result from i) the fact that potentially divergent North American populations were missing from our sampling and ii) the establishment, by chance, of different genotypes in different areas following multiple introductions (as shown in *Centaurea diffusa* [@pone.0017658-Marrs1]). Most importantly, we observed a major difference between North America and Europe in how genetic diversity was structured geographically, which gave an insight into the colonization process in Europe: i) we found significant isolation by distance (i.e. a positive correlation between genetic and geographic distances) in North America but not in Europe; ii) although we could identify two groups of European populations originating from distinct source regions, the pattern was not clear-cut and did not include all sampled populations, and the overall genetic structure was much more geographically organised in North America than in Europe (based on the Structure results); and iii) distinct haplotypes were significantly more related within populations than among populations in North America (*N* ~STcp~ \> *F* ~STcp~) but not in Europe. All these results were in agreement and, first, show that North American populations are at migration-drift equilibrium whereas European populations are not. Second, they indicate that range expansion in Europe occurred by a series of long-distance dispersal events and the establishment of outlying populations, similarly to what was found in e.g. invasive *Heracleum* taxa in Europe [@pone.0017658-Jahodova1] and *Centaurea diffusa* in North America [@pone.0017658-Marrs1], instead of a simple advancing wave front with stepwise colonisation events. Long distance dispersal events were probably human-mediated, and may have involved both transatlantic and within-Europe dispersal. Whereas the installation of new populations obviously required seed dispersal and establishment, most subsequent gene dispersal seems mediated through pollen, as indicated by the much stronger among-population differentiation at cpDNA (only dispersed by seeds) than at nDNA (dispersed by both seeds and pollen) markers. This is congruent with previous knowledge on pollen dispersal in *A. artemisiifolia*, which can reach hundreds of kilometres (although the duration of pollen viability is unknown [@pone.0017658-Friedmann1],[@pone.0017658-Martin1]). Genetic Structure At The Worldwide Scale {#s4e} ---------------------------------------- Genetic differentiation at the worldwide scale was low. This may be explained by weak founder effects when the species was introduced, on-going gene flow, and/or insufficient time for genetic drift to differentiate the populations since their establishment. In addition, there was little spatial component to the genetic structure: the geographic split between North America and Europe explained no significant part of the total genetic variance. Also, the combined use of several Bayesian algorithms allowed the delineation of five genetically-based clusters, but these clusters could not be related to clear geographic entities. Within-Population Genetic Structure And Mating System {#s4f} ----------------------------------------------------- Almost all within-population fixation indices *F* ~IS~ were significant, suggesting a deficit in heterozygotes. This was confirmed by the software Instruct, which estimated inbreeding coefficients of 0.3--0.4 within clusters. Although this result was congruent with previous population genetic studies of *A. artemisiifolia* [@pone.0017658-Chun1],[@pone.0017658-Genton1],[@pone.0017658-Gladieux1], it was surprising because the species has been shown to be outcrossing and self-incompatible [@pone.0017658-Friedmann1]. We hypothesise that selfing and/or biparental inbreeding, as well as a spatial substructuring within populations (i.e. Wahlund effect) may be involved. In earlier genetic surveys [@pone.0017658-Chun1],[@pone.0017658-Genton1],[@pone.0017658-Gladieux1], null alleles were proposed as the most likely explanation for positive *F* ~IS~ estimates. Although we do not exclude this possibility, we do not favor it for several reasons. First, we did not observe any repeated amplification failure for any given locus in any population (which is expected with null alleles since homozygotes for a null allele will produce no PCR amplification). Therefore, even if there are some null alleles, they occur at very low frequencies and contribute very little to overall heterozygosity deficit (and therefore to *F* ~IS~ calculations). Second, in our study, most populations displayed significant deficit of heterozygotes at several loci and significant monolocus *F* ~IS~ estimates were widely distributed across loci. Third, Genton et al. [@pone.0017658-Genton1] and Gladieux et al. [@pone.0017658-Gladieux1] used the same five nuclear microsatellite markers, which totally differ from the nine markers used by Chun et al. [@pone.0017658-Chun1], but both groups of loci lead to positive *F* ~IS~ values (in the present study, we used three markers in common with Genton et al. [@pone.0017658-Genton1] and Gladieux et al. [@pone.0017658-Gladieux1], and five markers in common with Chun et al. [@pone.0017658-Chun1]). This would mean that a high number of loci display null alleles, which does not seem very plausible. Fourth, Gladieux et al. [@pone.0017658-Gladieux1] explained high *F* ~IS~ in Eastern Europe by the fact that microsatellites were developed on French populations and that the genetic divergence of Eastern Europe may explain the occurrence of some mutations at primer sites, leading to null alleles. However, they also documented very high *F* ~IS~ estimates in France (mean *F* ~IS~ of 0.490 ± 0.0469). Altogether, these lines of evidence suggest that null alleles are probably not the main cause for the observed deficits of heterozygotes within populations. All *F* ~IS~ estimates were quite similar and there was no evidence for an evolutionary shift towards higher selfing rates in the introduced range, as has been suggested as a general pattern in invasive species [@pone.0017658-Barrett2]. A Wahlund effect is possible in the sampled populations given the very large size of most of them (sometimes counting more than 10000 plants and covering areas of several thousand square meters). Nevertheless, further studies on the breeding system of *A. artemisiifolia* and its potential variation across populations (or regions) would be interesting to better understand these positive *F* ~IS~ estimates. Such studies appear especially needed since *A. artemisiifolia* was long reported as self-compatible [@pone.0017658-Bassett1],[@pone.0017658-Jones1] and Friedmann & Barrett [@pone.0017658-Friedmann1], who showed the self-incompatibility in Canadian populations, acknowledged the possibility that some other populations may exhibit partial self-compatibility. Moreover, based on controlled pollinations, observations of pollen-tube growth and allozyme analyses in three populations from China, Li et al. [@pone.0017658-Li1] concluded that selfing was possible (although leading to lower seed sets than outcrossing) and estimated an average selfing rate of 0.22. Conclusions {#s4g} ----------- The present study shows how variable the history of distinct (but sometimes geographically close) invasive populations can be. This highlights the importance of sampling as many populations as possible to avoid biased inferences (see also [@pone.0017658-Muirhead1]). It also appears desirable to sample with no major geographic gap, especially in the native range. Gladieux et al. [@pone.0017658-Gladieux1], although with more populations from the native area than in the present survey, had poorer geographic coverage and could only conclude that Eastern European and French populations did not originate from the same source populations. Our geographically larger sampling area allowed us to document this pattern more precisely and propose possible source regions of the Eastern European *A. artemisiifolia* populations. We showed that *A. artemisiifolia* was introduced multiple times in most parts of its invasive range, leading to high levels of within-population and regional diversity. In Europe, introduction events probably mainly involved two different regions of the native area, with populations of Central Europe originating from eastern North America, and populations of Eastern Europe originating from more western North America. Our results indicate that the expansion of the European range mostly occurred through long-distance seed dispersal, explaining the weak association between genetic differentiation and geographic location in this area (in contrast to the native range, where isolation by distance was observed). Finally, heterozygote deficiencies may be explained by a Wahlund effect, but further investigations on the breeding system would provide useful information to better explain this result. Such data offer opportunities to study the ecological and/or evolutionary changes involved in the invasion process (e.g. [@pone.0017658-Lavergne1],[@pone.0017658-Colautti1]), and may help to predict the potential further expansion of the species. *A. artemisiifolia* exhibits latitudinal variation in flowering phenology both in the native range [@pone.0017658-Genton2] and in invasive populations in China, which may indicate on-going local adaptation and allow further expansion northwards of the invasive populations [@pone.0017658-Li1]. The same may be true in Europe, where the species is increasingly often observed flowering in Scandinavia, in spite of the short growing season [@pone.0017658-Skjth1]. The mechanisms underlying such potential, rapid adaptive processes, and their consequences would be worth examining more in depth, and in relation with global warming. Genetic data can also benefit the development of effective prevention and management strategies. More globally, this study adds to the growing body of data on the genetic patterns and processes involved in biological invasions, which will hopefully lead to an increased understanding and better management in order to minimize their negative impacts on biodiversity, economy, and also human health in the case of *A. artemisiifolia*. Supporting Information {#s6} ====================== Figure S1 ::: {.caption} ###### Relationship between geographic and genetic distances between all pairs of populations based on nDNA loci. **A**) In North America (P  =  0.002). B) In Europe (P  =  0.581). The geographic distance was expressed as the log~10~ of interpopulation distance in km; the genetic distance was expressed as *F* ~ST~/(1-*F* ~ST~). (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S1 ::: {.caption} ###### CpDNA haplotypic composition of the studied populations. Populations are grouped into three spatial groups (North America, Europe and non-European invasive populations) and roughly ordered from West to East. Haplotypes are ordered from the most frequent to the least frequent. Haplotypes that are in bold are private to one population. (XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S2 ::: {.caption} ###### Pairwise *F* ~ST~ indices among all pairs of populations, estimated at nDNA loci. Populations are grouped into three spatial groups (North America, Europe and non-European invasive populations) and roughly ordered from West to East. *F* ~ST~ estimates that are highlighted in grey correspond to non-significant exact tests of differentiation. (XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S3 ::: {.caption} ###### Results of the assignment test of invasive populations to North American populations, based on nDNA loci. The number of individuals assigned to each North American population is given, either considering all individuals or only individuals with assignment probabilities above 0.5. (XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: We wish to thank all the people who collected samples: L. Bassignot, J. Brodeur, M. Ellis, P. Fidalgo, J. Guzik, V. Hayova, T. Jankovics, A. Kiss, P. Lesica, G. May, R. McFadyen, J. B. Nelson, S. Renner, S. Reznik, R. E. Ricklefs, M. Ronikier, G. Rouhan, R. Sforza, R. Shaw, F. van Rossum, F. Verloove, F.-H. Wan and M. Wrzesień. We also thank the Mie University Mycological Herbarium (Tsu, Japan) and Mycological Herbarium of the Korea University (Seoul, South Korea) for allowing the removal of leaf material from herbarium specimens. We are grateful to F. Bretagnolle for sharing microsatellite primers before publication, to G. Rouhan for his help throughout the project, and to three anonymous reviewers for their comments on a previous version of the manuscript. Molecular work was performed at the BoEM lab of the Muséum National d\'Histoire Naturelle, Paris, France. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was funded by the Muséum National d\'Histoire Naturelle (Paris, France), the Université Paris-Sud (Orsay, France) and the Plant Protection Institute of the Hungarian Academy of Sciences (Budapest, Hungary). TG acknowledges the grant ANR 07-BDIV-003. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: MG. Performed the experiments: MG. Analyzed the data: MG. Contributed reagents/materials/analysis tools: MG TG LK JAS. Wrote the paper: MG. Commented on the paper: TG LK JAS.

PubMed Central

2024-06-05T04:04:19.820278

2011-3-10

PMC3053377

Introduction {#s1} ============ Identifying phenotypes of proteins is a central challenge of the modern genetics in post-genome era. The study on phenotypes always involves many major diseases, such as HIV [@pone.0017668-VanHoutte1], [@pone.0017668-Vasilev1], [@pone.0017668-Xu1], [@pone.0017668-Vermeiren1], [@pone.0017668-Foulkes1], different kinds of cancers [@pone.0017668-Lin1], [@pone.0017668-Bathen1], [@pone.0017668-Lakhani1], [@pone.0017668-Dwyer1], chronic liver diseases [@pone.0017668-Piruzyan1], Gaucher disease [@pone.0017668-Whitfield1]. The high-throughput phenotype assays [@pone.0017668-Drees1], [@pone.0017668-Dudley1] combining with gene perturbation technology [@pone.0017668-Fire1], [@pone.0017668-Winzeler1] provide fast identification for gene active in a response [@pone.0017668-Carter1]. For example, yeast mutant strain collections has become increasingly used to identify the phenotypes [@pone.0017668-Scherens1]. However, these assays are often trapped in the high false negative rates [@pone.0017668-McGary1]. On the other hand, the study on phenotypes is highly complex for the multifactorial trait often results from contributions of many proteins. Consequently, using experimental approaches alone is insufficient, and the computational methods should be applied for the identification of protein phenotypes [@pone.0017668-McGary1]. In principle, there are two kinds of computational methods: the sequence-based methods and network-based methods. A sequence-based method is often designed on a benchmark dataset, sequence features such as amino acid composition [@pone.0017668-Cedano1], pseudo amino acid [@pone.0017668-Chou1] (PseAAC), are used to represent the data (e.g. protein sequence), then a prediction model can be built according to the machine learning algorithm (e.g. nearest neighbor algorithm). In the past decade, a series of predictors have been designed for phenotype prediction. For example, Resch W *et al.* used a neural network model to identify the phenotype of HIV type 1 from loop 3 sequences [@pone.0017668-Resch1]. Pillai S *et al.* proposed a classifier based on support vector machine for V3 phenotype prediction [@pone.0017668-Pillai1]. Recently, Onuki R *et al.* also employed a support vector machine method for predicting phenotype from genotype data [@pone.0017668-Onuki1]. With the ever-increasing build-up of high-throughput techniques, biological data acquisition has never increased more rapidly. More and more biological networks, such as gene-regulatory networks and metabolic networks are constructed from multi data sources (e.g. microarrays, literature mining, and protein-protein interaction). Consequently, many network-based methods are proposed to contribute to various aspects of biology, including phenotype prediction. For instance, Keleta C *et al.* implemented the prediction of the 16 different growth phenotypes in E.*coli* based on regulated metabolic networks [@pone.0017668-Kaleta1]. McGary KL *et al.* demonstrate that the loss-of function *Saccharomyces cerevisiae* phenotypes are predictable in the functional gene network, and the proposed network-based method succeeded in the identification of yeast orthologs of human disease genes. In this research, we presented a new network-based method for predicting budding yeast protein phenotypes. Unlike previous methods, our method can rank the possible phenotypes associated with the query protein and shows a more comprehensive view of the protein\'s biological effects. With the results, we also demonstrated that using protein-protein network is effective for predicting protein phenotypes. Owing to many protein-protein network of other organisms are available, we suggest that this method will be widely applied. Materials and Methods {#s2} ===================== Data Set {#s2a} -------- Because of the complexity of phenotype research, we selected the budding yeast *Saccharomyces cerevisiae* (a well studied model organism [@pone.0017668-Gimeno1], [@pone.0017668-Lengeler1]) as a model system. The protein data used here was taken from CYGD [@pone.0017668-Guldener1] (the MIPS Comprehensive Yeast Genome Database, <ftp://ftpmips.gsf.de/yeast/>), which dedicated to information on the molecular structure and functional network of the budding yeast. Among the 6,732 proteins of the yeast proteome, only those with both sequence and phenotypic annotations were selected. Thus we obtained 1,460 such proteins belonging to 11 phenotypic categories (see [Table S1](#pone.0017668.s001){ref-type="supplementary-material"}). The number of proteins in each category was listed in the [Table 1](#pone-0017668-t001){ref-type="table"}, from which we can easily find that the total number of proteins (2,397) in 11 phenotypic categories is much larger than the total number of proteins (1,460). That is because many proteins exhibit more than one phenotype and this is the reason why we developed this method to predict the possible phenotypes with ranked scores, rather than only one predicted phenotype like previous tools. ::: {#pone-0017668-t001 .table-wrap} 10.1371/journal.pone.0017668.t001 Table 1 ::: {.caption} ###### Breakdown of 1,460 budding yeast proteins according to their 11 phenotypes. ::: {#pone-0017668-t001-1} Number Phenotype category Number of proteins -------- ------------------------------------------------------- -------------------- 1 Conditional phenotypes 536 2 Cell cycle defects 271 3 Mating and sporulation defects 198 4 Auxotrophies, carbon and nitrogen utilization defects 266 5 Cell morphology and organelle mutants 534 6 Stress response defects 147 7 Carbohydrate and lipid biosynthesis defects 46 8 Nucleic acid metabolism defects 218 9 Sensitivity to amino acid analogs and other drugs 124 10 Sensitivity to antibiotics 43 11 Sensitivity to immunosuppressants 14 See the texts of the paper for further explanation. ::: The yeast protein-protein interaction (PPI) network used here was retrieved from STRING [@pone.0017668-Jensen1] (<http://string.embl.de/>), whose primary mission is to provide researchers with both physical (direct) and functional (indirect) interactions. For each species, a PPI network is constructed by integrating huge information derived from numerous sources such as experimental repositories, computational methods, and text-mining methods. In the functional protein association network, the interaction unit consists of two nodes (proteins) and an edge between them. The interaction confidence score is used as the edge weight to represent the likelihood that a predicted association exists between two nodes. Weight confidence limits are as follows: low confidence −15% (or better), medium confidence −40%, high confidence −70%, highest confidence −90%. In this research, we chose the highest confidence limit −90% to obtain reliable yeast PPI network (see [Table S2](#pone.0017668.s002){ref-type="supplementary-material"}), which contains 32,513 functional linkages among 4,209 yeast proteins. Among the 1,460 proteins with phenotypic annotations, 1,267 proteins could be mapped to the yeast PPI network downloaded from STRING. Thus, the nodes in the network could be grouped into two kinds: those with phenotypic information, others without phenotypic information. Here, we called the protein with phenotypic annotation in the PPI network "seed protein", and the dataset consisting of 1,267 seed proteins "seed set", which were then used to test the network-based method. The availability of using the PPI network to predict protein phenotypes {#s2b} ----------------------------------------------------------------------- In the functional network, PPI contains both physical (direct) and functional (indirect) interactions. Physically interacting proteins exist in the same complex, while functional interacting proteins tend to participate in the same pathway or cellular process. Here, we investigated the relationships between complex/pathway and phenotype to explain the availability of using the PPI network to predict protein phenotypes. In order to analyze the relationship conveniently, we selected the proteins with single phenotype. The complex annotation of proteins was also downloaded from CYGD [@pone.0017668-Guldener1], and the pathway annotation of proteins was retrieved from KEGG [@pone.0017668-Kanehisa1] (Kyoto Encyclopedia of Genes and Genomes) (see [Table S3](#pone.0017668.s003){ref-type="supplementary-material"} and [Table S4](#pone.0017668.s004){ref-type="supplementary-material"}). Totally, these proteins belonged to 733 complexes and 86 pathways. Each protein was coded by the vectors:where if the protein belonged to the *i-th* complex/pathway, otherwise . Then *m-th* phenotype can be represented by the protein complex/pathway information as the vector:where *n* is the number of proteins that had the *m-th* phenotype. The similarity between any two phenotypes was calculated as:where is the vectors\' inner product, is the module of vector. Generally, two phenotypes are difficult to discriminated from each other using the complex/pathway if the value of the similarity of them is larger than 0.5. Using the protein complex information, the distribution of the similarities of 11 phenotypes was shown in [**Figure 1**](#pone-0017668-g001){ref-type="fig"}. Clearly, all the 55 similarities are smaller than 0.5. Because the proteins with the phenotype of sensitivity to immunosuppressants lacked the pathway annotation, the similarities of other 10 phenotypes were calculated using the protein pathway information. The distribution of the similarities of 10 phenotypes was shown in [**Figure 2**](#pone-0017668-g002){ref-type="fig"}, where two thirds of the similarities are smaller than 0.5. The results indicate the phenotypes can be classified by using protein complex/pathway information. Therefore, protein phenotypes can be predicted by using the functional PPI network. ::: {#pone-0017668-g001 .fig} 10.1371/journal.pone.0017668.g001 Figure 1 ::: {.caption} ###### The distribution of the similarities of 11 phenotypes that were represented by protein complex information. :::  ::: ::: {#pone-0017668-g002 .fig} 10.1371/journal.pone.0017668.g002 Figure 2 ::: {.caption} ###### The distribution of the similarities of 10 phenotypes that were represented by protein pathway information. :::  ::: Network-based Method {#s2c} -------------------- In the PPI network, when we were to predict the phenotypes of a node (protein), just like the weighted vote, not only the number of its neighbor nodes, but also the strengths of interactions (i.e., the edge weights) were considered by the method. The phenotypic categories of each protein in the network can be predicted as following. First, let us consider the PPI network consisting of proteins , in which seed proteins belonged to 11 phenotypic categories (), where represents the "Conditional phenotypes" category, the "Cell cycle defects", the "Mating and sporulation defects", and so forth (cf. [**Table 1**](#pone-0017668-t001){ref-type="table"}). And the phenotypes of the protein in the network can be denoted by where Towards a query protein , its interaction weights with *m* seed proteins can be defined as followswhere is the interaction weight (confidence score [@pone.0017668-Jensen1]) between and the protein in the seed set. If there is no edge between them, . Since we did not consider the self-interaction of protein, when . Subsequently, we proposed a new concept called "tethering potential" of protein to the phenotype to reflect the potential of protein belonging to the phenotype, which can be calculated as follows From this equation, we know that the proteins in seed set without association with the query protein do not contribute to the score of . Thus the tethering potential of protein to the phenotype can be also described as the sum of interaction weights of it with neighbor proteins of the phenotype in seed set. Obviously, the larger the value of is, the more likely the protein belongs to the phenotypic category. Therefore, the most likely phenotype of the query protein can be predicted to belong to the phenotypic category as followswhere stands for the argument of that maximizes the value of . However, many proteins in yeast give rise to more than one phenotype; the prediction result with only the most likely candidate phenotype is insufficient. In view of this, to make the method able to handle the proteins with multiple phenotypes and benefit biologists with more flexible information in prioritizing candidate phenotypes, we introduced a 11-D (dimensional) vector to reflect the likelihood that the query protein may give rise to each of the 11 phenotypes, which can be formulated as followswhere is a descending operator to sort the 11 scores of in descending order. Hence, we have . Accordingly, if , , , ..., then that the query protein gives rise to the 1^st^ phenotype (Conditional phenotypes) will have the maximum likelihood, that gives rise to the 7^th^ phenotype (Carbohydrate and lipid biosynthesis defects) will have the second maximum likelihood, that gives rise to the phenotype (Cell morphology and organelle mutants) will have the third maximum likelihood, and so forth (cf. [**Table 1**](#pone-0017668-t001){ref-type="table"}). In rare cases, when more than one element of the vector in Eq.6 has the same value, the order will be randomly sorted. Based on the descending order of Eq.6, the predicted results are respectively called the 1^st^-order predicted result, the 2^nd^-order predicted result, the 3^rd^-order predicted result, and so forth. Jackknife Cross-validation and Evaluation {#s2d} ----------------------------------------- In statistical prediction, three cross-validation methods are often used to examine the prediction quality: subsampling (K-fold) test, independent dataset test and jackknife test [@pone.0017668-Chou2]. Among the three methods, jackknife test is regarded as the most objective as discussed in Chou\'s work [@pone.0017668-Chou3], [@pone.0017668-Chou4] and has been used more and more frequently to test and evaluate various predictors [@pone.0017668-Cai1], [@pone.0017668-AfjehiSadat1], [@pone.0017668-Chen1], [@pone.0017668-Zeng1], [@pone.0017668-Ding1], [@pone.0017668-Zhou1], [@pone.0017668-Huang1], [@pone.0017668-Huang2], [@pone.0017668-Huang3], [@pone.0017668-Huang4]. In this research, the jackknife cross-validation was also applied to test the network-based method. During the validation, each protein in the seed set is in turn knocked out as a query protein sample, and the remaining proteins of the seed set in the PPI network are used for prediction by the network-based method. Thus, the *i-th* order prediction accuracy can be calculated as follows Where is the number of correctly predicted proteins of the *j-th* phenotypic category in the seed set, and *N* is the total number of proteins in the seed set. Finally, the 11-order prediction accuracies are obtained to evaluate the network-based method. The large with a small *i* and the small with a large *i* imply a good performance of the method. The average number of phenotypes that each protein in the network exhibits can be calculated as follows Therefore, another evaluation for the network-based method was proposed as the likelihood that the first *r-order* predicted results include all the phenotypes of proteins, which can be calculated as follows A large accompanied with a small *r* also implies a good performance of the method for the protein phenotype prediction. Prediction {#s2e} ---------- Besides the seed proteins, there are also 2,942 proteins in the PPI network. The tethering potential of such protein to the each phenotype can be calculated according to Eq. (7) and then ranked in descending order. In this manner, the phenotypes of these proteins can be predicted by the network-based method. Results and Discussion {#s3} ====================== Performance of Network-based method {#s3a} ----------------------------------- Through leave-one-out cross-validation, the overall 11-order success rates by the network-based method on the aforementioned 1,267 seed proteins are listed in [**Table 2**](#pone-0017668-t002){ref-type="table"}. As we can see from the table, the most likely (first-order) prediction accuracy is 65.4%, and the least likely (last-order) one is 3.39%. The former minus the latter equals 61%. Based on the prediction criteria, the bigger the difference value is, the better the method performs. According to [**Table 2**](#pone-0017668-t002){ref-type="table"}, a downward-slope curve is drawn in the [**Figure 3**](#pone-0017668-g003){ref-type="fig"}, showing that higher-order phenotype prediction is better than the lower-order one. This is the exact phenomenon that we want to see, and it may imply that the predicted phenotypic categories of proteins are well arranged by the method according to the prediction criteria. ::: {#pone-0017668-g003 .fig} 10.1371/journal.pone.0017668.g003 Figure 3 ::: {.caption} ###### A downward-slope curve to show the relations among the different order prediction accuracies. :::  ::: ::: {#pone-0017668-t002 .table-wrap} 10.1371/journal.pone.0017668.t002 Table 2 ::: {.caption} ###### The leave-one-out cross-validation (Jackknife test) success rates by a random guess and the network-based method. ::: {#pone-0017668-t002-2} Most likely category ---------------------- -------------- ------ ------ ------ ------ ------ ------ Order 1 2 3 4 5 6 Random Guess Accuracy (%) 15.5 15.5 15.5 15.5 15.5 15.5 Network-based Method 65.4 34.1 20.7 13.3 8.76 6.47 Least likely category ----------------------- -------------- ------ ------ ------ ------ ------ -- Order 7 8 9 10 11 Random Guess Accuracy (%) 15.5 15.5 15.5 15.5 15.5 Network-based Method 5.84 5.21 3.47 3.39 3.39 ::: The average number of phenotypes that each seed protein has is 1.7 according to Eq. (11). The chance that a random guess of a protein phenotype will succeed is 1.7/11 = 15.4%, much lower than the first order prediction success rate. As is shown in the [**Table 2**](#pone-0017668-t002){ref-type="table"}, the first 3 prediction accuracies are larger than the success rates of random guess. And the likelihood of the first 3-order predicted results including the phenotypic categories of the proteins in seed set is 70.6% according to the Eq. (12). These results may imply that our method performs well in the prediction of protein phenotypes in budding yeast. In genetics, mutations that cause the same phenotype are inferred to functionally associated, and vice versa [@pone.0017668-McGary1]. Phenotype is a multifactorial trait that often results from the contribution of many proteins. Because the interacting proteins are often in the same complex or pathway, it is rational to expect that interacting proteins often share the common phenotypes. For example, the interactions of seed protein YBR039W with the other seed proteins are listed in [**Table 3**](#pone-0017668-t003){ref-type="table"}. The complex information about those proteins is retrieved from CYGD [@pone.0017668-Guldener1]. We can easily see that protein YBR039W and its neighbors YBL099W, YDL004W, YDR298C, YLR295C, YML081C-A, YPL078C, YPL271W are members of the same F0/F1 ATP synthase (complex V) complex. Additionally, proteins YDR298C and YPL078C are also members of complex in [@pone.0017668-Ho1], and protein YPR024W is component of Yme1 protease complex. And these proteins share the common phenotype auxotrophies, carbon and nitrogen utilization defects. Therefore, when protein YBR039W is predicted as a test sample by the method, the first candidate phenotype will be assigned its real phenotype. For another example, the interactions of seed protein YDL028C with the other seed proteins are listed in [**Table 4**](#pone-0017668-t004){ref-type="table"}. The information of pathways that yeast proteins participate in is retrieved from Kyoto Encyclopedia of Genes and Genomes [@pone.0017668-Kanehisa2] (KEGG). Except proteins YDR168W, YKL042W, YPL209C with no pathway annotation, proteins YDL028C, YBL084C, YGL116W, YGR113W, YGR188C, YIL106W, YKL022C, YMR055C, YOR026W involve in the same pathway sce04111 (Cell cycle in budding yeast). The loss-of-function of any one of these 9 proteins likely disrupts the mitotic cell cycle progression and lead to cell cycle defects. Based on the interactions listed in the table, we can arrange the first, second candidate phenotype of protein YDL028C as the cell cycle defects, cell morphology and organelle mutants respectively according to the prediction criteria. The correct phenotype predictions of proteins YBR039W and YDL028C support the hypothesis that the functional associated proteins often share the same phenotypes. Therefore, the protein phenotypes can be predicted from the phenotypes of its interacting proteins by the method. ::: {#pone-0017668-t003 .table-wrap} 10.1371/journal.pone.0017668.t003 Table 3 ::: {.caption} ###### Interactions of protein YBR039W with its neighbor proteins. ::: {#pone-0017668-t003-3} Protein A Phenotype Complex Protein B Phenotype Complex Weight ----------- ----------- --------- ----------- ------------ --------- -------- YBR039W P1 C1 YBL099W P1 C1 999 YBR039W P1 C1 YDL004W P1 C1 999 YBR039W P1 C1 YDR298C P1 C1; C2 999 YBR039W P1 C1 YLR295C P1 C1 917 YBR039W P1 C1 YML081C-A P1; P2 C1 934 YBR039W P1 C1 YPL078C P1 C1; C2 999 YBR039W P1 C1 YPL271W P1 C1 997 YBR039W P1 C1 YPR024W P1; P2; P3 C3 986 C1 represents F0/F1 ATP synthase (complex V), C2 represents Complex in study [@pone.0017668-Zhou1], C3 represents Yme1 protease complex, P1 represents Auxotrophies, carbon and nitrogen utilization defects, P2 represents Cell morphology and organelle mutants, P3 represents Conditional phenotypes. ::: ::: {#pone-0017668-t004 .table-wrap} 10.1371/journal.pone.0017668.t004 Table 4 ::: {.caption} ###### Interactions of protein YDL028C with its neighbor proteins. ::: {#pone-0017668-t004-4} Protein A Phenotype Pathway Protein B Phenotype Pathway Weight ----------- ----------- ---------- ----------- ------------ ------------------------------ -------- YDL028C P4; P5 sce04111 YBL084C P4; P6 sce04111; sce04113; sce04120 929 YDL028C P4; P5 sce04111 YDR168W P4 no annotation 999 YDL028C P4; P5 sce04111 YGL116W P4 sce04111; sce04113; sce04120 956 YDL028C P4; P5 sce04111 YGR113W P4; P5 sce04111 999 YDL028C P4; P5 sce04111 YGR188C P5 sce04111; sce04113 999 YDL028C P4; P5 sce04111 YIL106W P4; P5 sce04111 988 YDL028C P4; P5 sce04111 YKL022C P4; P7 sce04111; sce04113; sce04120 929 YDL028C P4; P5 sce04111 YKL042W P4 no annotation 990 YDL028C P4; P5 sce04111 YMR055C P4 sce04111 984 YDL028C P4; P5 sce04111 YOR026W P4; P7 sce04111 978 YDL028C P4; P5 sce04111 YPL209C P4; P5; P7 no annotation 984 P4 represents Cell cycle defects, P5 represents Cell morphology and organelle mutants, P6 represents Nucleic acid metabolism defects, P7 represents Conditional phenotypes, Sce04111 represents cell cycle pathway in budding yeast. ::: Protein phenotype prediction with inactivating its interacting protein {#s3b} ---------------------------------------------------------------------- Here, we discuss the robustness of our method by applying the method to the proteins whose interacting proteins are inactivated. First, we chose a protein and took away one of its interacting proteins from the PPI network. Then the phenotype of the protein was predicted by the method based on the broken PPI network. In this way, the phenotypes of 6 proteins were predicted, as shown in [**Table 5**](#pone-0017668-t005){ref-type="table"}. The phenotypes predicted from the unbroken network and the recent phenotype studies focusing on these proteins are also listed in [**Table 5**](#pone-0017668-t005){ref-type="table"}. We found that the phenotypes predicted from the broken network were different from the phenotypes predicted from the unbroken network, while the proteins were verified to have these new phenotypes predicted from broken network in the recent studies. For example, with protein YOR196C in the network, the 1^st^ order predicted phenotype of protein YER178W by the method is "auxotrophies, carbon and nitrogen utilization defects", which is the same as the annotation from CYGD [@pone.0017668-Guldener1]. After inactivating protein YOR196C, the phenotype of protein YER178W is predicted as the "conditional phenotypes". In the study [@pone.0017668-Sinha1], protein YER178W was reported to have the phenotype-"Heat sensitivity: increased", which is one kind of "conditional phenotypes" according to the phenotype classification in CYGD. In the table, the new phenotypes of other proteins predicted from the broken network can also be supported by the literatures [@pone.0017668-Dudley1], [@pone.0017668-Morton1], [@pone.0017668-Watanabe1], [@pone.0017668-Altmann1], [@pone.0017668-Ungar1], [@pone.0017668-Cai2]. The examples listed in the table indicate that our method may provide new phenotypes for proteins and serve as a complementary tool for the existing resources. ::: {#pone-0017668-t005 .table-wrap} 10.1371/journal.pone.0017668.t005 Table 5 ::: {.caption} ###### Phenotypes of proteins predicted by our method with/without inactivating its interacting protein. ::: {#pone-0017668-t005-5} Protein Phenotype from CYGD [@pone.0017668-Guldener1] Phenotype predicted by our method without inactivating the interacting protein Inactivated interacting protein Phenotype predicted by our method with inactivating the interacting protein Phenotype from literatures --------- --------------------------------------------------------------------------------------------- -------------------------------------------------------------------------------- --------------------------------- ----------------------------------------------------------------------------- ----------------------------------------------------------------------------------------------------------------------------- YER178W Auxotrophies, carbon and nitrogen utilization defects Auxotrophies, carbon and nitrogen utilization defects YOR196C Conditional phenotypes Heat sensitivity: increased [@pone.0017668-Sinha1] YML035C Conditional phenotypes Conditional phenotypes YDR226W Cell morphology and organelle mutants Toxin resistance: increased [@pone.0017668-Morton1] YMR198W Cell cycle defects Cell cycle defects YPR141C Cell morphology and organelle mutants Bud morphology: abnormal [@pone.0017668-Watanabe1] YOR254C Conditional phenotypes Cell cycle defects Mating and sporulation defects Conditional phenotypes YKL073W Cell morphology and organelle mutants Mitochondrial morphology: abnormal [@pone.0017668-Altmann1] Telomere length: increased [@pone.0017668-Ungar1] YDL198C Conditional phenotypes Conditional phenotypes YPL240C Auxotrophies, carbon and nitrogen utilization defects Utilization of nitrogen source: absent [@pone.0017668-Cai2] Utilization of carbon source: decreased [@pone.0017668-Dudley1] YPR166C Auxotrophies, carbon and nitrogen utilization defects Cell morphology and organelle mutants Auxotrophies, carbon and nitrogen utilization defects YHR147C Conditional phenotypes Heat sensitivity: increased [@pone.0017668-Sinha1] ::: Application and improvement {#s3c} --------------------------- As is discussed above, the first 3-order predicted results (approximately double the average number of phenotypes 1.7) can be considered as the candidate phenotypes of the proteins concerned by the biologists. Genetic experiments can focus on these candidate phenotypes of the proteins, which may accelerate the research progress and decrease the cost. At least, the last three predicted phenotypes can be excluded because the last 3-order prediction accuracies are lower than 5% (See [**Table 2**](#pone-0017668-t002){ref-type="table"}). The effectiveness of the functional network for predicting phenotypes of proteins in yeast suggests the possibility of application to other species. The method is based on the functional protein association network. Besides an abundance of such networks in STRING [@pone.0017668-Jensen1] (Version 8.0 of STRING covered 630 networks of different organisms), the PPI networks can also derived from worm PPI database [@pone.0017668-Li1], fly database [@pone.0017668-Giot1], human PPI database [@pone.0017668-Rual1], [@pone.0017668-Lehner1], [@pone.0017668-Stelzl1], and so on. When in possession a series of proteins with known phenotypes, one can predict the possible phenotypes of other proteins in the networks. Therefore, the method can be easily applied to the prediction of protein phenotypes in other organisms, especially model organisms. The performance of our method can be improved if the following problems are solved. First, increase the quality of PPI network and exclude the false positive interaction; currently we used high confidence score cutoff to filter the network (See section Data Set). Second, proteins in the same complex or pathway may exert opposite effects on a phenotype, playing as actors or repressors [@pone.0017668-McGary1]. If the network can discriminate the positive or negative regulation, our method can be modified and the performance will be improved. Third, the performance of the network-based method depends on the number of seed proteins. This problem can be solved in future when the phenotypes of more proteins are investigated. In summary, identification of protein phenotypes is an extremely complicated work and there is a long way to go. Conclusion {#s3d} ---------- In this research, we proposed a multi-target model [@pone.0017668-Huang2] to predict phenotypes of proteins in budding yeast based on the protein-protein network. Because some proteins can give rise to more than one phenotype, rather than the most likely phenotype, a series of candidate phenotypes are predicted for each protein. With the performance of the method, it is anticipated that the promising approach may serve as a useful tool for annotating the phenotypes for uncharacterized protein sequences. Supporting Information {#s4} ====================== Table S1 ::: {.caption} ###### **The 1,460 proteins with both sequence and phenotype information retrieved from CYGD (the Comprehensive Yeast Genome Database) (Guldener U, Munsterkotter M, Kastenmuller G, Strack N, van Helden J, et al. (2005) CYGD: the Comprehensive Yeast Genome Database. Nucleic acids research 33: D364-368.).** The corresponding phenotype of the phenotype number can be found in [Table 1](#pone-0017668-t001){ref-type="table"}. (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S2 ::: {.caption} ###### **Yeast Protein-Protein Interaction Network.** The protein-protein interaction were downloaded from STRING (Search Tool for the Retrieval of Interacting Genes/Proteins) (Jensen LJ, Kuhn M, Stark M, Chaffron S, Creevey C, et al. (2009) STRING 8--a global view on proteins and their functional interactions in 630 organisms. Nucleic acids research 37: D412-416.) with the highest confidence, i.e., the confidence scores are not less than 900. (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S3 ::: {.caption} ###### **The proteins and the complexes they belong to in yeast.** The information was retrieved from CYGD (the Comprehensive Yeast Genome Database) (Guldener U, Munsterkotter M, Kastenmuller G, Strack N, van Helden J, et al. (2005) CYGD: the Comprehensive Yeast Genome Database. Nucleic acids research 33: D364-368.). (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S4 ::: {.caption} ###### **The proteins and the pathways they belong to in yeast.** The information was retrieved from KEGG (Kyoto Encyclopedia of Genes and Genomes) (Kanehisa M, Goto S, Hattori M, Aoki-Kinoshita KF, Itoh M, et al. (2006) From genomics to chemical genomics: new developments in KEGG. Nucleic Acids Res 34: D354-357.). (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: We thank CYGD and STRING for supply data to support not-for-profit research efforts. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**The authors have no support or funding to report. [^1]: Conceived and designed the experiments: LH YDC. Performed the experiments: LH TH. Analyzed the data: LH XJL YDC. Contributed reagents/materials/analysis tools: LH XJL. Wrote the paper: LH TH YDC.

PubMed Central

2024-06-05T04:04:19.827567

2011-3-10

PMC3053378

Introduction {#s1} ============ Bacteremia is a common disease associated with significant mortality [@pone.0017653-BrunBuisson1]. Early categorization of patients with different prognoses is difficult in the absence of a timely sensitive and specific biomarker. Although C-reactive protein (CRP) has been widely used as a prognostic marker in infectious diseases, its prognosic value in bacteremia and sepsis is weak [@pone.0017653-Silvestre1]. Pentraxins are multi-functional pattern-recognition receptors (PRR) interacting with selected viral, fungal and bacterial components. Long pentraxin 3 (PTX3), CRP and serum amyloid P (SAP) are the key members of the pentraxin superfamily. Although the long (PTX3) and short pentraxins (CRP and SAP) share common sequences they are encoded by different genes and are differentially regulated. CRP is produced in the liver, while PTX3 is an inflammatory mediator produced by various cells in peripheral tissues, for example macrophages, dendritic cells, endothelial cells, ovarian granulosa cells, fibroblasts, adipocytes and smooth muscle cells in response to the proinflammatory signals lipopolysaccharide (LPS), interleukin-1 (IL-1) and tumor necrosis factor-alfa (TNF-alfa) and Toll-like receptor activation [@pone.0017653-Bottazzi1], [@pone.0017653-Mantovani1]. PTX3 behaves as an acute-phase protein, as its blood levels, low in normal conditions (\<2 ng/ml in humans), increase rapidly in the plasma during inflammation (sepsis, endotoxin shock and other inflammatory conditions) [@pone.0017653-Bottazzi1]. PTX3 released in response to microbial recognition can bind specific pathogens such as fungi, bacteria and viruses, promoting phagocytosis and subsequent clearance of the pathogen via its binding to complement component C1q to induce classical complement activation [@pone.0017653-Bottazzi1], [@pone.0017653-Garlanda1], [@pone.0017653-Garlanda2]. Thus, PTX3 has a non-redundant role in the regulation of the innate immune response by contributing to the opsonization and clearance of apoptotic or necrotic cells [@pone.0017653-Bottazzi1]. In recently published studies, a high PTX3 level has been shown to be associated with mortality in severe sepsis and septic shock [@pone.0017653-Mauri1], and to be an early indicator of shock in severe meningococcal disease [@pone.0017653-Sprong1]. Elevated plasma levels of PTX3 predict disease severity in dengue virus infection [@pone.0017653-Mairuhu1] and in leptospirosis [@pone.0017653-Wagenaar1]. In critically ill patients PTX3 correlates with severity of disease and infection [@pone.0017653-Muller1]. It has been shown to predict bloodstream infection and severe disease in febrile patients admitted to emergency departments [@pone.0017653-deKruif1] and indicates acute respiratory distress syndrome (ARDS) in critically ill patients [@pone.0017653-Mauri2]. PTX3 has been shown to act as biomarker of acute lung injury [@pone.0017653-He1]. However, to the best of our knowledge no study has investigated the prognostic value of PTX3 in a cohort of patients with bacteremic infection. We have previously studied prognostic factors associated with case fatality in bacteremia and found obesity (BMI≥30) and smoking to be associated with poor outcome [@pone.0017653-Huttunen1]. We sought here to assess the prognostic value of plasma PTX3 in relation to other known prognostic factors in bacteremic patients. The prognostic values of two members of the pentraxin superfamily, CRP and PTX3, were compared. We show that PTX3 measurement may offer novel opportunities for the early prognostic stratification of bacteremic patients. Materials and Methods {#s2} ===================== Patients {#s2a} -------- The study material comprised 132 adult patients with bacteremia admitted to Tampere University Hospital, Tampere, Finland, from June 1999 to February 2004 ([Table 1](#pone-0017653-t001){ref-type="table"}). Patients were recruited from the emergency room, intensive care unit (ICU) and medical wards of the hospital. Patient recruitment, clinical data collection and sample collection were prospective. Samples for PTX3 were analyzed after hospitalization. ::: {#pone-0017653-t001 .table-wrap} 10.1371/journal.pone.0017653.t001 Table 1 ::: {.caption} ###### Baseline characteristics of the study population (132 patients). ::: {#pone-0017653-t001-1} Character ---------------------------------------------------------------- ------------------- Age, median (range) 62 (18--93 years) Gender (female/male) 62/70 **Causative organism** *S. aureus* 32 (24%) *Str. pneumoniae* 37 (28%) B- hemolytic streptococcus 22 (17%) *E. coli* 41 (31%) **Focus of infection (one patient may have several focuses)** Lung 33 (25%) Skin 33 (25%) Urinary 29 (22%) Osteomyelitis/spondylitis 13 (10%) Other or unknown focus 41 (31%) BMI (kg/m^2^), median (range)[a](#nt101){ref-type="table-fn"} 26 (15--39) Diabetes mellitus (type 1 or 2) 33 (25%) Current smoking[b](#nt102){ref-type="table-fn"} 33 (28%) Alcohol abuse 21 (16%) Cancer (solid or hematological) 23 (17%) At least one chronic disease 107 (81%) McCabe class II or III[c](#nt103){ref-type="table-fn"} 22 (17%) Cardiac disease[d](#nt104){ref-type="table-fn"} 41 (31%) SOFA score, median (quartiles)[e](#nt105){ref-type="table-fn"} 2 (1--6) ICU treament[f](#nt106){ref-type="table-fn"} 42 (32%) Died (d-30 case fatality) 18 (14%) a BMI data available on 101 patients, b smoking data available on 120 patients, c rapidly or ultimately fatal disease, d valvular, coronary artery disease, heart failure or cardiac myopathy, e sequential organ failure assessment, f intensive care unit. ::: In our hospital blood cultures are routinely taken in cases with symptoms or signs of systemic infection (fever or hypothermia, tachycardia or tachypnea combined with leukocytosis or leukopenia and/or elevated C-reactive protein (CRP)). The BACTEC 9240 (BD Diagnostic Systems, Sparks, MD, USA) blood culture system was used with standard media. Patients were identified according to microbiological blood culture finding, and only those with bacteremia caused by *S. aureus, Str. pneumoniae*, β-hemolytic streptococcus or *E. coli*, the most common causative organisms in community-acquired bacteremia, were included in the study, other microbes being excluded beforehand. Blood culture-negative patients with or without sepsis syndrome and those not consenting were not included. All patients included in the study had verified infection. Only patients at least 16 years of age were enrolled. The clinicians (J.S. or J.L.) were informed by the clinical microbiologist (R.V.) of a positive blood culture from Mondays to Thursdays and the patients were enrolled in the study whenever possible to adjust to the daily schedule. We were able to recruit zero to two patients per week during the study period. Since the clinicians had no knowledge of details regarding the patients or their disease severity prior to recruitment, selection was based solely on the blood culture finding. Upon notification by the clinical microbiologist the clinicians (J.L. and J.S.) asked patients to participate and interviewed and examined those consenting. Information was gathered from hospital records at the time of a hospital visit and hospital records were also reviewed subsequent to hospitalization (R.H.). Altogether 149 out of 152 patients agreed to participate. Samples for PTX3 determinations during 1--4 days after positive blood culture were available in 132 cases, and these patients were recruited as the final study population. The study was approved by the Ethics Committee of Tampere University Hospital. Written informed consent was obtained from patients or first-degree relatives. Underlying diseases and chronic conditions {#s2b} ------------------------------------------ Chronic diseases and sources of bacteremia were registered. Calculation of body mass index (BMI, kg/m^2^) was based on weight and height as reported by the patient on admission. Patients were defined as obese if their BMI was ≥30 kg/m^2^. Alcohol abuse was defined as consumption of 300 g absolute alcohol per week or a known social or medical problem due to alcohol use. Patients were defined as current smokers and nonsmokers, i.e. those who had never smoked or had stopped smoking. McCabe classification [@pone.0017653-McCabe1] was used to determine the severity of any underlying disease. Collection of clinical and laboratory data {#s2c} ------------------------------------------ Clinical data and laboratory findings were registered on admission and during 6 consecutive days. Alterations in mental status were evaluated on the Glasgow Coma Scale (GCS), possible mechanical ventilation and the need for intensive care unit (ICU) treatment were recorded. Mean arterial pressure (MAP) ((systolic+2 x diastolic blood pressure)/3) and SOFA score (sequential organ failure assessment) [@pone.0017653-Vincent1] were calculated. The maximum SOFA score (days 0--6) for every patient was used in analysis. Disease severity was assessed by SOFA score, severe disease being defined as a score ≥4. Laboratory tests included plasma C-reactive protein (CRP, mg/l), blood platelets (x10^9^/l), plasma bilirubin (µmol/l), plasma creatinine level (µmol/l) and blood leukocyte count (x10^9^/l). The case fatality rate was studied within 14 and 30 days after a positive blood culture (d--14 and d--30 case fatality). Determination of PTX3 plasma levels {#s2d} ----------------------------------- EDTA plasma samples for PTX3 determination were taken during patientś hospitalization and were stored at −70° until analyzed. PTX3 concentrations were determined in EDTA-plasma using a commercial solid-phase enzyme-linked immunosorbent assay (ELISA) according to the manufacturer\'s instructions (Quantikine DPTX 30; R&D Systems Inc., Minneapolis, USA). Samples in which the PTX3 concentration exceeded the detection range (n = 59) were serially diluted in assay diluent until they reached the dynamic range of the assay. According to the manufacturer, the mean detection limit for PTX3 is 0.025 ng/ml and the assay exhibits no cross-reactivity with either CRP or serum amyloid P. The plates were read with a Multiskan Ascent photometer (Thermo Scientific, Waltham, MA, USA) at 450 nm and corrected for readings at 540 nm. Samples for PTX3 determinations were taken in the acute phase (days 1 to 4) (n = 132 patients), on day 13--18 (13--18 days after blood culture) (n = 73 patients) and on recovery (\>25 days after positive blood culture) (n = 89 patients). Multiple samplings in the same patient were always performed on separate days. The maximum values for PTX3 for every patient measured during 1--4 days after positive blood culture were determined. Since patient recruitment was based on blood culture, which only became positive the following day, no samples for PTX3 were available on day 0 (blood culture day). Statistical analysis {#s2e} -------------------- An SPSS package (version 7.5 and version 10) was used for statistical analyses and a two-sided p-value \<0.05 was taken as cut-off for statistical significance. Categorical data were analyzed by *X^2^* test or Fisheŕs exact test when appropriate, nonparametric data by Mann-Whitney U-test or Kruskal-Wallis test. A logistic regression model was used to study the independent effect of high PTX3 activity on mortality models adjusted for potential confounders. Odds ratios (ORs) were expressed with their 95% confidence intervals (CI). The survival curve was calculated using the Kaplan-Meier method and survival differences between groups were compared using the log rank test. The accuracy of maximum PTX3 value and CRP in predicting case fatality was evaluated using ROC curves [@pone.0017653-Boyd1]. In this method, a test which is perfect has 100% sensitivity and no false-positives (1-specificity = 0) and will have an area under the curve (AUC) of 1.0, whereas a test of no diagnostic value would have an AUC of 0.5. The 95% confidence intervals were calculated. Results {#s3} ======= Baseline characteristics of bacteremia patients are shown in [Table 1](#pone-0017653-t001){ref-type="table"}. All subjects were treated with an empiric antibiotic regimen, and when necessary antimicrobial treatment was changed according to culture results. In all patients the causative organism proved susceptible to the first empiric antibiotic treatment selected on admission. Patients received antimicrobial therapy for a median of 17 days after blood culture. PTX3 values in bacteremic patients {#s3a} ---------------------------------- The median plasma PTX3 value in the acute phase (maximum value 1 to 4 days after blood culture) was 7.8 ng/ml (interquartile range 3.7--17.5 ng/ml) and 3.0 ng/ml on days 13--18 after blood culture (interquartile range 1.2--6.6 ng/ml). Values decreased on recovery; the median value \>25 days after blood culture was 1.1 ng/ml (interquartile range 0.5--2.0 ng/ml). Of chronic conditions, alcohol abusers had higher PTX3 values compared to patients without a history of alcohol abuse (maximum values 1 to 4 after blood culture 12.6 ng/ml compared to 7.3 ng/ml, p = 0.036, respectively). However, there was no difference between groups of patients in PTX3 levels stratified by other chronic conditions, age, sex or causative organism (data not shown). PTX3 and the outcome of bacteremia {#s3b} ---------------------------------- Median PTX3 values were significantly higher in nonsurvivors compared to survivors on days 1 to 2 (51.2 vs 9.7 ng/ml, p = 0.008), on day 3 (34.9 vs 5.3 ng/ml, p\<0.001) and on day 4 (24.6 vs 4.5 ng/ml, p\<0.001) after the initial diagnosis (blood culture day) ([Table 2](#pone-0017653-t002){ref-type="table"}). Maximum PTX3 values on days 1 to 4 after the initial diagnosis (blood culture day) were significantly higher in nonsurvivors compared to survivors (median values 44.8 vs 6.4 ng/ml, p\<0.001). ::: {#pone-0017653-t002 .table-wrap} 10.1371/journal.pone.0017653.t002 Table 2 ::: {.caption} ###### Pentraxins in patients with bacteremia[a](#nt108){ref-type="table-fn"}. ::: {#pone-0017653-t002-2} Days after diagnosis Plasma PTX3 value (ng/ml), median (quartiles) p-value Plasma CRP value (mg/l), median (quartiles) p-value --------------------------------- ----------------------------------------------- ----------------- --------------------------------------------- ---------------- ---------------- ------- **Day 1--2** 51.2 (40.8--113.5) 9.7 (6.4--17.7) 0.008 280 (206--368) 234 (158--335) 0.132 **Day 3** 34.9 (12.3--63.2) 5.3 (2.6--13.1) \<0.001 193 (163--238) 144 (68--232) 0.05 **Day 4** 24.6 (11.0--63.8) 4.5 (2.7--8.9) \<0.001 129 (105--160) 102 (48--168) 0.100 **Maximum value (days 1 to 4)** 44.8 (10.7--69.4) 6.4 (3.4--13.5) \<0.001 280 (206--368) 236 (155--334) 0.132 Plasma long pentraxin 3 (PTX3) and short pentraxin C-reactive protein (CRP) values 1 to 4 days after blood culture (diagnosis) in bacteremia nonsurvivors and in survivors. a PTX3 values available for 34 patients (5 nonsurvivors and 29 survivors) on day 1--2, 81 patients (12 nonsurvivors and 69 survivors) on day 3, 104 patients (17 nonsurvivors and 87 survivors) on day 4. CRP values available for 126 patients (18 nonsurvivors and 108 survivors) on day 1--2, 85 patients (12 nonsurvivors and 73 survivors) on day 3, 109 patients (15 nonsurvivors and 94 survivors) on day 4. ::: The optimal cut-off value for the maximum PTX3 values on days 1--4 in predicting fatal disease was estimated using ROC curve, illustrated in [Figure 1](#pone-0017653-g001){ref-type="fig"}. The PTX3 value at a cut-off level of 15 ng/ml showed a sensitivity of 72% and a specificity of 81% in detecting fatal disease, and this cut-off point was used to classify patients into those with high or low PTX3 value. High PTX3 values were associated with several endpoints indicative of severe disease ([Table 3](#pone-0017653-t003){ref-type="table"}). [Figure 2](#pone-0017653-g002){ref-type="fig"} shows cumulative 30-d survival in bacteremia patients with maximum plasma long pentraxin 3 (PTX3) level (1--4 days after blood culture) \>15 ng/ml compared to those with ≤15 ng/ml. ::: {#pone-0017653-g001 .fig} 10.1371/journal.pone.0017653.g001 Figure 1 ::: {.caption} ###### PTX3 and CRP ROC curves. Receiver operating characteristic (ROC) curves for maximal plasma long pentraxin 3 (PTX3) and C-reactive protein (CRP) levels detected on days 1--4 after positive blood culture in relation to case fatality in bacteremia patients. :::  ::: ::: {#pone-0017653-g002 .fig} 10.1371/journal.pone.0017653.g002 Figure 2 ::: {.caption} ###### PTX3 survival curves. Cumulative 30-d survival in bacteremia patients with maximum plasma long pentraxin 3 (PTX3) level (1--4 days after blood culture) \>15 ng/ml compared to those with ≤15 ng/ml. The survival curve was calculated using the Kaplan-Meier method, and survival differences between groups were compared by log-rank test. :::  ::: ::: {#pone-0017653-t003 .table-wrap} 10.1371/journal.pone.0017653.t003 Table 3 ::: {.caption} ###### Clinical characteristics of patients stratified by maximum plasma long pentraxin 3 PTX3 value (1 to 4 days after blood culture). ::: {#pone-0017653-t003-3} Characteristic High PTX3(\>15 ng/ml)N = 35 Low PTX3(≤15 ng/ml)N = 97 OR (95% CI) p-value ----------------------------------------------------------------------- ----------------------------- --------------------------- ------------------- --------- Died (d-30 case fatality) 13 (37%) 5 (5%) 10.9 (3.5--33.7) \<0.001 Died (d-14 case fatality) 11 (31%) 1 (1%) 44.0 (5.4--357.7) \<0.001 Hypotensive P\<70 mmHg) 26 (74%) 26 (27%) 7.9 (3.3--19.0) \<0.001 Needed ICU stay[a](#nt109){ref-type="table-fn"} 25 (71%) 17 (18%) 11.8 (4.8--29.0) \<0.001 Needed vasopressives 20 (57%) 6 (6%) 20.2 (7.0--58.6) \<0.001 Lowered Glasgow coma scale (\<15) 26 (74%) 27 (28%) 7.5 (3.1--18.0) \<0.001 Needed mechanical ventilation 16 (46%) 4 (4%) 19.6 (5.9--65.1) \<0.001 Highest SOFA score≥4[b](#nt110){ref-type="table-fn"} 29 (83%) 26 (27%) 13.2 (4.9--35.4) \<0.001 Lowest MAP[c](#nt111){ref-type="table-fn"} (mmHg), median (quartiles) 59 (52--73) 78 (68--92) \$ \<0.001 Highest SOFA score, median (quartiles) 9 (4--13) 2 (0--4) \$ \<0.001 Highest bilirubin level (µmol/l), median (quartiles) 23 (16--74) 17 (13--29) \$ 0.006 Highest creatinine level (µmol/l), median (quartiles) 146 (91--219) 98 (75--174) \$ 0.028 Median neutrophil count (x10^9^/l) (quartiles) (n = 112) 9.3 (8.1--13.2) 6.7 (3.8--9.8) \$ \<0.001 Lowest platelet count (x10^9^/l), median (quartiles) 86 (28--181) 171 (113--235) \$ \<0.001 a intensive care unit, b sequential organ failure assessment, c mean arterial pressure, \$continuous variable (OR and CI cannot be applied). ::: The independent effect of high (\>15 ng/ml) maximum PTX3 value on case fatality was studied in a logistic regression model adjusted for potential confounders ([Table 4](#pone-0017653-t004){ref-type="table"}). High maximun PTX3 value was studied together with one confounder at a time in a logistic regression model, as there were only 18 patients who died. The following grouping variables have previously been shown to be associated with case fatality in a univariate model in this material: obesity, smoking, alcohol abuse, and high SOFA score (≥4)[@pone.0017653-Huttunen1]. High PTX3 detected on days 1--4 after blood culture retained its significance in the logistic regression model in all combinations. Obesity and high SOFA score (≥4) also remained independent factors associated with case fatality when studied together with high PTX3. ::: {#pone-0017653-t004 .table-wrap} 10.1371/journal.pone.0017653.t004 Table 4 ::: {.caption} ###### The independent effect of high maximum plasma long pentraxin 3 (PTX3) value (\>15 ng/ml) on days 1--4 on case fatality in a logistic regression model adjusted for potential confounders. ::: {#pone-0017653-t004-4} Variables in the logistic regression model Odds ratio for high PTX3 (ng/ml) ------------------------------------------------------- ---------------------------------- **High PTX3 (\>15 ng/ml) +** male sex and age 11.3 (3.5--36.2) obesity (≥30 kg/m^2^)[a](#nt113){ref-type="table-fn"} 11.1 (2.3--54.1) alcohol abuse 10.0 (3.2--31.2) current smoking[b](#nt114){ref-type="table-fn"} 9.9. (2.7--36.4) McCabe class II or III 10.9 (3.5--34.0) SOFA score (≥4)[c](#nt115){ref-type="table-fn"} 4.8 (1.4--16.4) Obesity, smoking and alcohol abuse are included in the model as they proved to be factors significantly associated with case fatality in the univariate model in this material [@pone.0017653-Huttunen1]. a BMI data available on 101 patients, obesity also remained a significant factor associated with case fatality in the logistic regression model. b Smoking data available on 120 patients. c Sequential organ failure assessment score; also remained a significant factor associated with case fatality in the logistic regression model. ::: CRP and the outcome of bacteremia {#s3c} --------------------------------- The CRP level (maximum value on days 1 to 4) did not predict case fatality at any cut-off level in the ROC curve (p = 0.132) ([Figure 1](#pone-0017653-g001){ref-type="fig"}). The AUC^ROC^ for CRP (days 1 to 4 after blood culture) was 0.61 (95% CI 0.49--0.73). Thus, the optimal cut-off for CRP in detecting fatal disease could not be determined. Discussion {#s4} ========== The results presented here show high PTX3 values during the first days after diagnosis to be independently associated with case fatality in patients with bacteremia. High PTX3 values were associated with several variables indicative of severe disease. Compared to CRP, a member of the same pentraxin superfamily, PTX3 may act as a more specific prognostic marker in bacteremic patients. To the best of our knowledge, no previous study has investigated PTX3 in patients with blood culture-proven bacteremic infection. The present findings evidence the value of PTX3 in patients with blood culture-proven bacteremia caused by the four microbial organisms most commonly encountered in clinical practice. The study involved patients with non-severe disease and those admitted to the ICU due to severe infection, which enables study of the value of PTX3 in patients with distinct outcomes. In accord with the present findings, a high PTX3 level has previously been shown to be associated with mortality in severe sepsis and septic shock [@pone.0017653-Mauri1] and to be an early indicator of shock in severe meningococcal disease [@pone.0017653-Sprong1]. Furthermore, PTX3 is elevated in critically ill patients and in febrile patients admitted to the emergency room, correlating with severity of disease and infection in these patient groups [@pone.0017653-Muller1], [@pone.0017653-deKruif1]. The present results thus confirm those of previous studies on the prognostic value of PTX3 in infectious diseases. The study shows that PTX3 may be used to predict all variables indicative of severe disease, i.e. hypotension, mechanical ventilation, low platelet count, high SOFA score, need of ICU treatment and renal failure. PTX3 behaves as an acute-phase response protein, its blood levels being low in normal conditions (\<2 ng/ml in humans) but increasing rapidly in inflammatory and infectious conditions [@pone.0017653-Bottazzi1]. In accord with this, the PTX3 values in the present study were high in the acute phase and decreased on recovery. Only the day 1--4 values were included in the final ROC analysis, as the PTX3 values on the first days following blood culture (i.e. diagnosis) are most unlikely to be related to conditions other than bacteremia *per se*. The initially high values already 1 to 2 days after clinical suspicion of bacteremia (after blood culture) may reflect the central role of PTX3 in the first-step innate immune response, acting as a pattern recognition receptor, with subsequent activation of complement cascade and pathogen opsonization. In the present material, the values did not differ significantly between patients with infection caused by the four different culprit organisms. The precise clinical implications of PTX3 in infectious diseases remain elusive. In animal models, PTX3 even been found to protect from endotoxic shock and sepsis, but controversial results have also been published, highlighting the delicate balance among the various mediators which control the inflammatory response [@pone.0017653-Garlanda1], [@pone.0017653-Dias1], [@pone.0017653-Soares1]. There is evidence suggesting that PTX3 may contribute to acute lung injury (ALI) during inflammation, and a correlation between PTX3 expression and the severity of the lung injury has been documented [@pone.0017653-He1]. On the other hand, previous studies have shown that PTX3 is able to up-regulate tissue factor, a critical factor in the pathogenesis of coagulation/fibrinolysis dysregulation in sepsis [@pone.0017653-Mauri1], [@pone.0017653-Sprong1]. Sprong and associates have shown that PTX3 levels correlate significantly negatively with fibrinogen levels in sepsis. PTX3 may thus contribute to the pathological coagulation process in this condition [@pone.0017653-Sprong1]. High PTX3 concentration in severe disease may also reflect the role of pentraxins in the clearance of apoptotic cells. The currently available biomarkers or nonspecific physiologic criteria for the sepsis syndrome or the systemic inflammatory response syndrome (SIRS) do not adequately identify patients who might benefit either from conventional antimicrobial therapies or from therapies targeting specific mediators of inflammation, i.e. recombinant human activated protein C (rhAPC) [@pone.0017653-Marshall1]. Prognosis of patients is important in risk stratification and for efficient use of hospital resources [@pone.0017653-Marshall1], [@pone.0017653-Rhodes1]. The prognostic value of PTX3 as evaluated by ROC curve was better than that for CRP during the first days after diagnosis of bacteremic infection. This finding is in accord with those of a recent study in patients with severe sepsis and septic shock, showing the prognostic superiority of PTX3 over CRP [@pone.0017653-Mauri1]. Although both belong to the same pentraxin family, long pentraxins (i.e. PTX3) differ from short (SAP and CRP) in several respects, including their gene organization, chromosomal localization, cellular source and ligand-recognition and stimuli-inducing ability [@pone.0017653-Garlanda1]. Some limitations must be conceded here. PTX3 values for day 0 (blood culture day) were not available. Previous study in meningococcal disease has suggested that PTX3 may already peak during the first hours after hospital admission, this reflecting disease severity, which may suggest its utility as an early marker [@pone.0017653-Sprong1]. Also *in vivo* studies indicate rapid PTX3 induction following inflammatory stimulus [@pone.0017653-Garlanda1]. However, in the present study several PTX3 measurements per patient were available (median 2 measurements/patient during days 1 to 4), which reduces the possibility of bias compared to single measurement protocols. The present work was not designed to study the effects of antimicrobial therapy on PTX3 levels. Further research should also assess PTX3 levels in different settings, i.e. in patients with SIRS, in trauma patients and in viral infections. The prognostic value of PTX3 should also in subsequent studies be compared to that of procalcitonin. The biological role of PTX3 in bacteremia and sepsis calls for further elucidation. Possible interactions with coagulation process and PTX3 also warrant subsequent studies. Conclusions {#s4a} ----------- In conclusion, PTX3 proved to be a sensitive and specific independent prognostic marker in patients with bacteremia. It may serve as a more specific indicator of severe disease than the other member of the pentraxin superfamily, CRP. PTX3 measurement may offer novel opportunities for the early prognostic stratification of bacteremia patients in order to target various therapeutic interventions. We thank Mrs Sinikka Repo-Koskinen and Mrs Mirja Ikonen for technical assistance. This work was carried out in Tampere University Hospital and in the University of Tampere Medical School, Tampere, Finland. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was financially supported by the Competitive Research Funding of Tampere University Hospital (Pirkanmaa Hospital District), the Finnish Medical Foundation, and the Orion-Farmos Research Foundation. The authors\' work was independent of the funders (the funding sources had no involvement). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: RH JA JS JL RV HH MH JJ. Performed the experiments: JJ MH JA RV. Analyzed the data: JA RH JS HH. Contributed reagents/materials/analysis tools: JJ MH JA. Wrote the manuscript: RH. Participated in the study design and wrote the first draft of the manuscript: RH. Contributed to the design and approved the final version of the manuscript: RH MH JA HH RV JL JJ JS. Responsible for PTX3 measurements: MH JJ. Constructed the ROC and Kaplan Meier survival curves: JA HH. Statistician: HH. Checked the statistical methods used in the study: HH. Analyzed the blood culture results: RV JA. Recruited the study participants: JS JL.

PubMed Central

2024-06-05T04:04:19.830952

2011-3-10

PMC3053379

Introduction {#s1} ============ The major contribution of the mitochondrion to plant growth and development is the production of energy [@pone.0017662-McBride1]. In some higher plants, the economic significance of mitochondria is boosted by its role in the determination of cytoplasmic male sterility (CMS). To outline characteristics of the plant mitochondrial genome and identify CMS-related genes, the mitochondrial genomes of more than 10 fertile or sterile plant species have been sequenced, including *Arabidopsis thaliana* [@pone.0017662-Unseld1], *Beta vulgaris* [@pone.0017662-Kubo1], [@pone.0017662-Satoh1], *Oryza sativa* [@pone.0017662-Notsu1], [@pone.0017662-Tian1], [@pone.0017662-Fujii1], *Brassica napus* [@pone.0017662-Handa1], *Zea mays* [@pone.0017662-Clifton1], [@pone.0017662-Allen1], *Nicotiana tabacum* [@pone.0017662-Sugiyama1], *Triticum aestivum* [@pone.0017662-Ogihara1], *Vitis vinifera* [@pone.0017662-Goremykin1], *Citrullus lanatus,* and *Cucurbita pepo* [@pone.0017662-Alverson1]. Research revealed that unique deleterious genes inserted into the mitochondrial genome confer CMS traits on natural plant germplasm [@pone.0017662-Linke1], [@pone.0017662-He1]. In higher plants, natural CMS has been identified in more than 150 species over the past century [@pone.0017662-Carlsson1]. CMS emergence undoubtedly stems from alteration of the mitochondrial genome [@pone.0017662-Linke1], [@pone.0017662-Carlsson1], but the question is whether this kind of genomic alteration is the result of toxic genes being inserted during the past century or from the substoichiometric shifting of mitochondrial DNA molecules. Research in a few crops has demonstrated that substoichiometric shifting results in the inherent CMS-triggering mitotype dominating mitochondrial genome constituents [@pone.0017662-Janska1]--[@pone.0017662-Feng1], which results in the occurrence of CMS. In the common bean, a progenitor mitochondrial form containing a three-molecule CMS-associated configuration is universally present in the germplasm [@pone.0017662-Janska1], and substoichiometric shifting can explain the emergence of CMS. In cybrids between CMS radish and *B. napus*, mitochondrial genomic components of the two species coexist in stable rapeseed progenies, and the phenomena of substoichiometric shifting has been observed between male fertile and sterile plants [@pone.0017662-Bellaoui1]. In *B. napus,* at least five natural occurrences of CMS have been independently reported [@pone.0017662-Thompson1]--[@pone.0017662-Fu2]. Genetic experiments and restriction mapping have shown that the CMS lines can be classified as two cytoplasmic groups, one including T-CMS and S-CMS [@pone.0017662-Thompson1]--[@pone.0017662-Shiga2], which is commonly referred to as the "nap" cytoplasm present in normal fertile accessions, and the other group as "pol" cytoplasm including pol-CMS, MI CMS, and shaan 2A CMS [@pone.0017662-Fu1]--[@pone.0017662-Fu2], which has been the most widely applied line in the world. During the past few decades, the fact that the grouped CMS accessions from multiple independent findings have the same genetic regulation has been a perplexing issue for researchers [@pone.0017662-Yang1]. In order to identify the mechanism for natural CMS in *B. napus,* the main mitochondrial genome of pol-CMS was sequenced and compared to the sequence of Westar (nap cytoplasm), uncovering structural and evolutionary differences. Based on the mitotype-specific sequences of the mitochondrial genome in *B. napus*, the constituents of the mitochondrial genome were detected by PCR amplification. Results {#s2} ======= The *pol* mitochondrial genome {#s2a} ------------------------------ A single circular mitochondrial genome 223,412 bp (EMBL accession number: FR715249) in size was obtained by shotgun sequencing. Because of heteroplasmy in *B. napus*, we refer to the mtDNA sequence of the fertile rapeseed variety Westar as the *nap* mitotype, in contrast to the polima-derived *pol* mitotype. The *pol* mitotype sequence is 1,559 bp longer than that of the *nap* mitotype, and the G+C content of the two mitotypes is 45.22% and 45.19%, respectively. Sequence alignment showed that 3.63% of *nap* and 4.53% of *pol* is mitotype-specific. The *pol* sequence encodes 34 proteins, three ribosomal RNAs, and 18 tRNAs, constituting 17.34% of the genome ([Figure 1](#pone-0017662-g001){ref-type="fig"}). The sequence includes two near identical copies of *trnH* that differ by only one base pair, whereas only a single copy is present in the *nap* mitotype. Of the 55 genes present in *pol*, 48 have an identical copy in *nap*, implying that the known mitochondrial genes are well conserved in *B. napus*. Five of the six remaining genes in *nap* differ only marginally from their counterparts in *pol* (the exception was *orf224*). The *atp1* sequences differ by one synonymous single nucleotide polymorphism (SNP), as do the *trnC* and *trnE* sequences. One non-synonymous SNP in *cox1* produces a proline to leucine switch, but this has no effect on gene function. Similarly, a single SNP produces a proline to serine switch in *cox2*. The alignment of the *pol* and *nap* versions of *orf224*, which is generally accepted as the determinant of CMS in *pol*, revealed 84% nucleotide similarity and 77% peptide similarity. ::: {#pone-0017662-g001 .fig} 10.1371/journal.pone.0017662.g001 Figure 1 ::: {.caption} ###### Organization of the *pol* mitotype in *B. napus*. A total of 55 genes with known function were identified, including 34 protein-coding genes (red), 3 ribosomal RNA genes (yellow), and 18 tRNA genes (green). Nine (out of 44) putative ORFs encoding at least 150 amino acids are marked in blue. :::  ::: We found that the *pol* mitotype comprises 44 putative open reading frames (ORFs) encoding at least 100 amino acid residues. Of these ORFs, 35 are identical to their equivalents in *nap*, but the other ORFs are different ([Table S1](#pone.0017662.s004){ref-type="supplementary-material"}). *Orf122* and *orf132* are both present in *pol* but not in *nap*, whereas *orf117b* is unique in *nap*. None of these three sequences aligned with any known sequence in GenBank. Thus, the sequences were inferred to be the result of large insert sequences in the mitochondrial genome. *Orf261* in *nap* and *orf265 in pol* partially aligned and were inferred to be generated by the rearrangement of syntenic regions in the mitochondrial genome. Reorganization of the mitochondrial genome {#s2b} ------------------------------------------ Analysis of the syntenic regions of *pol* and *nap* using the bl2seq algorithm suggested that the sequence as a whole consists of 11 syntenic regions (similarity \>99%, length \>2 kb). The orientation of eight of the regions is identical, but in the other three it is reversed ([Figure 2](#pone-0017662-g002){ref-type="fig"}). The syntenic regions are largely discrepant in the two genomic positions, though the genetic sequences are well conserved ([Figure 2](#pone-0017662-g002){ref-type="fig"}). A minimum of five recombination events is required to account for the structural differences between the two mitotypes. ::: {#pone-0017662-g002 .fig} 10.1371/journal.pone.0017662.g002 Figure 2 ::: {.caption} ###### Alignment of the *nap* and *pol* mitotype genomes. The numbers refer to the syntenic regions with the *pol* mitotype sequence as a reference. Comparisons between *nap* (horizontal axis) and *pol* (vertical axis) indicated that the nucleotide sequences of the syntenic region are highly conserved, but the syntenic orders and directions were evolutionally rearranged. :::  ::: Repeated sequences in the *pol* mitotype were analyzed. A pair of large repeats 2427 bp in size identified in *pol* are also found in *nap* [@pone.0017662-Handa1] and related to the formation of the multipartite structure of the *Brassica* mitochondrial genome, including one master circle and two smaller subgenomic circles through homologous recombination [@pone.0017662-Palmer1], [@pone.0017662-Palmer2]. The *pol* sequence also contains multiple copies of various 30--500 bp repeats; applying a 90% sequence similarity criterion to define a repeat, these sequences account for 4.75% of the genome and are arranged in both direct and inverse orientation. The repeats probably provide the opportunity for mtDNA sequence evolution as illustrated in [Figure 3](#pone-0017662-g003){ref-type="fig"}. The pair of inverted 146 bp repeats (F) on either side of syntenic region 10 may account for the changed orientation of this region in the two mitotypes ([Figure 3A](#pone-0017662-g003){ref-type="fig"}). A pair of inverted 233 bp repeats (H) is present at the end of syntenic regions 8 and 6 in *pol,* whereas the whole of regions 8 and 7 is inverted in *nap* following rearrangement, accompanied by the deletion of one short repeat ([Figure 3B](#pone-0017662-g003){ref-type="fig"}). Short repeats seem to be associated with inserted or deleted sequences of mtDNA because the 50 bp sequence (K) between syntenic regions 4 and 5 in *pol* is present as two copies in the *nap* mitotype, along with the insertion of an anonymous 44 bp sequence ([Figure 3C](#pone-0017662-g003){ref-type="fig"}). Short repeats providing cues to the genome reorganization may correspond to recombination hotspots in plant mitochondrial genomes [@pone.0017662-Scotti1]. ::: {#pone-0017662-g003 .fig} 10.1371/journal.pone.0017662.g003 Figure 3 ::: {.caption} ###### Syntenic reordering and structural genomic changes caused by short repetitive sequences. The numbers refer to the syntenic regions as in [Figure 2](#pone-0017662-g002){ref-type="fig"}. The orientation of the sequence is shown by an arrow. (**A**) A pair of inverted repeats (in red, F) may have induced the re-orientation of region 10. (**B**) In *pol*, a pair of inverted repeats (in green, H) is located at the end of linked regions 8 and 7, but in *nap*, this region is re-orientated and contains a small deletion of a sequence of unknown origin and fewer repeats due to an overlap of H at the terminus of region 6. (**C**) A 50 bp sequence (in blue, K) is located between syntenic regions 4 and 5 in *pol*, whereas two copies of K are present at this location in *nap*, accompanied by a 44 nucleotide insertion of unknown origin. :::  ::: A more detailed analysis of the *orf224* region, which is reported to be the CMS-associated region for *pol* [@pone.0017662-Singh1], was carried out. The *pol* and *nap* mitotypes differed by at least two recombination events ([Figure 4](#pone-0017662-g004){ref-type="fig"}). The first recombination event, within *orf265* of pol mitotype, broke the region into three parts. In the *nap* mitotype, the frontal 78 bp and terminal 467 bp of the *orf265* sequence were separated, together with regions 3 and 4, respectively, whereas a 253 bp segment in the middle of the ORF was lost in the course of the rearrangement. The 467 bp *orf265* sequence was then attached to a neighboring sequence, generating the *nap* form of *orf261.* The second recombination event occurred at the end of region 2. Another putative gene, *orf188,* in the *nap* mitotype appeared because of the relocation of region 2. Only 90 bp of the *nap orf188* sequence remained in *pol*, probably due to evolutionary events that resulted in a sequence loss. Summarily, this region is more frequently rearranged than other regions in genomic sequences. ::: {#pone-0017662-g004 .fig} 10.1371/journal.pone.0017662.g004 Figure 4 ::: {.caption} ###### Structural polymorphism in the region surrounding *orf224*. The numbers refer to the syntenic regions as in [Figure 2](#pone-0017662-g002){ref-type="fig"}. At least two recombination events are inferred in this region. One of these, located within *orf265* in *pol*, splits this sequence into three segments, and in nap the terminus of *orf265* generates *orf261* by fusing to a neighboring sequence; the second event occurs between regions 2 and 3 resulting in the formation of *orf188,* a combination of the terminal 90 bp of region 2 and a new neighboring sequence in the *nap* mitotype. :::  ::: Substoichiometrically different *pol* and *nap* mitotypes coexist in *B. napus* {#s2c} ------------------------------------------------------------------------------- In order to demonstrate our inference about the coexistence of *pol* and *nap* mitotypes in *B. napus* cytoplasm, we designed 11 pairs of mitotype-specific PCR primers according to the two entire mitotype sequences. Using mtDNA extracted from pol-CMS line NH3A and its nap cytoplasm maintainer line NH3B as templates, PCR amplification showed that all primers amplified fragments regardless of the line ([Figure 5](#pone-0017662-g005){ref-type="fig"}). The profile generated by the multiple-primer P10 consisted of two bands of contrasting intensity([Figure 5](#pone-0017662-g005){ref-type="fig"}). The results indicated that the mitochondrial population in both cytotypes was mixed. In addition, 14 materials with different origins were used in the PCR reactions. PCR with the P1 (*nap*-specific) and P5 (*pol*-specific) primers indicated that both primers were effective at amplifying all of the templates ([Figure S1](#pone.0017662.s001){ref-type="supplementary-material"} and [S2](#pone.0017662.s002){ref-type="supplementary-material"}), demonstrating the coexistence of the two mitotypes. Sequencing of the various amplicons generated from the 11 primer pairs confirmed that all PCRs amplified their target sequences. The PCR profiles successfully established the coexistence of the two mitotypes within one *B. napus* plant. However, these results do not mean that the main genomes in the cytoplasm of different genetic lines are the same. Under identical PCR conditions with the same controlled mtDNA template concentration, the amplification was greater for the pol-CMS line than its maintainer when *pol* mitotype-specific primers were used ([Figure 5](#pone-0017662-g005){ref-type="fig"}). On the other hand, when *nap* mitotype-specific primers were used, the amplification was greater in the maintainer line compared to the CMS line. These results indicated that differences exist at the substoichiometric level of mtDNA molecules between the pol-CMS line and its maintainer. The main genome in the pol-CMS plant is the *pol* mitotype, whereas the main genome in the nap cytoplasm is *nap*. In other words, the cytoplasmic difference may be explained by the substoichiometric complexity of the rapeseed mitochondrial genome, such as that observed in the common bean [@pone.0017662-Janska1], [@pone.0017662-ArrietaMontiel1]. These interesting and surprising truths upset the conceptual framework of the rapeseed mitochondrial genome in regards to the pol and nap cytotypes. ::: {#pone-0017662-g005 .fig} 10.1371/journal.pone.0017662.g005 Figure 5 ::: {.caption} ###### PCR amplification of the pol-CMS line and its maintainer. The *pol* and *nap* mitotypes are both present in pol-CMS line NH3A (**A**) and its maintainer NH3B (**B**), but are substoichiometrically differentiated in these cytotypes. Primer pairs P1 and P2 targeted the *nap* mitotype-specific sequences *orf117b* and *orf222*, respectively, whereas P3-9 and P11 targeted *pol* mitotype-specific sequences. P10 amplified two distinct fragments 841 bp (*pol*-specific) and 226 bp (*nap*-specific) in size. All PCR reactions used the 10 ng of mtDNA template. All primer pairs were able to amplify target fragments in both lines, but amplification differed in the two lines. The PCR assay indicated that the *nap* and *pol* mitotypes coexisted in rapeseed, but the content was substoichiometrically different in the two cytoplasm types. M: DNA ladder. :::  ::: *Orf222* to *orf224* copy number ratio in *B. napus* {#s2d} ---------------------------------------------------- The ratio of *orf224* and *orf222* copy numbers, which are popularly accepted as CMS-related genes, was estimated using a TaqMan qPCR platform. The results indicated that the ratio was cytoplasm-dependent. In pol-CMS lines, the *orf224* copy number was greater than that of *orf222*, but the predominant gene was *orf222* in maintainer lines and other male fertile cultivars carrying *nap* cytoplasm ([Table 1](#pone-0017662-t001){ref-type="table"}). These quantifying results are consistent with the PCR assays mentioned above. Obviously, some substoichiometric shifting of *orf224/orf222* occurred between pol-CMS lines and their maintainers with *nap* cytoplasm. Substoichiometric shifting from *orf222* to *orf224* (or from *orf224* to *orf222*) in different materials represents a cytoplasmic difference, and the substoichiometric shifting appears to be the determinant of the male fertility/sterility phenotype in *B. napus*. ::: {#pone-0017662-t001 .table-wrap} 10.1371/journal.pone.0017662.t001 Table 1 ::: {.caption} ###### Copy number ratio of *orf222* and *orf224* in different accessions of *B. napus.* ::: {#pone-0017662-t001-1} *pol* cytoplasm *orf224/orf222* *nap* cytoplasm *orf222/orf224* ----------------- ----------------- ----------------- -------------------------------------------- NH3A 15.00 NH3B 1097.71 NH12A 1929.68 NH12B 7479.48 NH15A 287.25 NH15B 15771.91 NH18A 88.48 NH18B 1630.26 NH21A 15652.38 NH21B 4401.30 NH923A 18.39 NH923B 445.26 Westar 11337.91 Tapidor 36.17 Huaiyin 16 155209.38 Zheshuang 72 135117.61 Westar in 2010 164059.11[\*](#nt101){ref-type="table-fn"} Tapidor in 2010 746.43[\*](#nt101){ref-type="table-fn"} \* repeated determination in 2010. The rest were determined in 2009. ::: However, we also noted variation in the copy number ratio among cultivars sharing the same cytotype. In cultivars carrying *pol* cytoplasm, the *orf224*:*orf222* ratio ranged from 15 to \>15,000; in cultivars with *nap* cytoplasm, the *orf222*:*orf224* ratio ranged from 36 to \>150,000 ([Table 1](#pone-0017662-t001){ref-type="table"}). The abundance of sublimons (i.e. alternative genomes) was roughly two to six orders of magnitude less than that of the prevalent mitotype. This dramatic variation in the substoichiometric ratio among accessions sharing the same cytoplasm may be under nuclear control. The *orf222*:*orf224* copy number ratio in the Westar cultivar was consistently higher than that of the Tapidor cultivar, though the ratio was not stable from year to year ([Table 1](#pone-0017662-t001){ref-type="table"}), presumably as a result of an interaction with the environment. Sequence evolution in *B. napus* {#s2e} -------------------------------- A sequence comparison of the *pol* and *nap* mitotypes identified 197 SNPs (102 transitions, 95 transversions), equivalent to a polymorphism rate of 8.9 bp per 10,000 bases. As many as 104 of the SNPs were located within the *orf224*/*orf222* genes. Only six SNPs were identified in other genes of known function, indicating a much higher level of sequence conservation. Although reconstructing the evolution of the *pol* and *nap* mitotypes is not possible, the former is likely more primitive. The evidence for this conclusion is that the structure of the *pol* mitotype is closely related to that of a 124 kb *Brassica rapa* mitochondrial BAC clone (GenBank accession number AC172860.1) that is the likely ancestor of *B. napus* ([Figure 6](#pone-0017662-g006){ref-type="fig"}). Another reason is that restorer in the natural germplasm population is sparse for the pol-CMS line, but popular for the nap-CMS line, implying that *nap* cytoplasm is more accommodated evolutionarily than *pol* cytoplasm. ::: {#pone-0017662-g006 .fig} 10.1371/journal.pone.0017662.g006 Figure 6 ::: {.caption} ###### Alignment of the *B. rapa* mitochondrial segment with the two *B. napus* mitotype sequences. The 124 kb *B. rapa* mitochondrial segment was plotted on the horizontal axis against the *pol* subgenome (**A**) or *nap* subgenome (**B**) on the vertical axis. :::  ::: Discussion {#s3} ========== The coexistence of mitotypes has been discovered by only a few gene detection experiments in wheat [@pone.0017662-Hattori1], maize [@pone.0017662-Small1], pearl millet [@pone.0017662-Feng1], and cybrids [@pone.0017662-Bellaoui1], and is probably common in higher plants. In the common bean, coexisting mitochondrial genome molecules have been systematically demonstrated through restriction mapping. We showed here for the first time using a combination of genome sequencing and PCR assays that the *pol* and *nap* mitotypes coexist within one *B. napus* plant. The mitochondrial genome in higher plants is very stable, maintaining plant survival in natural environments to achieve its important functions in living activities, so it is not probable that so many cases of CMS emerge only by evolution of the mitochondrial genome over one century. Substoichiometric shifting of coexisting mitotypes in plants may be an explanation of the phenomena related to CMS findings. The mitotype diversification cannot be explained by evolutionary events occurring over a short time, and the mitotypes are probably evolutionary products of a long time-span and coexisting in the mitochondrial population. The presence of mitochondria carrying alternative, rearranged genomes (sometimes referred to as "sublimons") has been associated with phenotypic switching, tissue differentiation, cytotype evolution, and various nuclear-cytoplasmic interactions in higher plants [@pone.0017662-Woloszynska1]. The means by which sublimons are maintained, propagated, and transmitted largely remain unexplored. However, the sublimons do appear to be stable and, thus, able to affect pollen fertility via substoichiometric shifting. We provided evidence that the *B. napus* mitochondrial genome includes at least two distinct mitotypes that are present in variable ratios in both CMS and male fertile cultivars. *B. napus* cultivars sharing the same cytotype can vary considerably with respect to the copy number ratio; therefore, the value of the copy number ratio is controlled, at least to some extent, by the nuclear genotype. In the common bean, a dominant allele in a nuclear gene can act to limit the copy number of mtDNA containing the CMS-associated *pvs-orf239* sequence. Environmental factors also exert an influence on this ratio [@pone.0017662-Janska1]. Both the nap and pol CMS trait was selected during the course of conventional breeding, apparently because in a particular confluence of nuclear genotype and environment, the copy number of the critical sublimon amplified sufficiently to exert a phenotypic effect ([Figure S3](#pone.0017662.s003){ref-type="supplementary-material"}). Our results, as well as previous conclusions in plant species [@pone.0017662-ArrietaMontiel1], show that the conditions for the substoichiometric shifting of mitotypes are natural and common. Generally, when the alternative mitotype with malicious genes is accompanied by particular nuclear genotypes, the plants may manifest CMS phenotypes [@pone.0017662-Janska1]. The model for the natural emergence of pol-CMS is shown in [Figure S3](#pone.0017662.s003){ref-type="supplementary-material"} and probably applies to other plants. In higher plants, mtDNA molecules exist in many forms [@pone.0017662-Woloszynska1], but a multiple partite structure may still not be denied [@pone.0017662-Andre1]. Based on the two 2.4 kb sequence repeats in the mitotypes, the *pol* mitotype is inferred as containing one master circle accompanied by two smaller circles 86.2 kb and 137.1 kb in size ([Figure 7B](#pone-0017662-g007){ref-type="fig"}), which is different from the *nap* mitotype ([Figure 7A](#pone-0017662-g007){ref-type="fig"}). Assuming that the two mitotypes are present in an equimolar amount in an oilseed rape cell, the mitochondrial genome as a whole can be expected to form a six-circle structure comprising two master and four smaller circles. ::: {#pone-0017662-g007 .fig} 10.1371/journal.pone.0017662.g007 Figure 7 ::: {.caption} ###### The tripartite mitochondrial genomic structure of the mitotypes. The numbers refer to the syntenic regions as in [Figure 2](#pone-0017662-g002){ref-type="fig"}. Highly or completely homologous regions are indicated by color. (**A**) *nap;* (**B**) *pol*. R: 2.4 kb repetitive sequence. :::  ::: A major difference between *nap* and *pol* mitotypes is localized sequence rearrangement. The short repeated sequences in higher plant mitochondria are usually inactive and play a central role in irreversible recombination to produce a new stable mitochondrial genome structure [@pone.0017662-Andre1]. The *pol* mitotype also contains numerous short repeated sequences, including direct repeats and inverted repeats; those located at the edge of rearranged syntenic regions may represent the vestiges of past recombination events that restructured the genome. *Nap* and *pol* mitotypes also have significant differences at the nucleotide level. In the pol-CMS of *B. napus*, expression analysis has shown that the co-transcription of *orf224* and *atp6* alters the expression of *atp6*, which is associated with CMS [@pone.0017662-Singh1], [@pone.0017662-LHomme1]. In addition to the sequence polymorphism within *orf224*, both *orf118* and *orf265* differ between *pol* and *nap*, whereas *orf122* and *orf132* are *pol*-specific. Whether the *pol*-specific ORFs play any determining role in CMS is still unknown. Materials and Methods {#s4} ===================== Plant material {#s4a} -------------- Six pol-CMS lines, their corresponding maintainers, and four commercial cultivars were analyzed for mitotype constitution and copy number: NH3A/B (CMS line/maintainer), NH12A/B, NH15A/B, NH21A/B, NH18A/B, and NH923A/B (bred by the present authors). The four commercial cultivars were Westar (from Canada), Tapidor (from France), and Huaiyin 16 and Zheshuang 72 (from China). The source of mitochondrial DNA for genome sequencing was pol-CMS line NH12A. These plants were planted in Jiangpu experimental station of Nanjing Agricultural University in the Autumn of 2008, harvested in 2009, and mitotype copy number determined in the Summer of 2009 with their abundant seeds. The determinations for Westar and Tapidor were repeated in 2010. Isolation of mtDNA {#s4b} ------------------ The mtDNA was purified from 7-day-old etiolated seedlings germinated on an artificial medium in the dark at 25°C. The DNA isolation method was a modification of the method used in tobacco [@pone.0017662-Bland1]. All of the extraction steps were performed at 4°C unless stated otherwise. A 250 g sample of seedling material was homogenized in 1000 ml buffer (0.5 M sucrose, 50 mM Tris-HCl, 10 mM EDTA, 1% bovine serum albumin, 5 mmol/L *β*-mercaptoethanol, 1.5% polyvinyl-pyrrolidone, pH 7.5). The homogenate was filtered through a layer of cheesecloth and two layers of miracloth, and centrifuged at 3000 g for 15 min. The resulting supernatant was then centrifuged at 17,000 g for 15 min and the pellet re-suspended in buffer A (0.5 M sucrose, 50 mM Tris-HCl, 10 mM MgCl~2~, and 25 µg/ml DNase (Roche 104159), pH 7.5) at a 5∶1 (g/ml) ratio and held at room temperature for 1 h. The suspension was then centrifuged at 18,000 g for 20 min and the pellet re-suspended in 30 ml buffer B (0.6 M sucrose, 10 mM Tris-HCl, 20 mM EDTA, pH 7.5). The suspension was centrifuged once more at 18,000 g for 20 min. Finally, the pellet was re-suspended in the same buffer and layered onto a step gradient consisting of 1.45 M and 1.2 M sucrose solution containing 10 mM Tris-HCl and 20 mM EDTA (pH 7.5). The gradient was centrifuged at 72,000 g for 90 min. Purified mitochondria were removed from the 1.45 M-1.2 M interphase, diluted with two volumes of buffer C (10 mM Tris-HCl, 20 mM EDTA, pH 7.5), and centrifuged at 18,000 g for 20 min. The final pellet was gently re-suspended in 10 ml lysis buffer (50 mM Tris--HCl, 10 mM EDTA, 1% SDS, and 200 µg/ml proteinase K(Sigma)) and incubated at 37°C for 3 h. Ammonium acetate was added to reach a concentration of 0.8 M, and the preparation was extracted twice in phenol/chloroform. The mtDNA was precipitated overnight in 2.5 volumes of 100% ethanol and centrifuged at 18,000 g for 20 min. The pellet was washed twice with 70% ethanol, and then air-dried and re-suspended in sterile distilled water. Sequencing strategy {#s4c} ------------------- The mtDNA was sonicated randomly and size-fractionated by agarose gel electrophoresis. The 1.6--4 kb fraction was cloned into the pUC19 vector. The inserts were subjected to cycle sequencing (BigDye terminator v3.1 Cycle Sequencing kit, Applied Biosystems, USA). A set of 2,592 reads was obtained, which is equivalent to 11x coverage of the *B. napus* mitochondrial genome. The individual reads were assembled into five contigs using Phred-Phrap software (<http://www.phrap.org/>). Gaps between the contigs were filled by PCR. Detection of coexisting mitotypes {#s4d} --------------------------------- A set of 11 PCR primer pairs ([Table S2](#pone.0017662.s005){ref-type="supplementary-material"}) was based on the *pol* and *nap* mitotype sequences, of which eight (P3-9, P11) are *pol*-specific. P1 targeted *orf117b* and P2 *orf222*, both of which are *nap*-specific. P10 was a multiple-primer designed to amplify both *pol* and *nap* mitotype templates, specifically generating an 841 bp *pol*-specific and 226 bp *nap*-specific amplicon. The mtDNA samples were amplified using the primers in 25 µl PCR reactions containing 50 µM dNTPs, 0.2 µM of each primer, 1 µl template, 2.5 µl Taq polymerase buffer, and 1 U Taq polymerase (TIANGEN, China). The PCR was programmed as follows: an initial step of 94°C/5 min, then 30 cycles of 94°C/30 s, 53°C--60°C/30 s, and 72°C/90 s, followed by a final amplification step of 72°C/10 min. All PCR products were sequenced directly. Determination of copy number ratios {#s4e} ----------------------------------- A quantitative PCR platform (TaqMan) was used to estimate the copy number of specific mitotypes. The sequences of the relevant primers and TaqMan probe were designed from the *orf224* and *orf222* sequences, which were specific for the *pol* and *nap* mitotypes, respectively ([Table S3](#pone.0017662.s006){ref-type="supplementary-material"}). The probe was 5′-labeled with FAM and 3′-labeled with TAMRA. All primers and probes were obtained from Invitrogen Biotechnology (China). Each 20 µl PCR contained 10 µl premix EX Taq (2x) (TaKaRa), 0.3 µM of each primer, 0.2 µM TaqMan probe, 0.4 µl Rox reference dye II, and 5 µl template. The two step PCR was carried out on an ABI PRISM 7500 Fast sequence detection system (Applied Biosystems) using the following cycling program: 95°C/50 s, then 40 cycles of 95°C/3 s, 58.5°C/30 s. The analysis of each mtDNA sample was performed in triplicate. In order to check the amplification efficiency (E) of each primer pair/probe, validation experiments were performed using a 10-fold serial dilution of mtDNA. The E of each primer pair/probe was 99.7% and 100.1%, respectively. The copy number ratio of *orf224* and *orf222* was determined from the *Ct* values and E. Sequence analysis {#s4f} ----------------- The NCBI database was searched using the blast network service and bl2seq (<http://www.ncbi.nlm.nih.gov/>). A tRNA gene search was carried out using the tRNA scan-SE service (<http://www.genetics.wustl.edu/eddy/tRNAscan-SE/>). Small dispersed repeats were defined as sequences of identical length (30--500 bp) that were found more than once in the genome and at least 90% identical. SNPs were identified by applying default settings in Mauve software [@pone.0017662-Darling1]. Peptide and nucleotide sequences were aligned by ClustalX 1.81 using default settings [@pone.0017662-Thompson2]. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### **PCR analysis of a panel of male fertile and pol-CMS cultivars using P1**. Analysis was based on the nap-specific primer pair P1, which targeted *orf117b*, but amplification occurred in both nap- and pol-CMS lines. Huaiy16: Huaiyin16; Zhesh72: Zheshuang72; M: DNA ladder. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### **PCR analysis of a panel of male fertile and pol-CMS cultivars using P5.** Analysis was based on the pol-specific primer pair P5. Amplification of the target fragment occurred in both nap- and pol-CMS lines. Huaiy16: Huaiyin16; Zhesh72: Zheshuang72; M: DNA ladder. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S3 ::: {.caption} ###### **The coexistence and stoichiometric shifting of the** ***nap*** **and** ***pol*** **mitotypes in** ***B. napus.*** *Nap* and *pol* mitotypes coexisted in *B. napus*, but the relative content of the two mitotypes was different in the two cytoplasm types. In the nap cytotype, *nap* is the prevalent mitotype and *pol* is present as a sublimon. In the pol cytotype, *pol* is the prevalent mitotype and *nap* is present as a sublimon. Under certain conditions, including nuclear genotypes and environmental stress, sublimons may be amplified and accumulate to take over the role of the main genome. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S1 ::: {.caption} ###### **Different ORFs between** ***nap*** **and** ***pol*** **mitotypes.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S2 ::: {.caption} ###### **PCR primer sequences.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S3 ::: {.caption} ###### **Primers and probes for TaqMan qPCR.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: The authors wish to thank Shanghai Majorbio Bio-pharm Biotechnology Company (China) for their help with the shotgun sequencing. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was supported by the National Basic Research Program of China (973 Program) (2011CB109300); National Natural Science Foundation of China (30970289); National Key Technology R&D Program (2010BAD01B02) in China; and the Ministry of Agriculture of China for Transgenic Research (2009ZX08009-040B). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: JC RG. Performed the experiments: JC SC TD. Analyzed the data: JC RG HZ. Contributed reagents/materials/analysis tools: HX. Wrote the paper: RG JC HX.

PubMed Central

2024-06-05T04:04:19.833597

2011-3-10

PMC3053380

Introduction {#s1} ============ During our studies of the antimicrobial activity of human milk, we identified a complex of alpha-lactalbumin (ALA) and oleic acid that induces apoptosis in tumor cells, without affecting healthy, differentiated cells [@pone.0017717-Hakansson1], [@pone.0017717-Svensson1]. The same complex showed strong bactericidal activity against specific pathogens of the oral cavity and respiratory tract, with the highest activity against the gram-positive organism *Streptococcus pneumoniae* [@pone.0017717-Hakansson2]. In the complex, designated HAMLET for "human alpha-lactalbumin made lethal to tumor cells", alpha-lactalbumin is present in a partially unfolded conformation that is stabilized under physiological conditions by C18:*n* cis unsaturated fatty acids, the most prevalent fatty acids in human milk [@pone.0017717-Svensson1], [@pone.0017717-Svensson2]. The native, folded form of ALA, with lactose synthase activity, has no tumoricidal or bactericidal effect, however [@pone.0017717-Hakansson1], [@pone.0017717-Hakansson2]. Programmed cell death or apoptosis in eukaryotes is executed by the consecutive activation of specific biochemical pathways that produce a dying cell, with typical morphology, such as cell shrinkage, membrane blebbing, chromatin condensation, as well as distinct DNA fragmentation [@pone.0017717-Taylor1]. This type of programmed cell death represents an important mechanism to regulate tissue function and homeostasis in multicellular organisms [@pone.0017717-Majno1] but is also used by unicellular eukaryotes to regulate their optimal adaptation to their environment [@pone.0017717-Ameisen1]. Although, primitive forms of programmed cell death and terminal differentiation have been described in prokaryotes also [@pone.0017717-Lewis1], there has been no information to date to suggest that bacterial death show similarities to eukaryote apoptosis. In this study, we demonstrate that HAMLET triggers DNA fragmentation, as well as morphological and biochemical changes in *S. pneumoniae,* resembling apoptosis in tumor cells. We also describe similarities in the responses to HAMLET between mitochondria and bacteria. Our studies suggest for the first time that bacteria contain basic cell death programs that are similar to those involved in eukaryotic cell apoptosis. Results {#s2} ======= 1. HAMLET kills tumor cells and bacteria {#s2a} ---------------------------------------- Cell death in response to HAMLET was quantified in parallel in Jurkat leukemia cells and *S. pneumoniae* ([Fig. 1A](#pone-0017717-g001){ref-type="fig"}). HAMLET killed Jurkat leukemia cells in a dose-dependent manner with 50% death occurring at a concentration of 200 µg/ml after 6 hours. At the same concentration, HAMLET reduced the viability of *S. pneumoniae* by more than 6 log~10~ within 1 hour, with complete eradication of the inoculum (8 log~10~) obtained at 250 µg/ml ([Fig. 1A](#pone-0017717-g001){ref-type="fig"}). These concentrations are within the physiological range, as the concentration of ALA is especially high (1,000--2,000 µg/ml) in human milk [@pone.0017717-Heine1]. The effect appeared to be general among pneumococci, as over 25 pneumococcal strains of nine different capsule types were equally sensitive to HAMLET-induced death ([Table 1](#pone-0017717-t001){ref-type="table"}). In accordance with earlier results [@pone.0017717-Svensson1], [@pone.0017717-Hakansson2] the calcium bound holo-form of ALA had no tumoricidal or bactericidal activity even at concentrations up to 10,000 µg/ml ([Fig. 1A](#pone-0017717-g001){ref-type="fig"}). ::: {#pone-0017717-g001 .fig} 10.1371/journal.pone.0017717.g001 Figure 1 ::: {.caption} ###### Bacterial and tumor cell death induced by HAMLET. A\) Jurkat cells and *S. pneumoniae* D39 were incubated with increasing concentrations of HAMLET or human alpha-lactalbumin (ALA) and viability was monitored after 6 h or 1 h of incubation for Jurkat and bacterial cells, respectively. Viability of Jurkat cells are presented on the right Y axis in per cent viable cells in the suspension as determined by trypan blue staining and viability of bacteria are presented on the left Y axis as colony forming units (CFUs) per ml suspension (detection limit in the assay was 10^1^ CFU/ml). ALA (hatched lines) did not kill any of the organisms whereas HAMLET (solid lines) killed both Jurkat and bacterial cells in a dose-dependent manner. The data represent the mean of three individual experiments with standard deviation error bars. B) Role of bacterial virulence factors in HAMLET-induced killing of *S. pneumoniae.* Pneumococcal strains lacking capsule, Pneumococcal surface proteins A and C (PspA and PspC), autolysin (LytA), pneumococcal surface adhesin A (PsaA), or dihydrolipoamide dehydrogenase (DLDH), all associated with virulence, were treated with 50, 120 and 200 µg/ml of HAMLET for 1 hour at 37°C and viable organisms were determined by plating organisms on solid agar and counting colony forming units after overnight growth. The graph depicts the mean of three experiments. There was no significant difference in sensitivity related to lack of these virulence factors. :::  ::: ::: {#pone-0017717-t001 .table-wrap} 10.1371/journal.pone.0017717.t001 Table 1 ::: {.caption} ###### Strains of *S. pneumoniae* tested for HAMLET-sensitivity. ::: {#pone-0017717-t001-1} Strain name Capsule type Bacterial viability (CFU/ml) ------------- -------------- ------------------------------ --------- --------- L82006 1 2×10^8^ 9×10^4^ 8×10^2^ D39 2 3×10^8^ 3×10^4^ 5×10^2^ DBL2 2 9×10^7^ 1×10^5^ 7×10^2^ WU2 3 2×10^8^ 7×10^3^ \<10 A66 3 2×10^8^ 4×10^4^ 8×10^2^ 3JYP2670 3 1×10^8^ 7×10^4^ 4×10^2^ ATCC 6303 3 1×10^8^ 2×10^4^ 9×10^2^ EF10197 3 3×10^8^ 1×10^5^ 2×10^3^ TIGR4 4 5×10^8^ 8×10^4^ 5×10^2^ EF3296 4 5×10^8^ 6×10^4^ 3×10^2^ L81905 4 8×10^7^ 2×10^4^ 8×10^2^ DBL5 5 2×10^8^ 3×10^4^ 7×10^2^ BG9273 6A 4×10^8^ 5×10^4^ 1×10^3^ BG9163 6B 5×10^8^ 3×10^4^ 6×10^2^ BG7322 6B 3×10^8^ 1×10^4^ 4×10^2^ L82016 6B 3×10^8^ 2×10^3^ \<10 BG30-11 6B 2×10^8^ 9×10^4^ 5×10^2^ EF3559 14 3×10^8^ 1×10^5^ 8×10^2^ EF1488 15A 3×10^8^ 6×10^4^ 4×10^2^ L82013 19F 3×10^8^ 8×10^3^ \<10 EF3030 19F 5×10^8^ 1×10^4^ \<10 EF10175 19F 4×10^8^ 7×10^3^ \<10 ATCC 49619 19F 1×10^8^ 4×10^4^ 2×10^2^ BG8826 23F 9×10^7^ 9×10^3^ \<10 ::: To examine if death in response to HAMLET was modified by the virulence of the *S. pneumoniae* strain, we used the wild-type strain D39 and deletion mutants lacking each of five virulence-associated pneumococcal surface molecules (capsule, pneumococcal surface proteins A or C, pneumococcal surface adhesin A, and dihydrolipoamide dehydrogenase [@pone.0017717-Hakansson3]--[@pone.0017717-Campos1]. These molecules are known to interact with host targets, including some molecules present in human milk and to bind and inactivate bactericidal host defense molecules such as complement proteins and lactoferrin. Wild type or mutant strains were exposed to HAMLET and the loss of viability was quantified by viable counts. No differences were recorded, suggesting that these molecules are not involved in HAMLET\'s bactericidal activity ([Fig. 1B](#pone-0017717-g001){ref-type="fig"}). HAMLET-induced killing of *S. pneumoniae* was accompanied by lysis of the bacterial cells, detected by a parallel decrease in viability and turbidity of the bacterial suspension ([Fig. 2A](#pone-0017717-g002){ref-type="fig"}). To address the mechanism of bacterial lysis, autolysin-negative mutants in *S. pneumoniae* D39 were exposed to HAMLET. Bacterial lysis was autolysin (LytA)-dependent but LytA-negative bacteria were killed as efficiently as wild type bacteria ([Fig. 2B](#pone-0017717-g002){ref-type="fig"}), indicating that lysis was independent and occurred downstream of the initiation of HAMLET-induced death. In this regard, HAMLET\'s activity was similar to that of the well-characterized bile salt deoxycholate (DOC), which was used as a positive control [@pone.0017717-Tomasz1]. ::: {#pone-0017717-g002 .fig} 10.1371/journal.pone.0017717.g002 Figure 2 ::: {.caption} ###### Role of autolysin in HAMLET-induced lysis and killing of *S. pneumoniae.* Pneumococcal strains D39 and AL2 (D39 **Δ** *lytA*) were treated with 150 µg/ml of HAMLET (HL) or 0.1% of the bile salt deoxycholate (DOC) over 90 minutes. As pneumococci are bile salt sensitive streptococci and lysis from bile salts is LytA dependent, deoxycholate was added as a LytA-dependent control. *A*) The optical density of the suspension was monitored at 600 nm every 15 minutes to assess the lysis of bacteria. *B*) After 10 minutes and 60 minutes bacteria were serially diluted and plated on blood agar overnight to determine viable colony forming units (detection limit in the assay was 10^2^ CFU/ml). The data represent the mean of three individual experiments with standard deviation error bars. HAMLET and DOC killed both strains equally well, but lysis was only observed in the autolysin positive D39 strain. :::  ::: 2. Apoptosis-like morphology of HAMLET-killed pneumococci {#s2b} --------------------------------------------------------- To examine if HAMLET interacts with analogous targets in tumor cells and bacteria we first examined apoptosis-associated phenotypes in A549 carcinoma cells and pneumococci treated with HAMLET. Chromatin condensation and DNA fragmentation, two morphological hallmarks of apoptosis, were investigated by microscopy and gel electrophoresis, respectively. HAMLET-treated tumor cells displayed typical apoptotic morphology with nuclear fragmentation and chromatin condensation ([Fig. 3A](#pone-0017717-g003){ref-type="fig"}). A similar change in DNA morphology was observed in pneumococci as well. All pneumococci treated with HAMLET displayed condensed chromatin, seen as a reduction of the DNA volume. Associated with the DNA condensation, the chromatin was no longer homogenously stained in the bacterial cells but displayed fragmentation resulting in a punctate or patched staining pattern ([Fig. 3A](#pone-0017717-g003){ref-type="fig"}, see arrows). ::: {#pone-0017717-g003 .fig} 10.1371/journal.pone.0017717.g003 Figure 3 ::: {.caption} ###### Chromatin condensation and fragmentation induced by HAMLET in tumor cells and *S. pneumoniae*. *A*) Chromatin morphology in HAMLET-treated A549 carcinoma cells and *S. pneumoniae* AL2 (D39 **Δ** *lytA*) cells after 6 h and 1 h of incubation, respectively. All cells were fixed in 4% paraformaldehyde solution and stained with 300 nM DAPI to visualize DNA. Arrows in the bacterial image indicated chromatin that is condensed and fragmented. Inset shows a 2.5X additional magnification of a representative bacterium from each image. *B*) High molecular weight DNA fragments were induced by HAMLET in A549 carcinoma cells (Tumor cells) and *S. pneumoniae* D39 cells and detected after 6 h and 1 h of incubation, respectively. The concentration of HAMLET used for each lane is presented in µg/ml. Increasing concentrations of HAMLET resulted in accumulation of DNA fragments over time. Low molecular weight oligonucleosomal DNA fragments were not observed in either tumor cells or bacteria treated with HAMLET (lower panel). :::  ::: Treatment of A549 cells with increasing concentrations of HAMLET also induced the accumulation of apoptosis-associated high molecular weight DNA fragments in the 600, 300 and 50 kbp ranges ([Fig. 3B](#pone-0017717-g003){ref-type="fig"}). No oligonucleosomal fragmentation was detected, however, consistent with cell death in response to HAMLET being caspase-independent [@pone.0017717-Tait1] ([Fig. 3B](#pone-0017717-g003){ref-type="fig"}). High molecular weight DNA fragmentation with similar fragment sizes was also detected in pneumococci, but as with the A549 cells, this fragmentation did not proceed to oligonucleosomal fragment sizes ([Fig. 3B](#pone-0017717-g003){ref-type="fig"}). DNA fragmentation was not detected in pneumococci treated with detergents such as deoxycholate or SDS, suggesting that HAMLET activates specific mechanisms responsible for executing the DNA fragmentation. The mechanism of DNA processing in response to HAMLET has not been determined, but likely involves the nucleases that cleave DNA during apoptosis [@pone.0017717-Parrish1]. In an attempt to identify the factors responsible for cleaving the bacterial chromatin we first investigated whether HAMLET itself had DNAse activity. HAMLET was incubated with purified bacterial or tumor cell DNA in the presence of the DNAse cofactors calcium and magnesium and DNA processing was detected after separation of DNA by gel electrophoresis ([Fig. 4A](#pone-0017717-g004){ref-type="fig"}). HAMLET had no DNAse activity under these conditions and failed to cleave either bacterial or tumor cell DNA. DNAse I from bovine pancreas was used as a positive control and effectively processed both DNA samples ([Fig. 4A](#pone-0017717-g004){ref-type="fig"}). ::: {#pone-0017717-g004 .fig} 10.1371/journal.pone.0017717.g004 Figure 4 ::: {.caption} ###### Nuclease activity in HAMLET induced DNA fragmentation. *A*) HAMLET was mixed with either pneumococcal chromosomal DNA or chromosomal DNA from Jurkat cells in the presence of both 1 mM each of Ca^2+^ and Mg^2+^ and the mixture was incubated for 1 hour and digestion was examined by gel electrophoresis. Bovine pancreas DNAse I was used as a positive control for DNA cleavage. HAMLET was unable to cleave either DNA preparation. (*B*) *S. pneumoniae* R6 (WT) and the isogenic strains 577 (**Δ** *endA*), 641 (**Δ** *exoA*) and 642 (**Δ** *endA,* **Δ** *exoA*) were incubated with 100 or 200 µg/ml of HAMLET for 2 h and high molecular weight DNA fragmentation was examined. All strains displayed an accumulation of HMW DNA fragments. :::  ::: Next we assessed the role of pneumococcal endonucleases in HAMLET-induced DNA fragmentation. The *S. pneumoniae* genome contains several open reading frames with potential nuclease activity [@pone.0017717-Tettelin1]. Two nucleases necessary for uptake of DNA during genetic transformation (EndA and ExoA) have been thoroughly characterized [@pone.0017717-Claverys1]. Pneumococcal strains lacking either EndA or ExoA or a double mutant lacking both enzymes were treated with HAMLET and bacterial viability and DNA fragmentation was investigated ([Fig. 4B](#pone-0017717-g004){ref-type="fig"}). No difference was detected, suggesting that these nucleases were not involved in HAMLET-induced DNA fragmentation. 3. HAMLET induces a calcium-dependent depolarization of the mitochondrial and bacterial membranes {#s2c} ------------------------------------------------------------------------------------------------- HAMLET targets mitochondria in tumor cells [@pone.0017717-Kohler1], [@pone.0017717-Kohler2]. In view of the common origin of bacteria and mitochondria [@pone.0017717-Gray1], we investigated if HAMLET-treated bacteria and mitochondria might undergo similar, apoptosis-like biochemical changes. To compare the binding of HAMLET to mitochondria and *S. pneumoniae,* Alexa Fluor 488™-conjugated HAMLET was incubated with isolated rat liver mitochondria or with *S. pneumoniae* D39 for 30 min at 37°C. Binding was quantified by flow cytometry and visualized by confocal microscopy ([Fig. 5](#pone-0017717-g005){ref-type="fig"}). HAMLET bound strongly to isolated mitochondria in a dose-dependent manner. At a sublethal concentration of HAMLET (25 µg/ml), mitochondrial fluorescence was 5.5±0.3 times above the fluorescence of untreated mitochondria, detected by flow cytometry (*P*\<0.01). At a lethal HAMLET concentration (100 µg/ml) the fluorescence increased to 20.2±0.5 times above untreated mitochondria (*P*\<0.01). By confocal microscopy, binding to mitochondria was clearly detected but the sub-organelle distribution could not be determined due to limits of resolution ([Fig. 5](#pone-0017717-g005){ref-type="fig"}). ::: {#pone-0017717-g005 .fig} 10.1371/journal.pone.0017717.g005 Figure 5 ::: {.caption} ###### Association of HAMLET with mitochondria and pneumococci. Confocal micrographs of mitochondria (left) and *S. pneumoniae* D39 (right), incubated with a cytotoxic concentration of Alexa Fluor 488-conjugated HAMLET (100 µg/ml, green) for 1 hour at 37°C, and counterstained with DAPI (300 nM, pseudo-stained red). A light microscopy image (DIC) of each section is included (bottom panels). HAMLET associated with both bacteria and isolated mitochondria. :::  ::: HAMLET\'s association with pneumococci was less pronounced. At the sublethal concentration (25 µg/ml) the bacterial fluorescence was 1.7±0.2 times higher than in untreated pneumococci, as quantified by flow cytometry (*P*\<0.05). At the lethal concentration (100 µg/ml) the fluorescence increased to a mean of 5.5±0.5 times that of untreated bacteria (*P*\<0.01). At this concentration, all bacteria had bound HAMLET, with one population of bacteria showing intense HAMLET-staining ([Fig. 5](#pone-0017717-g005){ref-type="fig"}). This population also stained with propidium iodide, a cell membrane impermeable DNA stain, indicating that these bacteria had ruptured membranes. In previous studies, HAMLET was shown to cause mitochondrial depolarization, followed by mitochondrial swelling, permeability transition, and release of apoptogenic factors to the cytosol, resulting in activation of the cell death program [@pone.0017717-Kohler1]. To further compare the response to HAMLET between mitochondria and bacteria, depolarization of the mitochondrial membrane potential was quantified as the loss of staining with the membrane potential-sensitive dye TMRE or, using the distribution of tetraphenyl-phosphonium ion. In whole tumor cells, mitochondrial depolarization was detected as a decrease in TMRE staining within the first 30 min of HAMLET exposure ([Fig. 6A](#pone-0017717-g006){ref-type="fig"}). Using isolated mitochondria, binding of HAMLET was followed by dissipation of the mitochondrial membrane potential, with complete depolarization detected after approximately 10 minutes ([Fig. 6B](#pone-0017717-g006){ref-type="fig"}) [@pone.0017717-Kohler1]. To examine if the loss of membrane potential required Ca^2+^ the mitochondria were preincubated with the Ca^2+^-transport inhibitor Ruthenium Red and exposed to HAMLET. The response was markedly reduced (78%, *P*\<0.01), suggesting that HAMLET-induced mitochondrial membrane depolarization requires Ca^2+^-transport ([Fig. 6B](#pone-0017717-g006){ref-type="fig"}). ::: {#pone-0017717-g006 .fig} 10.1371/journal.pone.0017717.g006 Figure 6 ::: {.caption} ###### Membrane depolarization and death induced by HAMLET in mitochondria and pneumococci. *A*) Visualization of the membrane potential after HAMLET-treatment of bacteria and tumor cells for 30 minutes. Confocal micrographs depict untreated (left) and HAMLET-treated (right) *S. pneumoniae* AL2 bacteria (D39 **Δ** *lytA*) and A549 carcinoma cells. Bacterial membrane potential was visualized using the anionic bis-oxonol dye DiBAC~4~(3) that accumulates in depolarized bacteria and the tumor cell mitochondrial potential was visualized with the cationic dye TMRE that dissipates from depolarized mitochondria. Treatment of the cells with HAMLET resulted in dissipation of both the bacterial and mitochondrial membrane potential in tumor cells seen by an increased staining with DiBAC~4~(3) and a decreased staining with TMRE, respectively. *B*) Membrane potential and *C*) membrane rupture measurements in *S. pneumoniae* AL2 (D39 **Δ** *lytA*), or in isolated rat liver mitochondria. Bacterial membrane potential was monitored by the DiBAC~4~(3) and membrane rupture by an influx of propidium iodide, after treatment with 31 (HL31), 62 (HL62), or 125 (HL125) µg/ml HAMLET or 125 µg/ml ALA in the absence or presence of 30 µM Ruthenium Red (RuR). Each experiment was repeated six times and the data represents the mean ratio of the six experiments. Membrane potential in isolated mitochondria was measured by the distribution of TPP^+^ ions in the suspension in the presence of 40 nmoles of Ca^2+^ per mg protein after treatment with 50 µg/ml HAMLET in the presence or absence of 10 µM RuR. Arrow indicates addition of mitochondria. The experiment was repeated three times. The graph represents one of the three traces obtained. *D)* Effect of calcium transport inhibition on HAMLET-induced pneumococcal death. *S. pneumoniae* D39 was incubated with increasing concentrations of HAMLET in the presence or absence of 30 µM Ruthenium Red (hatched lines) and viability was monitored after 1 h of incubation by viable plate counts after overnight culture. Viability of bacteria is presented as colony forming units (CFUs) per ml suspension (detection limit in the assay was 10^2^ CFU/ml). Ruthenium Red significantly reduced HAMLETs bactericidal activity. The data represent the mean of four individual experiments with standard deviation error bars. :::  ::: Depolarization of bacterial cell membranes was detected by microscopy as an accumulation of the fluorescent dye DiBAC~4~ (3) ([Fig. 6A](#pone-0017717-g006){ref-type="fig"}). By fluorometry, the rapid and dose-dependent loss of membrane potential was confirmed ([Fig. 6B](#pone-0017717-g006){ref-type="fig"}), with kinetics similar to isolated mitochondria, as described above. Depolarization was followed by membrane rupture detected by an influx of propidium iodide starting approximately 3--5 minutes after the addition of HAMLET ([Fig. 6C](#pone-0017717-g006){ref-type="fig"}). The dose-dependent depolarization was directly associated with a dose-dependent killing (1.6, 2.8, and 5.0 log~10~ viability reduction at 31, 62, and 125 µg/ml of HAMLET after 1 hour; [Fig. 6D](#pone-0017717-g006){ref-type="fig"}). Similar to mitochondria, Ruthenium Red (30 µM) reduced the loss of membrane potential in *S. pneumoniae* by 88.5% (*P*\<0.0001; [Fig. 6B](#pone-0017717-g006){ref-type="fig"}), protected against membrane rupture by 97.6% (*P*\<0.0001; [Fig. 6C](#pone-0017717-g006){ref-type="fig"}) and reduced log-death by 50.5% (*P*\<0.0001; [Fig. 6D](#pone-0017717-g006){ref-type="fig"}). The results suggest that HAMLET directly influences the membrane potential, in *S. pneumoniae* and mitochondria, and indicate that Ca^2+^ transport is involved in both systems. 4. Role of protease activity in HAMLET-induced death {#s2d} ---------------------------------------------------- Mitochondrial permeability transition leads to a release of apoptogenic factors from mitochondria, which induce apoptosis in eukaryotic cells, either with or without caspase activation [@pone.0017717-Tait1]. Even though exposure of tumor cells to HAMLET results in cytochrome c-release from mitochondria and activation of caspases, inhibitors of these pathways do not prevent cell death and cell death is not regulated by Bcl-2 or p53 family proteins [@pone.0017717-Hallgren1], [@pone.0017717-Rammer1], demonstrating that the effects of HAMLET on cell viability are independent of common apoptosis effector molecules. The pan-caspase inhibitor zVAD-fmk also does not inhibit high molecular weight DNA fragmentation in response to HAMLET [@pone.0017717-Hallgren1]. To address the role of other proteases, tumor cells were preincubated with the calcium-dependent cysteine protease inhibitor calpeptin or the serine protease inhibitor dichloroisocoumarin (DCI) and effects on HAMLET-induced tumor cell death and DNA fragmentation were recorded ([Fig. 7A](#pone-0017717-g007){ref-type="fig"}). Calpeptin inhibited HAMLET-induced log-death by 43% (*P*\<0.001) and effectively blocked DNA fragmentation in response to HAMLET. DCI significantly reduced death in response to HAMLET (12%; *P*\<0.05) and almost completely blocked DNA fragmentation. ::: {#pone-0017717-g007 .fig} 10.1371/journal.pone.0017717.g007 Figure 7 ::: {.caption} ###### Role of serine proteases in HAMLET-induced death of tumor cells and pneumococci. A\) A549 carcinoma cells were preincubated for 10 minutes with diluent, 25 µM zVAD-fmk (pan-caspase inhibitor), 100 µM dichloroisocoumarin (serine protease inhibitor), or 100 µM calpeptin (calcium-dependent cysteine protease inhibitor) before being treated with 300 µg/ml of HAMLET. After 6 hours of incubation cell viability was measured using trypan blue exclusion. The graph depicts the mean death in % obtained after 3 individual experiments. The error bars represent the standard deviation. \* and \*\*\* represent *P*\<0.05 and *P*\<0.001, respectively. B and C) *S. pneumoniae* AL2 (D39 **Δ** *lytA*) were preincubated for 10 minutes with diluent, 25 µM Aprotinin (serine protease inhibitor), 25 µM zVAD-fmk (pan-caspase inhibitor), or 10 µM E-64 (cysteine protease inhibitor) before being treated with 50 µg/ml of HAMLET. After 2 hours viability was determined and samples were analyzed for high molecular weight DNA fragmentation. *B)* Viability. The graph depicts the mean log~10~ death obtained from five individual experiments. The error bars represent the standard deviation. \*\* represents *P*\<0.01. C) DNA fragmentation. Untr indicates untreated bacteria. The remaining samples were treated with HAMLET in the presence of diluent (none) or proteasee inhibitors. Only the serine protease inhibitors aprotinin rescued pneumococci from death and DNA fragmentation. :::  ::: To evaluate the role of proteases in bacterial death we first searched the pneumococcal genome for caspase-homologues, but no none were identified. To address if bacteria express other proteases with caspase-like activity, *S. pneumoniae* D39 lysates were incubated with the three major groups of caspase-substrates YVAD-, IETD-, and DEVD-amc in a fluorescence-based assay but no activity was detected. To verify the lack of caspase-related activity, bacteria were incubated with the pan-caspase inhibitor zVAD-fmk or with the cysteine protease inhibitor E-64. No significant effect on bacterial viability was observed and DNA fragmentation remained unchanged ([Fig. 7B and C](#pone-0017717-g007){ref-type="fig"}). In contrast, pretreatment with the serine protease inhibitor aprotinin reduced HAMLET-induced log-death by 39% ([Fig. 7B](#pone-0017717-g007){ref-type="fig"}, *P*\<0.01) and inhibited DNA fragmentation ([Fig. 7C](#pone-0017717-g007){ref-type="fig"}), indicating a role for this family of proteases in the execution of HAMLET-induced bacterial death. 5. Effect of HAMLET on other bacterial species {#s2e} ---------------------------------------------- In addition to *S. pneumoniae*, HAMLET effectively kills other streptococcal species as well as the respiratory pathogen, *Haemophilus influenzae* [@pone.0017717-Hakansson2]. This pathogen was selected to investigate if the apoptosis-like response also occurs in other bacterial species. HAMLET was found to bind to *H. influenzae* cells ([Fig. 8A](#pone-0017717-g008){ref-type="fig"}) and binding resulted in depolarization ([Fig. 8B](#pone-0017717-g008){ref-type="fig"}) of the bacterial membrane. Although the depolarization was more rapid than in pneumococci it showed less pronounced changes (2--3 fold increase in fluorescence compared to 5--6 fold increase in pneumococci). Membrane depolarization resulted in rupture of the bacterial membrane ([Fig. 8C](#pone-0017717-g008){ref-type="fig"}) and loss of viability ([Fig. 8D](#pone-0017717-g008){ref-type="fig"}), both less pronounced than for pneumococci. As for pneumococci, Ruthenium Red inhibited depolarization and death, with membrane potential being reduced by 52.8% (*P*\<0.01; [Fig. 8B](#pone-0017717-g008){ref-type="fig"}), membrane rupture by 72.1% (*P*\<0.01; [Fig. 8C](#pone-0017717-g008){ref-type="fig"}) and log~10~ death decreased by 62.4% (*P*\<0.05; [Fig. 8D](#pone-0017717-g008){ref-type="fig"}). ::: {#pone-0017717-g008 .fig} 10.1371/journal.pone.0017717.g008 Figure 8 ::: {.caption} ###### Apoptosis-like changes in HAMLET-treated *H. influenzae*. A\) Association of HAMLET with bacteria. Confocal micrographs of *H. influenzae* 2019, incubated with a cytotoxic concentration of Alexa Fluor 488-conjugated HAMLET (250 µg/ml, green) for 1 hour at 37°C, and counterstained with DAPI (300 nM, pseudo-stained red). A light microscopy image (DIC) of each section is included in the bottom row. *B*) Membrane potential and *C*) membrane rupture measurements in *H. influenzae* 2019. Bacterial membrane potential was monitored by the DiBAC~4~(3) and membrane rupture by an influx of propidium iodide, after treatment with 62 (HL62), 125 (HL125), or 250 (HL250) µg/ml HAMLET or 250 µg/ml ALA in the absence or presence of 30 µM Ruthenium Red (RuR). Each experiment was repeated six times and the data represents the mean ratio of the six experiments. *D)* Effect of calcium transport inhibition on HAMLET-induced pneumococcal death. *H. influenzae* 2019 was incubated with increasing concentrations of HAMLET in the presence or absence of 30 µM Ruthenium Red (hatched lines) and viability was monitored after 1 h of incubation by viable plate counts after overnight culture. Viability of bacteria is presented as colony forming units (CFUs) per ml suspension (detection limit in the assay was 10^2^ CFU/ml, (mean of four experiments with standard deviation error bars). E) Chromatin fragmentation induced by HAMLET in *H. influenzae*. High molecular weight DNA fragments were induced by HAMLET in *H. influenzae* 2019 cells and detected after 1 h of incubation. (HAMLET concentration in µg/ml). Increasing concentrations of HAMLET resulted in accumulation of DNA fragments over time. Low molecular weight oligonucleosomal DNA fragments were not observed (lower panel). :::  ::: DNA fragmentation in response to HAMLET was also detected in *H. influenzae,* with a dose-dependent accumulation of high molecular weight DNA fragments in the 600, 300 and 50 kbp ranges, similar to the DNA fragmentation which occurred in HAMLET-treated tumor cells and pneumococci ([Fig. 8E](#pone-0017717-g008){ref-type="fig"}). No oligonucleosomal fragmentation was detected. These results suggest that HAMLET\'s bactericidal mechanism is similar in different HAMLET-sensitive bacterial species and not restricted to gram-positive or gram-negative organisms. Discussion {#s3} ========== Apoptotic cell death is critical for eukaryotic cell turn over, tissue development and homeostasis. Here we describe that features of apoptosis can be activated also in bacterial cells. Striking similarities were observed in cellular responses of eukaryotic and prokaryotic cells to the human milk complex HAMLET, including cell death accompanied by DNA fragmentation and a change in morphology with cell shrinkage and DNA condensation. HAMLET bound to eukaryotic and prokaryotic cell membranes and induced a calcium-dependent depolarization of the plasma- or mitochondrial membranes as well as bacterial cells, followed by downstream degradation pathways involving protease and endonuclease activity. The similarities between mitochondrial and bacterial responses to HAMLET are consistent with their shared evolutionary origin and suggest that HAMLET activates targets that are conserved in tumor cells and certain bacteria. The results support our speculation that death activation pathways evolved early, and suggest that similar mechanisms may be shared by prokaryotic and eukaryotic organisms. Characterizing these basic mechanisms of cell death regulation may be important for future disease therapies involving both eukaryotic and prokaryotic cells. HAMLET shows a specific activity against certain bacterial species but fails to kill others [@pone.0017717-Hakansson2]. *S. pneumoniae* and many other streptococci are sensitive, undergoing apoptosis like changes including DNA fragmentation, but most other species, including *Escherichia coli* and Staphylococci, are not killed. The membranes of sensitive species are rapidly depolarized by HAMLET. In a screen of bacterial HAMLET sensitivity, we observed that resistant species such as Staphylococci respond to HAMLET with low membrane depolarization and a difference in signaling threshold between sensitive and resistant bacteria is further supported by our observation that both HAMLET-sensitive and resistant species can undergo death with apoptosis-like morphology in response to other death stimuli, such as starvation. Apoptotic response pathways may thus be present in most bacteria but differentially activated depending on the agonist. The difference in HAMLET sensitivity was not related to the bacterial structure as defined by gram-positive or gram-negative staining, but it may be speculated that sensitivity has evolved to fit the niche the organisms inhabit. In the case of HAMLET, sensitive species are found primarily in the oral cavity and respiratory tract, which are exposed to human milk and its constituents during breast feeding. HAMLET triggered high molecular weight DNA fragmentation in both carcinoma cells and bacteria but no oligonucleosomal fragments were observed. In eukaryotic cells, proteolytic cleavage of lamins and other structural DNA-associated proteins is necessary for the early formation of high molecular weight DNA fragments ranging from 50--600 kbp [@pone.0017717-Ankarcrona1]. These DNA fragments represent excised DNA loops and oligomers that precede the oligonucleosomal DNA fragmentation [@pone.0017717-Taylor1]. The shared DNA fragmentation pattern displayed by eukaryotic and bacterial cells after HAMLET-treatment suggests that the topology and packing of the DNA may be more similar than currently appreciated. Protease activation is commonly required for the unpacking of DNA and for endonuclease activation during eukaryote death [@pone.0017717-Taylor1], [@pone.0017717-Susin1]. In tumor cells, HAMLET required calpain and serine protease activity to exert its effects, as death and DNA fragmentation were effectively inhibited by calpeptin and DCI, respectively. For *S. pneumoniae,* we show that HAMLET-induced DNA fragmentation was independent of the two well-characterized endonucleases EndA and ExoA but required serine proteases, as inhibition by aprotinin effectively rescued HAMLET-treated pneumococci from DNA fragmentation and death. While the exact mechanisms remain to be identified, our observations suggest that DNA processing in response to HAMLET may be a universal aspect of cell death in both eukaryotic and certain prokaryotic cell kingdoms. In pneumococci, this leaves a number of open reading frames with potential endonuclease activity, which might be involved in DNA repair and DNA catabolism [@pone.0017717-Tettelin1]. HAMLET\'s binding to mitochondria and bacteria caused a rapid loss of membrane potential. This effect was inhibited by Ruthenium red, which blocks calcium fluxes. In eukaryotic cells, calcium signaling plays a clear and prominent role in the regulation of many cellular processes including cell death [@pone.0017717-Lemasters1]--[@pone.0017717-Pinton1]. An increase in cytosolic calcium invariably stimulates mitochondrial uptake through the mitochondrial uniporter and other systems and excessive calcium uptake by mitochondria causes depolarization of the mitochondrial membrane and opening of the permeability transition pore, which leads to the release of apoptogenic factors that trigger the execution phase of apoptotic cell death [@pone.0017717-Lemasters1], [@pone.0017717-Crompton1], [@pone.0017717-Kroemer1]. A role for calcium in HAMLET-induced permeability transition in isolated mitochondria has previously been suggested, based on observations that EGTA inhibited HAMLET-induced swelling of mitochondria and the release of cytochrome c [@pone.0017717-Kohler1]. The role of calcium in bacterial cell signaling is more elusive [@pone.0017717-Dominguez1], [@pone.0017717-Norris1]. There is evidence for a role of calcium in responses to environmental stresses as well as a potential role in cell cycle progression and differentiation processes such as sporulation and fruiting body development, but no information is available regarding calcium\'s involvement in bacterial cell death [@pone.0017717-Dominguez1]. While Ruthenium Red inhibited HAMLET-induced depolarization of the bacterial membrane and rescued the bacteria from death, the calcium transport mechanism induced by HAMLET remains unknown. Pneumococci express a P-type Ca^2+^-ATPase for calcium efflux and calcium transport through a sodium/calcium exchanger has been proposed to regulate competence, DNA uptake and lysis, but the transporter has not been identified and is not annotated in published genomes [@pone.0017717-Trombe1], [@pone.0017717-Trombe2]. Furthermore, BLAST analysis of both the pneumococcal and *H. influenzae* genomes against the transport classification database (<http://www.tcdb.org/>) failed to identify any other calcium transporters, suggesting that sequences and motifs differ from those characterized so far. In multicellular organisms, controlled elimination of aged, faulty or potentially harmful cells is an important feature to maintain functional homeostasis of the organism, and the need for such mechanisms is easily apparent. It is especially important to degrade defective DNA to avoid the persistence of mutated DNA that could become detrimental to the organism. Bacteria may require similar mechanisms, especially since they often grow in aggregated communities or biofilms. Specialization of bacterial function (terminal differentiation) or sacrifice of some individual cells in favor of others (altruism) within these communities appears to exist and elimination of cells using a genetically inherent pathway could be advantageous [@pone.0017717-Lewis1], [@pone.0017717-Bayles1]. This is especially evident for *S. pneumoniae* where it was recently shown that the release of the intracellular virulence factor pneumolysin, as well as DNA, is due to the predation of genetically competent organisms on non-competent organisms surrounding them, a phenomenon named fratricide, which has been seen also in other bacterial species [@pone.0017717-Claverys2]. This would also indicate that the release of fragmented DNA in response to lethal stimuli can benefit the bacterial "community" by increasing the spread of genetic information, including antibiotic resistance, to surrounding bacteria and be used to form an intercellular matrix in biofilms [@pone.0017717-Bayles1]. Indeed, such DNA release may well be critical to the very efficient pneumococcal genetic transformation originally described by Avery [@pone.0017717-Avery1]. Molecules such as HAMLET may thus have evolved to help the infant combat potentially harmful infections early in life [@pone.0017717-Lawrence1], and/or to help the infant regulate cell proliferation during the early, rapid growth and development of mucosal tissues [@pone.0017717-Davis1]. Materials and Methods {#s4} ===================== Reagents {#s4a} -------- DEAE-Trisacryl M was from BioSepra (Villeneuve la Garenne, France). SeaKem GTG agarose and SeaPlaque GTG Low melting temperature agarose gel and were from SeaKem, FMK Bioproducts (Rockdale, USA). Trypan blue was from Chroma Gesellschaft, Schmid & Co (Stuttgart, Germany). Production of HAMLET {#s4b} -------------------- HAMLET was produced by converting native alpha-lactalbumin in the presence of oleic acid (C18:1) as described [@pone.0017717-Svensson1]. Briefly, native alpha-lactalbumin was purified from human milk by ammonium sulfate precipitation and phenyl sepharose chromatography [@pone.0017717-Lindahl1]. Apo alpha-lactalbumin was generated from 25 mg of native alpha-lactalbumin dissolved in Tris (10 mM Tris/HCl pH 8.5) by addition of 3.5 mM EDTA. Conversion of apo-alpha-lactalbumin to HAMLET was achieved by ion exchange chromatography on DEAE-Trisacryl M matrix conditioned by addition of 10 mg of C18:1 fatty acid. The protein was eluted by applying increased concentrations of NaCl in a Tris buffer devoid of EDTA. Cells {#s4c} ----- Jurkat leukemia cells were obtained from the European Collection of Cell Cultures (Wiltshire, UK) and A549 cells (CCL-185) were obtained from the American Type Culture Collection (Manassas, VA, USA). Both cell types were cultured in RPMI 1640 medium supplemented with 10% fetal calf serum, 2 mM glutamine, non essential amino acids, sodium pyruvate, and 50 µl gentamicin/ml, at 37°C in a humidified atmosphere containing 5% CO~2~. The effect of HAMLET and ALA on cell viability was assessed by measuring the exclusion of trypan blue (Invitrogen) in the cell population. Bacteria {#s4d} -------- The following isolates of *S. pneumoniae* were used in the study: DBL5 [@pone.0017717-Yother1], WU2 [@pone.0017717-Briles1], TIGR4 [@pone.0017717-Tettelin1], L82006, L81905, L82016, L82013, BG9273, BG7322, BG30-11, BG8826 [@pone.0017717-McDaniel1], EF10197, EF10175, EF3030, EF3296, EF3559, EF1488 [@pone.0017717-Andersson1], [@pone.0017717-Andersson2], ATCC 6303 [@pone.0017717-Briles2], 3JYP2670 [@pone.0017717-Roche1], ATCC 46919 [@pone.0017717-Jorgensen1] A66, and D39 [@pone.0017717-Avery1]. Furthermore D39 pneumococci lacking the autolysin LytA, AL2 [@pone.0017717-Berry1], PspA [@pone.0017717-Yother2], PspC [@pone.0017717-Balachandran1], PsaA [@pone.0017717-Berry2], pneumolysin [@pone.0017717-Berry3], and DLDH [@pone.0017717-Smith1] were used to evaluate the role of virulence factors for HAMLET-induced pneumococcal death. These mutants were all produced through insertion duplication mutagenesis where the target gene was interrupted with a plasmid carrying erythromycin resistance. Strains 577, 641, and 642, lacking nucleases were kindly provided by Dr Sanford Lacks, New York, USA [@pone.0017717-Puyet1], [@pone.0017717-Puyet2]. *H. influenzae* strain 2019 was kindly provided by Dr Campagnari, University at Buffalo, SUNY [@pone.0017717-Campagnari1]. *H. influenzae* strain Eagan (type b) [@pone.0017717-Lysenko1] and Rd (type e) [@pone.0017717-Alexander1] were kindly provided by Jeffrey Weiser, University of Pennsylvania, Philadelphia, PA. The pneumococcal strains were stored in glycerol stocks at −80°C, and frozen stocks were used to seed Todd Hewitt medium containing 0.5% Bacto-Yeast extract. *Haemophilus influenzae* were grown on chocolate agar and seeded into Brain heart infusion broth containing 10 ml/L IsoVitaleX enrichment solution (BD Biosciences), 5% fetal bovine serum (Invitrogen), and 25 mg/L hemin (Sigma). In late logarithmic growth phase, the bacteria were harvested by centrifugation at 1200× *g* for 20 minutes and suspended in phosphate-buffered saline (PBS; 30 mM Na~2~HPO~4~, 10 mM KH~2~PO~4~, 120 mM NaCl, pH 7.4). Appropriate dilutions of the bacteria were suspended in PBS. The effect of HAMLET on bacterial viability was assessed by viable counts on blood agar or chocolate agar plates, respectively. Chromatin condensation and fragmentation {#s4e} ---------------------------------------- For detection of high molecular weight DNA fragments, treated bacteria and cells were pipetted into gel plugs that were treated with proteinase K for 24 hours as described [@pone.0017717-Hakansson1]. Gel electrophoresis was run at 12°C, 175 V in 1% agarose gels in 0.5× TBE (45 mM Tris, 1.25 mM EDTA, 45 mM boric acid, pH 8.0), with the ramping rate changing from 0.8 seconds to 30 seconds for 24 hours, using a forward to reverse ratio of 3:1. DNA fragment size was calibrated using two sets of pulse markers: chromosomes from *Saccharomyces cerevisiae* (225--2200 kbp) and a mixture of lDNA Hind III fragments, lDNA and lDNA concatemers (0.1--200 kbp) from Sigma. To study DNA morphology of bacteria or cells, treated bacteria were fixed in 4% paraformaldehyde and exposed to 300 nM DAPI (Invitrogen) and viewed by fluorescence microscopy using a Leica DMI6000 microscope (Leica Microsystems, Bannockburn, IL). Isolation of mitochondria {#s4f} ------------------------- Jurkat cells were pelleted and washed in buffer, containing 100 mM sucrose, 1 mM EGTA and 20 mM MOPS, resuspended in 5% Percoll, 0.01% digitonin and protease inhibitors and incubated on ice for 10 min, followed by centrifugation at 2,500× g for 5 minutes. The supernatant was subjected to an additional centrifugation at 10,000× g for 15 minutes, mitochondrial pellet was collected in 300 mM sucrose, 1 mM EGTA, 20 mM MOPS and protease inhibitors and kept at −70°C. Mitochondria (3 mg/ml) were transferred into buffer, containing 250 mM sucrose, 10 mM MOPS, 5 mM succinate, 3 mM KH~2~PO~4~, 10 µM EGTA and 10 mM Tris, pH 7.5 and after incubation at 30°C centrifuged at 10,000× g for 15 minutes. Binding of HAMLET to mitochondria and bacteria {#s4g} ---------------------------------------------- HAMLET protein was directly labelled with Alexa Fluor™ 488 (Molecular Probes Inc) according to manufacturer\'s instructions. Bacteria (10^8^/ml, 100 µl) or mitochondria (3 mg/ml protein concentration, 25 µl) were incubated with fluorescently labelled HAMLET at 37°C for various 30 min. The fluorescence intensity of the bacteria were analyzed in a FACSCalibur flow cytometer (BD) using a 520 nm band-pass filter or bacteria and mitochondria were counterstained with 300 nM DAPI and inspected by confocal microscopy. Measurement of membrane potential {#s4h} --------------------------------- For a visual depiction of the membrane potential in whole cells, Jurkat cells were treated with 0.3 mg/ml of HAMLET for 30 min and 25 nM of TMRE was added 15 minutes before the cells were inspected by confocal microscopy. The potential over the mitochondrial membrane of isolated mitochondria was measured using an electrode sensitive to the cation tetraphenylphosphonium as described [@pone.0017717-Kohler1]. Bacteria were pelleted by centrifugation at 2,400× *g* and washed twice by resuspension in PBS (pH 7.2) followed by centrifugation. The bacterial pellet was resuspended in half the original volume of PBS and energized with 50 mM glucose for 15 min at 37°C. To energized bacteria, propidium iodide (40 µg/ml) and DiBAC~4~(3) (0.5 µM) were added and 100 µl bacterial suspension was added to 100 µl of PBS, HAMLET or ALA in each well of a 96-well microtiter plate (Falcon, BD Biosciences) resulting in a final concentration of 25 mM glucose, 20 µg/ml PI and 250 nM DiBAC~4~(3) in each well. The plate was immediately placed into a 37°C pre-warmed Synergy II microplate reader (Biotek, Winooski, VT) and the DiBAC~4~(3) fluorescence (485/20 nm excitation, 520/25 nm emission) and PI fluorescence (485/20 nm excitation, 590/35 nm emission) were read every minute for 60 minutes to monitor the change membrane polarity and integrity, respectively. Protease activity assays {#s4i} ------------------------ Bacterial extracts were produced as described above. Aliquots containing bacterial extract were then transferred to a 96-well plate and 50 µl of freshly prepared substrate buffer (100 mM HEPES, 10% sucrose, 0.1% CHAPS, 5 mM DTT 10^−6^% NP-40, pH 7.25) containing either of the substrates Ac-YVAD-amc, Ac-DEVD-amc or Ac-IETD-amc was added per well. The final concentration of Ac-YVAD-amc, Ac-DEVD-amc or Ac-IETD-amc was 14.4 µM, 33.3 µM and 13.6 µM, respectively. The enzymatic reaction was carried out at 37°C and the rates of hydrolysis were measured by release of amc from the substrates using an ELISA reader. Experiments were performed in duplicates and the activity was expressed as change in fluorescence units per min per 10^6^ cells. For protease inhibition studies, *S. pneumoniae* AL2 (D39 **Δ** *lytA*) were preincubated for 10 minutes in the presence of diluent, 25 µM aprotinin, 10 µM E-64, or 25 µM ZVAD-fmk before being treated with 50 µg/ml of HAMLET for 2 hours. Bacterial cells were then diluted and plated for determination of colony forming units on blood agar after overnight culture. Nuclease activity assays {#s4j} ------------------------ Nuclease activity of HAMLET was tested as the ability to cleave chromosomal DNA from D39 bacteria or Jurkat cells. Chromosomal DNA from cells and bacteria was prepared by standard procedures using phenol/chloroform extraction. For tests of nuclease activity 1 µg of *S. pneumoniae*, or Jurkat chromosomal DNA were mixed with 50 µg/ml HAMLET in phosphate buffered saline (30 mM Na~2~HPO~4~, 10 mM KH~2~PO~4~, 120 mM NaCl, pH 7.4) with 1 mM Ca^2+^ and 1 mM Mg^2+^, and incubated for 1 hour at 37°C, and run by agarose gel electrophoresis in 1.5% gels with a constant voltage of 100 V. DNAse I from bovine pancreas (100 Kuntz units/ml) was used as a positive control. Statistical analysis {#s4k} -------------------- Quantitative data were analysed from a minimum of three repeats using Student\'s T-test with a 2-tailed *P*-value. The *n* for each analysis is presented in the figure legends. The authors thank Emily Clementi, Camilla Köhler, Alexander Smith, and Boris Zhivotovsky for technical assistance. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This study was supported by grants from Bill and Melinda Gates Foundation (<http://www.gatesfoundation.org/>, Grant 53085; AH), the JR Oishei Foundation (<http://www.oisheifdt.org/>, AH), The Royal Physiographic Society, Lund, Sweden (<http://www.fysiografen.se/> AH), The Swedish Society for Medical Research, Stockholm, Sweden (<http://www.ssmf.se/> AH), John and Augusta Perssons Foundation for Scientific Medical Research, Lund, Sweden (AH), Anna-Lisa and Sven-Eric Lundgrens Foundation for Medical Research, Malmö, Sweden (AH), Gunnar, Arvid and Elisabeth Nilssons Foundation for Fighting Cancer, Helsingborg, Sweden (AH), Maggie Stephen\'s Foundation, Lund, Sweden (AH), The Swedish Cancer Foundation (<http://www.cancerfonden.se/> Grants no. 3807-B97-01XAB and 96-1719, CS), The American Cancer Society (<http://www.cancer.org/> Grant no. RPG 97-157-01, CS), The Swedish Heart and Lung Foundation (<http://www.hjart-lungfonden.se/> CS) and the American Lung Association (Grant RG-123721-N). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: AH CS. Performed the experiments: AH HR-H A-KM. Analyzed the data: AH. Contributed reagents/materials/analysis tools: A-KM. Wrote the paper: AH CS.

PubMed Central

2024-06-05T04:04:19.836382

2011-3-10

PMC3053381

Introduction {#s1} ============ Multiple Sclerosis (MS) is a complex demyelinating disease of the central nervous system (CNS). Although the initiating insult in MS remains unknown, it is clear that the pathology of MS involves complex interactions between many systems and cell types, including neurons, glia and both the innate and adaptive immune systems. The majority of research in MS has focussed on the indisputably important role of adaptive immunity in the disease. However, it has recently been posited that, at least in some cases, the primary insult could be directed at the oligodendrocyte. In a seminal study in 2004, Barnett and Prineas examined newly-forming lesions from MS patients and found that prior to the onset of extensive demyelination, and in the absence of T-cell infiltration, oligodendrocyte death could be detected, generally in association with increased numbers of microglia [@pone.0017727-Barnett1]. This work suggests that oligodendrocyte apoptosis and the innate immune response could have important roles to play in the initial development of an MS lesion. A family of receptor protein tyrosine kinases known as the TAMs, comprised of three closely related proteins (Tyro3, Axl and Mertk), has been shown to be centrally important in the regulation of both oligodendrocyte survival and microglial activation. The three TAM receptors were identified as a distinct receptor PTK subfamily in 1991 [@pone.0017727-Lai1]. The extracellular, ligand-binding regions of these receptors each have a defining arrangement of two tandem immunoglobulin-related domains and two fibronectin type III repeats. These domains are followed by a single-pass transmembrane domain, and a catalytically competent, cytoplasmic PTK [@pone.0017727-Janssen1], [@pone.0017727-OBryan1], [@pone.0017727-Rescigno1]. Like all other receptor PTKs, the TAMs signal as dimers. They are activated by the binding of two closely-related ligands, Gas6 (Growth Arrest Specific Gene 6) and Protein S (ProS), which also dimerize (reviewed in [@pone.0017727-Lemke1]). We, and others, have recently shown that TAM signalling can regulate myelination, microglial activation and phagocytosis during both demyelination and remyelination in the CNS. We showed that Gas6 knockout (KO) mice subjected to cuprizone-induced demyelination for three weeks exhibited significant differences in both myelination and microglial activation in comparison with wild-type (WT) mice: in the absence of Gas6, oligodendrocyte survival was compromised, demyelination was worse and there were fewer myelinated axons [@pone.0017727-Binder1]. In contrast, a study by Hoehn *et al.* of cuprizone challenge to Axl KO mice did not observe an increased loss of oligodendrocytes in comparison with wild-type mice after 6 weeks of cuprizone-challenge [@pone.0017727-Hoehn1]. This apparent disparity between the relative effect of Gas6 and Axl deficiency upon oligodendrocyte survival in mice challenged with cuprizone suggests that a loss of signalling through a single receptor (Axl) has a less profound effect than a reduction in signalling through all three receptors, such as in the absence of the ligand Gas6. Alternatively, it could be that either Tyro3 or Mer is the main TAM receptor responsible for transducing the anti-apoptotic effect of Gas6 in this context. In addition to the studies in animal models of demyelination described above, recent evidence has shown that that TAM receptor signalling is involved in both the etiology and pathogenesis of MS. We recently conducted an association study to identify single nucleotide polymorphisms (SNPs) within genes encoding the TAM receptors and their ligands associated with MS. We identified polymorphisms within the *MERTK* gene associated with the risk of developing MS [@pone.0017727-Ma1]. Further evidence linking TAM signalling to the pathogenesis of MS also has emerged from studies of human MS lesions, revealing that soluble Axl and Mer proteins, which act as decoy receptors to sequester Gas6, are upregulated in chronic MS lesions, and are negatively correlated with expression of Gas6 [@pone.0017727-Weinger1]. These data suggest that a reduction in protective Gas6 signalling could correlate with extended lesion activity. Given the known role that Gas6 and the TAM receptors play in regulating myelination, we examined the role of these molecules during recovery from cuprizone-induced demyelination, and in an *in vitro* model of myelination. Here, we demonstrate that the absence of Gas6 delays, but does not ultimately prevent, remyelination following cuprizone-induced demyelination, and that this delay is correlated with a reduction in oligodendrocyte numbers. This effect may be contributed to by a direct effect of Gas6 upon myelination by oligodendrocytes, as we also show that exogenous Gas6 can enhance myelination of dorsal root ganglion (DRG) neurons *in vitro*. Materials and Methods {#s2} ===================== Animals and reagents {#s2a} -------------------- Gas6^−/−^ mice [@pone.0017727-AngelilloScherrer1] were backcrossed onto the C57Bl6 background as previously described [@pone.0017727-Binder1]. Sprague-Dawley rats for primary cell culture were obtained from the Animal Resource Centre (Canning Vale, WA, Australia). All chemical reagents were obtained from Sigma-Aldrich (St Louis, MO) unless otherwise indicated. Recombinant human Gas6 (rhGas6) was a kind gift of Dr Patrick Jones (Berlex Biosciences, Richmond, CA). All cell culture plasticware was purchased from Nunc (Rochester, NY). All cell culture media and reagents were purchased from Invitrogen (Carlsbad, CA) unless otherwise indicated. All secondary antibodies were purchased from Jackson Immunochemicals (West Grove, PA) unless otherwise indicated. Ethics Statement {#s2b} ---------------- All animal experiments were approved by the Florey Neuroscience Institutes Animal Ethics Committee (Approval \#07-063) and conducted according to National Health and Medical Research Council (NH&MRC) guidelines, with all appropriate effort made to minimise animal suffering. Purification and culture of rat oligodendrocyte precursor cells (OPCs) {#s2c} ---------------------------------------------------------------------- Oligodendrocyte precursor cells were purified from the cortices of P7 Sprague-Dawley rat pups using sequential immunopanning as previously described [@pone.0017727-Barres1]. Purified cells were cultured in 75 cm^2^ tissue culture flasks coated with poly-D-lysine (PDL) in modified Bottenstein-SATO medium [@pone.0017727-Bottenstein1] containing N-acetyl-cysteine (60 µg/ml), forskolin (5 µM), penicillin and streptomycin, neurotrophin-3 (NT-3, 5 ng/ml, Peprotech, Rocky Hill, NJ) and platelet-derived growth factor-AA (PDGF-AA, 10 ng/ml, Peprotech, Rocky Hill, NJ). Co-culture of rat dorsal root ganglia (DRG) and OPCs {#s2d} ---------------------------------------------------- Rat DRG-OPC co-cultures were established based on previously described protocols [@pone.0017727-Kleitman1], [@pone.0017727-Chan1]. Briefly, rat DRG neurons were purified with antimitotic medium in the presence of NGF (100 ng/ml) for 2--3 weeks as described previously [@pone.0017727-Xiao1]. OPCs were seeded onto coverslips containing purified DRG in a small volume at a density of 200,000 OPCs per 22 mm coverslip or 100,000 OPC per well of 4-well chamber slides and incubated overnight to facilitate attachment. DRG-OPC co-cultures were maintained for 14 days in a defined co-culture media containing 50∶50 DMEM∶Neurobasal medium with Sato and B27 supplements, N-acetyl cysteine and D-biotin with or without Gas6 according to the experimental paradigm. Analysis of myelination in DRG-OPC co-culture {#s2e} --------------------------------------------- For analysis of protein levels using western blot, the lysates of DRG-OPCs co-cultures were separated on an SDS-PAGE gel and transferred to a PVDF membrane, as described previously [@pone.0017727-Xiao1]. Membranes were probed with specific antibodies against myelin proteins (2′,3′-cyclic nucleotide 3′-phosphodiesterase \[CNPase, Chemicon, Temecula, CA\], myelin associated glycoprotein \[MAG, Chemicon, Temecula, CA\] and myelin basic protein \[MBP, Chemicon, Temecula, CA\]), then incubated with HRP-conjugated secondary antibodies (Cell Signalling, Danvers, MA). Anti-β-actin antibody (Sigma, St. Louis, CA) was used as a loading control. Blots shown are representative of 3 independent experiments. For determination of the number of MBP segments, immunocytochemical analysis of myelinating co-cultures was undertaken, as described previously [@pone.0017727-Xiao1]. Briefly, the co-cultures were fixed with 4% paraformaldehyde, then blocked with 20% calf serum in 0.2% Triton X-100 in PBS. Myelin segments were visualized with an anti-MBP antibody (AB980, Chemicon, Temecula, CA), followed by incubation with secondary antibodies, and images were captured by Zeiss confocal microscopy (Carl Zeiss, Inc. Thornwood, NY, USA). For quantitative analysis, 4 random fluorescent images were taken as above, and the number of MBP positive myelin segments in each field was counted. Determination of cell death {#s2f} --------------------------- Co-cultures of OPCs and DRG were established, as described above. Cells were cultured for either 1, 2, 7 or 14 days with appropriate factors depending on the experimental paradigm. Ethidium homodimer-1 (4 mM) and calcein AM (4 mM) were added to each well. Cells were then visualised on an inverted fluorescent microscope (Zeiss, Thornwood, NY) and live (calcein AM positive) and dead (ethidium homodimer-1 positive) cells counted. For all assays 3 wells were counted per condition and assays were repeated at least two times with independent cell isolations. All results are expressed as % live cells±SEM. Induction of demyelination/remyelination {#s2g} ---------------------------------------- Cuprizone mediated demyelination was induced by feeding 8--10 week old mice powdered feed (Barastoc, Pakenham, Victoria, Australia) containing 0.2% cuprizone (w/w: bis-cyclohexanone-oxaldihydrazone) for 5 weeks. Mice were then returned to a normal diet for either 0, 2, 4 or 10 weeks, according to the experimental paradigm. During the 5 week demyelination phase, feed was refreshed each day. Wild-type littermates were used as controls for induction of demyelination in Gas6^−/−^ cohorts. In one cohort (cohort 3), the wild-type, control animals were supplemented with separately bred C57Bl6 mice to increase available numbers. For each cohort an equal mix of male and female Gas6^+/+^ and Gas6^−/−^ mice were subjected to cuprizone challenge. Consistent with previous findings [@pone.0017727-Taylor1], no gender effect was identified. The brains of animals from cohorts 1 and 2 were embedded, cut and analysed in sagittal orientation, whereas those from cohort 3 were embedded, cut and analysed in the coronal orientation. Histology {#s2h} --------- Mice were anaesthetised and perfused intracardially with PBS followed by 4% paraformaldehyde and the brains embedded in paraffin or, for cryosectioning, were equilibrated successively in 12%, 16% and 18% sucrose in PBS for cryoprotection and then frozen on dry ice. Cryostat sections of 15 µm were collected onto chrom-alum coated slides. For coronal analysis, frozen sections were selected from appropriate regions of the brain: rostral sections were as close to Bregma 0.38 as the series of sections would allow; middle from Bregma −0.70; caudal from Bregma −0.94 [@pone.0017727-Paxinos1]. Sagittal paraffin sections (10 µm) were selected as close to lateral 0.12 mm as the series of sections would allow [@pone.0017727-Paxinos1]. Myelination in cuprizone challenged animals was assessed using luxol fast blue (LFB)-periodic acid Schiff (PAS) reagent using paraffin embedded sections. For electron microscopic evaluation of myelin, mice were perfused as above and processed for resin embedding. Semi-thin sections (0.5 µm) were cut to evaluate quality and orientation. Representative samples were then chosen and ultra-thin sections (90 nm) cut and images captured using a Siemens Stereoskop Transmission Electron Microscope (Siemens, Munich, Germany) at 3000×. Sections corresponding to the aforementioned Bregma locations were imaged and a region of interest (ROI) randomly selected for quantification by an observer blind to genotype and treatment. Quantification and morphometric analysis of axon myelination {#s2i} ------------------------------------------------------------ Images of the appropriate ROI were imported into Image J64 (US National Institutes of Health, Bethesda). The threshold function of Image J64 was used to outline the myelin within each image, with threshold levels adjusted manually. The number of myelinated axons within each ROI was counted manually. The measure function was used to determine both the internal area of each myelinated axon and the area of the whole axon including myelin wrappings. The diameter of each axon and total fibre diameter was calculated mathematically from the internal area or the whole axon area respectively, as the diameter of a circle of equivalent area [@pone.0017727-Stidworthy1]. The *g*-ratio for each axon was calculated as a ratio of axon diameter/fibre diameter. Counts of myelinated axons, axon diameter and g-ratios are expressed as mean± SEM. Quantification of oligodendrocyte and microglia number {#s2j} ------------------------------------------------------ Sections of frozen tissue were prepared as described above. Sections were fixed in 4% paraformaldehyde. Primary antibodies were used at 1/500 and sourced as follows: anti-Olig2 (Millipore, Billerica, MA); anti-IBA1 (Wako Chemicals, Richmond, VA); anti-GFAP (Dako, Carpinteria, CA); anti-PDGFRα (Fitzgerald, Concord, MA); anti-APC(CC1) (EMD Chemicals, Gibbstown. NJ). Appropriate secondary antibodies were used at 1/500 and the final antibody incubation included Hoescht 33342 (1/2000, Invitrogen) to visualise the nuclei of all cells. For cell counts, images were captured from 3 sections for each individual animal using a Carl Zeiss Axioplan microscope at 20× objective. All immunopositive cells in the corpus callosum contained within the imaged region were counted, and the area of the corpus callosum measured using NIH ImageJ64 (US National Institutes of Health, Bethesda). Counts are expressed as number of positive cells/mm^2^± SEM. Quantification of LFB {#s2k} --------------------- Image analysis of LFB stained sections was performed according to procedures previously described [@pone.0017727-Emery1]. Briefly, all images from a given experiment were acquired in a single session using the same light intensity and filter settings with the white balance of all images standardised. Images were captured using a Carl Zeiss Axioplan microscope with a 5× objective. For quantification of LFB density, images were divided into equal thirds corresponding to rostral, middle and caudal segments. The area of the corpus callosum was determined using the NIH ImageJ64 (US National Institutes of Health, Bethesda). Density measurements are expressed as raw mean intensity±SEM. Statistical analysis {#s2l} -------------------- All statistical analysis was performed using GraphPad Prism 5 software (GraphPad Software Inc., La Jolla). Differences between genotypes were compared using Students t-tests. Single-factor, multiple condition experiments were analysed using one-way ANOVA with Bonferroni\'s post-hoc tests. A *p*-value of less than 0.05 was considered to be statistically significant. Results {#s3} ======= Early remyelination is influenced by the absence of Gas6 {#s3a} -------------------------------------------------------- We initially investigated whether the absence of Gas6 affects the rate or level of remyelination following cuprizone-induced demyelination. Given our previous results showing that demyelination is greater in the absence of Gas6 following 3 weeks of cuprizone challenge [@pone.0017727-Binder1], we chose a 5 week time-point of cuprizone challenge to provide nadir levels of demyelination in both Gas6^+/+^ and Gas6^−/−^ mice. After 5 weeks of cuprizone challenge, mice were returned to a normal (cuprizone-free) diet for either 0, 2 or 4 weeks. To analyse the extent of remyelination in wild-type and Gas6 knockout mice, the corpus callosum was segmented lengthwise into 3 equal segments as previously described [@pone.0017727-Binder1]: rostral, middle and caudal (outlined in [Fig. 1A](#pone-0017727-g001){ref-type="fig"}), and the level of myelination assessed using LFB staining as described in the [Materials and Methods](#s2){ref-type="sec"}. Representative images of all groups are shown in [Figure 1 (A--F)](#pone-0017727-g001){ref-type="fig"}. We detected an effect of the loss of Gas6 in the rostral and middle segments, with Gas6^−/−^ mice showing lower myelin density than wild-type mice after 4 weeks of recovery following 5 weeks of cuprizone-challenge, although in the middle segment this was only a trend (rostral p = 0.039, [Fig. 1G](#pone-0017727-g001){ref-type="fig"}; middle p = 0.052, [Fig. 1H](#pone-0017727-g001){ref-type="fig"}). ::: {#pone-0017727-g001 .fig} 10.1371/journal.pone.0017727.g001 Figure 1 ::: {.caption} ###### LFB staining demonstrates a reduction in the extent of remyelination in the absence of Gas6. **A--F** Wild-type and Gas6 KO mice were subjected to cuprizone challenge for 5 weeks followed by recovery in the absence of cuprizone for 0, 2 or 4 weeks. **A**. The corpus callosum was divided into three segments for image analysis: rostral (R), Middle (M) and Caudal (C) **G--I** Myelin density was assessed using image analysis as described in the [Materials and Methods](#s2){ref-type="sec"}. **G**. A significant reduction in myelination was observed in the rostral segment of the corpus callosum (p = 0.039) and a strong trend towards reduction was observed in the middle segment (p = 0.052) (**H**). Scale bar = 1 mm. :::  ::: To clarify the changes in remyelination detected using LFB analysis, we next performed an ultrastructural analysis of all segments using electron microscopy. The number of myelinated axons was determined for both Gas6^+/+^ and Gas6^−/−^ mice, as well as the diameter of each myelinated axon and the thickness of the myelin surrounding each axon. Representative images of each group are shown in [Figure 2 (A--F)](#pone-0017727-g002){ref-type="fig"}. In agreement with the LFB analysis, we detected an effect of the loss of Gas6, with Gas6^−/−^ mice showing fewer myelinated axons than wild-type mice following 4 weeks of recovery. However, in contrast to the LFB analysis this only reached significance in the caudal segment ([Fig. 2I](#pone-0017727-g002){ref-type="fig"}; p = 0.032). ::: {#pone-0017727-g002 .fig} 10.1371/journal.pone.0017727.g002 Figure 2 ::: {.caption} ###### EM analysis demonstrates a reduction in the number of myelinated axons during the course of remyelination in the absence of Gas6. **A--F** Wild-type and Gas6 KO mice were subjected to cuprizone challenge for 5 weeks followed by recovery in the absence of cuprizone for 0, 2 or 4 weeks. The number of myelinated axons per mm^2^ was quantified and shown in **G--I**. A significant reduction in myelination was observed in the caudal (**I**) segments of the corpus callosum in Gas6^−/−^ mice compared with Gas6^+/+^ mice (p = 0.032). Scale bar = 2 µm. :::  ::: In addition to determining the number of myelinated axons in each segment, each myelinated axon was assessed for diameter and myelin thickness (expressed as g-ratio) as described in the [Material and Methods](#s2){ref-type="sec"}, with the data shown in [Table 1](#pone-0017727-t001){ref-type="table"}. No differences between wild-type and Gas6^−/−^ mice were observed in any of the measures, except for a significant increase following 5 weeks of cuprizone challenge in the mean diameter of myelinated axons ([Table 1](#pone-0017727-t001){ref-type="table"}, p = 0.009) and a decrease in the thickness of myelin surrounding the axons in the rostral segment of the corpus callosum of Gas6^−/−^ mice ([Table 1](#pone-0017727-t001){ref-type="table"}, p = 0.007). ::: {#pone-0017727-t001 .table-wrap} 10.1371/journal.pone.0017727.t001 Table 1 ::: {.caption} ###### Axon Diameter and *g*-ratios in Gas6^+/+^ and Gas6^−/−^ mice during remyelination. ::: {#pone-0017727-t001-1} Weeks Recovery Genotype Axon diameter (µm^2^) *g*-ratio ---------------- ----------- ---------------------------------------------- ------------- ------------- ---------------------------------------------- ------------- ------------- 0 Gas6^+/+^ 0.604±0.020 0.781±0.163 0.601±0.041 0.688±0.005 0.703±0.033 0.688±0.014 Gas6^−/−^ 0.718±0.018[\*](#nt101){ref-type="table-fn"} 0.645±0.035 0.597±0.060 0.746±0.010[\*](#nt101){ref-type="table-fn"} 0.684±0.005 0.648±0.026 2 Gas6^+/+^ 0.704±0.039 0.637±0.037 0.588±0.028 0.711±0.010 0.678±0.014 0.651±0.023 Gas6^−/−^ 0.683±0.018 0.657±0.064 0.571±0.062 0.686±0.008 0.715±0.025 0.672±0.017 4 Gas6^+/+^ 0.665±0.046 0.626±0.038 0.585±0.011 0.704±0.019 0.677±0.009 0.662±0.013 Gas6^−/−^ 0.722±0.056 0.598±0.027 0.579±0.015 0.705±0.023 0.689±0.012 0.660±0.002 \*p\<0.05. ::: Taken together, the analysis of myelination in the absence of Gas6 indicates a significant effect on the remyelination process, with an apparently lower level of myelination after 4 weeks of recovery following 5 weeks of cuprizone challenge, with the maximum effect observed in any parameter being a reduction of approximately 43% in the number of myelinated axons in the caudal segment following 4 weeks of recovery (4.68×10^5^ vs 2.70×10^5^ axons/mm^2^ in Gas6^+/+^ vs Gas6^−/−^ mice respectively; [Fig. 2I](#pone-0017727-g002){ref-type="fig"}). The absence of Gas6 delays but does not prevent recovery of myelination {#s3b} ----------------------------------------------------------------------- Given the data showing a decrease in myelination following 4 weeks of recovery, we next examined whether this decrease was chronic or if myelin levels recovered to wild-type levels over time. To test this, an independent cohort of mice was subjected to cuprizone challenge for 5 weeks, and then returned to a cuprizone-free diet for either 0 or 10 weeks. As for the previous cohort, ultrastructural analysis was performed using electron microscopy and the number of myelinated axons per mm^2^ quantified. Representative images of each group are shown in [Figure 3 (A--L)](#pone-0017727-g003){ref-type="fig"}. In contrast to the shorter time-points of recovery, after 10 weeks of recovery following 5 weeks of cuprizone challenge, no significant difference in the number of myelinated axons/mm^2^ was observed between Gas6^−/−^ and wild-type mice in any segment. ::: {#pone-0017727-g003 .fig} 10.1371/journal.pone.0017727.g003 Figure 3 ::: {.caption} ###### EM analysis demonstrates that, following 10 weeks of recovery, Gas6 KO mice remyelinate to the same extent as Gas6 WT mice. **A--L** Wild-type and Gas6 KO mice were subjected to cuprizone challenge for 5 weeks followed by recovery in the absence of cuprizone for 0 or 10 weeks. The number of myelinated axons/mm^2^ was quantified in the rostral, middle and caudal segments of the corpus callosum (**M--O**). No significant differences were observed between Gas6^−/−^ or Gas6^+/+^ mice at any time point or in segment (p\>0.05). Scale bar = 2 µm. :::  ::: In addition, each myelinated axon was assessed for diameter and myelin thickness ([Table 2](#pone-0017727-t002){ref-type="table"}). No difference in either the diameter of myelinated axons or in the thickness of myelin surrounding axons was observed between Gas6^+/+^ and Gas6^−/−^ mice following 10 weeks of remyelination. However, as detected in the previous cohort, small differences were once again observed following 5 weeks of cuprizone challenge, with Gas6−/− mice showing a significant increase in the mean diameter of myelinated axons in the caudal segment ([Table 2](#pone-0017727-t002){ref-type="table"}, p = 0.019), and significantly thinner myelin in the middle segment ([Table 2](#pone-0017727-t002){ref-type="table"}, p = 0.034). ::: {#pone-0017727-t002 .table-wrap} 10.1371/journal.pone.0017727.t002 Table 2 ::: {.caption} ###### Axon Diameter and *g*-ratios in Gas6^+/+^ and Gas6^−/−^ mice during long-term remyelination. ::: {#pone-0017727-t002-2} Weeks Recovery Genotype Axon diameter (µm^2^) *g*-ratio ---------------- ----------- ----------------------- ------------- ---------------------------------------------- ------------- ---------------------------------------------- ------------- 0 Gas6^+/+^ 0.656±0.030 0.688±0.050 0.511±0.032 0.693±0.017 0.711±0.009 0.687±0.021 Gas6^−/−^ 0.588±0.049 0.764±0.129 0.662±0.039[\*](#nt102){ref-type="table-fn"} 0.680±0.012 0.762±0.020[\*](#nt102){ref-type="table-fn"} 0.686±0.009 10 Gas6^+/+^ 0.642±0.018 0.625±0.022 0.504±0.031 0.699±0.008 0.682±0.011 0.653±0.029 Gas6^−/−^ 0.652±0.016 0.595±0.027 0.571±0.044 0.706±0.010 0.669±0.011 0.659±0.014 \*p\<0.05. ::: Taken together, these data indicate that, in contrast to the findings after 4 weeks of recovery following 5 weeks of cuprizone challenge, there is no difference in either the number or size of myelinated axons, or in the thickness of myelin, following 10 weeks recovery. Expression of oligodendrocyte markers is reduced following 4 weeks of remyelination {#s3c} ----------------------------------------------------------------------------------- We next examined whether the delay in the myelination observed in the absence of Gas6 correlated with a reduction in oligodendrocyte numbers. A separate cohort of mice was challenged with cuprizone for 5 weeks and then returned to a cuprizone-free diet for 0, 2, 4 or 10 weeks. Coronal sections were collected as described in the [Materials and Methods](#s2){ref-type="sec"} then stained with anti-APC(CC1) to determine the number of mature oligodendrocytes, with adjacent sections double-stained with anti-Olig2 and anti-PDGFRα to determine the number of oligodendrocyte precursors. The number of positive cells in all 3 segments - rostral, middle and caudal - was determined. Representative images from the middle segment at each time-point are shown in [Figure 4(A)](#pone-0017727-g004){ref-type="fig"}. We observed a reduction in the number of mature oligodendrocytes in the absence of Gas6 following cuprizone challenge, as measured by expression of APC(CC1), with the caudal segment showing a significant loss of APC(CC1) oligodendrocytes following 4 weeks of remyelination ([Fig. 4D](#pone-0017727-g004){ref-type="fig"}, 1331+/−74 cells/mm2 vs 717+/−129 cells/mm^2^, Gas6^+/+^ vs Gas6^−/−^ mice respectively, p = 0.007). Additionally, the middle segment showed a trend towards a reduction in the number of APC(CC1) positive oligodendrocytes following 4 weeks of remyelination ([Fig. 4C](#pone-0017727-g004){ref-type="fig"}, p = 0.067). Furthermore, a strong trend towards a reduction in the number of APC(CC1) positive oligodendrocytes was observed in the caudal segment following 10 weeks of recovery ([Fig. 4D](#pone-0017727-g004){ref-type="fig"}, p = 0.052). ::: {#pone-0017727-g004 .fig} 10.1371/journal.pone.0017727.g004 Figure 4 ::: {.caption} ###### Expression of oligodendrocyte lineage markers are altered in the absence of Gas6 during remyelination. Wild-type and Gas6 KO mice were subjected to cuprizone challenge for 5 weeks followed by recovery in the absence of cuprizone for 0, 2, 4 or 10 weeks. **A**. Representative images of the middle segment of the corpus callosum, showing cells positive for APC(CC1) or double positive for Olig2/PDGFRα. **B--G**. A significant reduction in the number of APC(CC1) positive cells was observed in the caudal segment of the corpus callosum in Gas6^−/−^ mice compared with Gas6^+/+^ mice (p = 0.0070) (**D**). The density of Olig2+/PDGFRα+ OPCs was significantly increased in Gas6^−/−^ mice compared with Gas6^+/+^ mice following both 4 and 10 weeks of remyelination (4 weeks, p = 0.024; 10 weeks, p = 0.003) (**E**). Scale bar = 20 µm. :::  ::: Taken together, these data indicate that the delay in myelination observed in the absence of Gas6 parallels a reduction in the number of APC(CC1) immunopositive oligodendrocytes. In particular, the only segment to show a significant reduction in APC(CC1) positive oligodendrocytes, the caudal segment following 4 weeks remyelination, was also the sole segment that displayed a significant reduction in the number of myelinated axons at the same time-point. To assess whether numbers of OPCs were also affected by Gas6 deficiency, the density of PDGFRα/Olig2 double immunopositive cells was determined. The middle and caudal segments showed no apparent differences in OPC numbers at any time-point ([Fig. 4H,I](#pone-0017727-g004){ref-type="fig"}). In contrast, the rostral segment of the corpus callosum showed a significant effect, with more OPCs observed in the absence of Gas6 following both 4 weeks of remyelination ([Fig. 4E](#pone-0017727-g004){ref-type="fig"}, p = 0.025) and 10 weeks of remyelination ([Fig. 4E](#pone-0017727-g004){ref-type="fig"}, p = 0.003). Reduction in oligodendrocyte markers is not correlated with an increase in numbers of microglia in the absence of Gas6 {#s3d} ---------------------------------------------------------------------------------------------------------------------- We next wished to determine if the reduction of oligodendrocyte markers in the absence of Gas6 was correlated with an increase in the number of microglia. We therefore assessed the number of IBA1 positive microglia in mice after 5 weeks of cuprizone challenge and following 0, 2, 4 or 10 weeks recovery in the absence of cuprizone. The number of IBA1 positive cells was determined in both Gas6^−/−^ and Gas6^+/+^ mice in all three segments of the corpus callosum - rostral, middle and caudal. Representative images of the middle segment are shown in [figure 5A](#pone-0017727-g005){ref-type="fig"}. No significant differences were observed between Gas6^−/−^ and Gas6^+/+^ mice at any time-point examined, with the exception that an increase in the number of IBA1 positive microglia was seen in the middle segment following 10 weeks of remyelination ([Fig. 5C](#pone-0017727-g005){ref-type="fig"}, p = 0.013). However, as no difference was observed in the number of APC(CC1) positive oligodendrocytes at this time in this segment, these data indicate that the increased loss of oligodendrocyte markers observed in the absence of Gas6^−/−^ is not directly linked to an increase in the number of microglia. ::: {#pone-0017727-g005 .fig} 10.1371/journal.pone.0017727.g005 Figure 5 ::: {.caption} ###### The number of IBA1 positive microglia is minimally altered by the absence of Gas6 during remyelination. Wild-type and Gas6 KO mice were subjected to cuprizone challenge for 5 weeks followed by recovery in the absence of cuprizone for 0, 2, 4 or 10 weeks. **A**. Representative images of the middle segment of the corpus callosum, showing cells positive for IBA1. **B--D** The number of IBA1 positive cells/mm2 was quantified. A significant increase in the number of IBA1-positive microglia was observed in Gas6^−/−^ mice compared with Gas6^+/+^ mice following 10 weeks of remyelination (p = 0.031). Scale bar = 20 µm. :::  ::: Exogenous Gas6 can directly increase myelination of neurons *in vitro* {#s3e} ---------------------------------------------------------------------- We additionally wished to assess if, in addition to affecting numbers of oligodendrocytes, Gas6 could also directly affect the capacity of oligodendrocytes to myelinate axons. In order to assess this, we employed an *in vitro* system, co-culturing OPCs with retinal ganglion cells in the presence or absence of exogenous rhGas6, as described in the [Materials and Methods](#s2){ref-type="sec"}. Upon addition of rhGas6, a dose-dependent significant increase in the number of myelin basic protein (MBP)-positive segments was observed ([Figure 6 A--E](#pone-0017727-g006){ref-type="fig"}, p\<0.0001). This increase in MBP-positive segments paralleled an increase in the expression of myelin components as determined using Western analysis for CNPase, myelin associated glycoprotein (MAG) and MBP. This increase in myelination did not appear to be a result of improved viability of OPCs in this assay, as no improvement in survival of cells could be detected at 2, 7 or 14 days in the presence of exogenous Gas6 (data not shown). These data indicate that exogenous Gas6 can directly increase the number of myelinated axons in an *in vitro* culture system, suggesting that the delay observed in recovery from cuprizone-challenge in the absence of Gas6 may be partially a result of a deficit in myelination, rather than simply reflecting a loss of oligodendrocytes. ::: {#pone-0017727-g006 .fig} 10.1371/journal.pone.0017727.g006 Figure 6 ::: {.caption} ###### Exogenous Gas6 increases the number of myelinated DRG neurons when co-cultured with oligodendrocytes. **A--D** Oligodendrocytes and DRG neurons were co-cultured as described in the [Materials and Methods](#s2){ref-type="sec"}. Exogenous Gas6 was added to the cultures at either 0 ng/ml (A), 1 ng/ml (B), 10 ng/ml (C) or 100 ng/ml (D) and the number of myelinated segments quantified, with the results shown in (E). Exogenous Gas6 significantly increased the number of myelinated segments in a dose-dependant manner (p\<0.0001). **F** Western analysis showing exogenous Gas6 increased the expression of myelin compenents. Scale bar = 100 µm. :::  ::: Discussion {#s4} ========== In this study we show that the absence of Gas6 affects the efficiency of remyelination following a demyelinating insult, induced by cuprizone, such that remyelination is delayed, although not ultimately prevented. This delay in remyelination is paralleled by a significant decrease in the number of mature oligodendrocytes. This decrease in oligodendrocytes was specifically observed in the caudal corpus callosum during remyelination, the segment of the corpus callosum most affected by cuprizone. Interestingly, this reduction in oligodendrocyte markers is not correlated with an increase in the number of microglia in the absence of Gas6, suggesting a direct effect of the absence of Gas6 on oligodendrocytes in the remyelination phase. In support of this, we show that exogenous Gas6 can directly enhance the myelination of DRG neurons *in vitro* by oligodendrocytes. Collectively, these data indicate that Gas6 and TAM receptor signalling plays an important role in the process of remyelination following a demyelinating insult, and that this effect could be transduced, at least in part, via a direct effect of Gas6 on myelination by oligodendrocytes. In demyelinating diseases such as MS, whilst remyelination occurs, it is often inadequate or eventually fails. This failure is thought to result from a variety of factors, including a deficiency of OPCs, or the failure of these OPCs to reach the correct target, or alternatively, failure of differentiation of OPCs resident within lesions (reviewed in [@pone.0017727-Franklin1]). In this study, we show that in the absence of Gas6 there is a significant delay in the remyelination process following cuprizone-induced demyelination. Although remyelination is not ultimately prevented, the decrease in the efficiency of remyelination observed in the absence of Gas6 could lead to an increase in the vulnerability of axons [@pone.0017727-Irvine1], [@pone.0017727-Irvine2]. As the loss of axons has been shown to be closely related to disability in MS (reviewed in [@pone.0017727-Trapp1]), the apparent role of Gas6 in increasing the efficiency of remyelination could be an important factor in preventing or reducing axon loss. What mediates this delay in remyelination? As indicated above, a number of factors has been postulated to be important in ensuring the efficiency of remyelination, not least of which is the number and localisation of OPCs. In this study, we have shown that whilst there were some changes in the number of PDGFRα/Olig2 positive OPCs in mice in the absence of Gas6 compared with WT mice, specifically in the rostral segment of the corpus callosum, these changes did not clearly correlate with the changes observed in remyelination efficiency. This suggests that the effect of the loss of Gas6 upon remyelination is not upon either the number or recruitment of OPCs to the damaged region. This further suggests that the loss of Gas6 has a direct effect upon either extant or new mature oligodendrocytes during remyelination. Our examination of mature oligodendrocytes showed that the reduction in myelination observed following 4 weeks of recovery paralleled a significant reduction in the expression of the oligodendrocyte marker APC(CC1). It is not possible to determine from these data whether the loss of marker expression indicates an increase in the death of mature oligodendrocytes or an increase in dysfunction following initial recovery. We and others have previously shown that Gas6 is an important anti-apoptotic signal for oligodendrocytes [@pone.0017727-Binder1], [@pone.0017727-Shankar1], [@pone.0017727-Shankar2], and it is possible that a reduction in survival signals could lead to the observed effects. However, recent work from Hesse *et al.* [@pone.0017727-Hesse1] indicates that the vast majority of oligodendrocytes undergo apoptosis during the first 3 weeks of cuprizone challenge. Given that differences in the number of mature oligodendrocyte markers between Gas6^−/−^ and Gas6^+/+^ were not observed until 4 weeks post-cuprizone challenge, and that prior to this time the number of both APC(CC1) and Olig2 positive oligodendrocytes was equal between both genotypes, it seems unlikely that the differences observed are the result of cell death, but are more likely to result from either a dysregulation in the expression of these markers and/or in the maturation of these cells. The mature oligodendrocyte marker APC(CC1) is an antagonist of *β-*catenin and results in a reduction in Wnt signalling [@pone.0017727-Papkoff1]. It has previously been shown by Fancy *et al.* (2009) that whilst Wnt-*β*-catenin signalling is upregulated during normal developmental myelination and during remyelination, the timely inhibition of the Wnt-*β-*catenin signalling pathway is required or remyelination is blocked [@pone.0017727-Fancy1]. We observed a significant reduction in APC(CC1) expression in the caudal segment of the corpus callosum and a strong trend for a reduction in APC expression in the middle segment of the corpus callosum after 4 weeks of recovery following cuprizone challenge, the time at which we observed a reduction in myelin density and a reduction in myelinated axons in the absence of Gas6. These findings suggest the intriguing hypothesis that dysregulation in the Wnt-*β*-catenin signalling pathway could be responsible for the reduced efficiency of remyelination observed in the absence of Gas6, a possibility that merits specific interrogation in future experiments. It is also possible that the observed effects on remyelination could result from other effects of the absence of Gas6 on myelination. These effects could be either direct or indirect. We have previously shown that during demyelination in the absence of Gas6, there was both a loss of oligodendrocytes and a concomitant increase in the number of IBA1-positive oligodendrocytes in the corpus callosum following 3 weeks of cuprizone challenge [@pone.0017727-Binder1]. However, it was not clear whether the increase in oligodendrocyte loss was a primary effect of the absence of Gas6, or whether it was secondary to the increase in microglial numbers. Hoehn *et al.* (2008) examined a number of time-points in WT and Axl KO cuprizone-mediated demyelination and, in contrast to the findings in Gas6 KO mice, observed a significant decrease in microglia in Axl KO mice after 4 weeks of cuprizone challenge. There was also a delay in phagocytosis and in the removal of myelin debris in these mice [@pone.0017727-Hoehn1]. Here we show that the reduction in mature oligodendrocyte markers in Gas6^−/−^ mice occurs in the absence of a differential increase in numbers of IBA1 positive microglia, suggesting a direct effect of the absence of Gas6 on mature oligodendrocytes. Our *in vitro* data showing that exogenous Gas6 can increase the myelination of DRG neurons by OPCs is also strongly suggestive of a direct effect of Gas6 on myelination by oligodendrocytes following a demyelinating challenge. The apparent disparity in the results between the Axl^−/−^ mice and the Gas6^−/−^ mice could reflect the ability of alternate members of the TAM receptor family to compensate for the loss of Axl, whereas the overall reduction in signalling that accompanies the loss of an important ligand could outstrip the ability of, for example, ProS to compensate for this loss. The results presented in this study indicate an important role for Gas6 and TAM receptor signalling during remyelination, in addition to the known role of this signalling pathway during demyelination. In the absence of Gas6, remyelination is slowed, potentially increasing the length of time to which axons are exposed to damage. Additional work is required to determine the mechanisms by which remyelination efficiency is reduced in the absence of Gas6, in particular whether Gas6 can directly increase myelination by oligodendrocytes *in vivo* as well as *in vitro*. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was supported by a grant from the National Health and Medical Research Council Australia (grant number 566830). GZMM is supported by an Australian Postgraduate Award. JX is the recipient of a Betty Cuthbert Fellowship (\#454330) from National Health and Medical Research Council Australia and Multiple Sclerosis Research Australia. The Florey Neuroscience Institutes is supported by funding provided from the Operational Infrastructure Scheme of the Department of Innovation, Industry and Regional Development, Victoria, Australia. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: MB JX SM TK. Performed the experiments: MB JX DK GM. Analyzed the data: MB JX DK GM. Contributed reagents/materials/analysis tools: MB JX DK SM GM TK. Wrote the paper: MB JX SM TK.

PubMed Central

2024-06-05T04:04:19.840155

2011-3-10

PMC3053386

Introduction {#s1} ============ Direct, pair-wise (binary), physical protein-protein interactions (PPIs) are the foundation of all biological processes. Efforts to elucidate the interaction network of all proteins within a cell or organism --- termed the interactome --- has helped identify the architectural and functional blueprint of cellular processes in various model eukaryotic organisms, such as yeast [@pone.0017701-Uetz1]--[@pone.0017701-Hazbun1], *Drosophila* [@pone.0017701-Giot1]--[@pone.0017701-Stanyon1], *C. elegans* [@pone.0017701-Simonis1]--[@pone.0017701-Walhout1], *Plasmodium* [@pone.0017701-LaCount1], *Arabidopsis* [@pone.0017701-Boruc1]--[@pone.0017701-Hackbusch1], mouse [@pone.0017701-Suzuki1] and humans [@pone.0017701-Rual1]--[@pone.0017701-Colland1]. Mapping PPIs has forwarded our understanding of key biological processes such as the mitotic spindle [@pone.0017701-Wong1], cell polarity [@pone.0017701-Drees1], the proteasome [@pone.0017701-Cagney1] and the editosome [@pone.0017701-Schnaufer1]. Furthermore, it has helped assign roles to proteins of previously unknown function [@pone.0017701-Hazbun1] and has increased our understanding of and progress against human diseases [@pone.0017701-Lim1]--[@pone.0017701-Goehler1]. There are two main methods of observing direct PPIs *in vivo*: the yeast two-hybrid (Y2H) and its many derivatives [@pone.0017701-Fields1] and more recently, the protein-fragment complementation assay (PCA) [@pone.0017701-Michnick1]. In the Y2H, the interaction of bait and prey fusion proteins within the nucleus reconstitutes a transcription factor that up-regulates the expression of a reporter gene. PCA works similarly to the Y2H but occurs in the cytoplasm and replaces the transcription-reporter system with a reconstituted reporter protein capable of metabolizing a toxic compound. The PPIs of the yeast *Saccharomyces cerevisiae* have been extensively explored. There are currently three genome-wide high-throughput yeast two-hybrid (HT-Y2H) surveys [@pone.0017701-Uetz1]--[@pone.0017701-Yu1] and one genome-wide PCA study of the yeast interactome [@pone.0017701-Tarassov1]. However, while these large-scale Y2H and PCA screening projects have established proteome-wide protein interaction networks (PINs) for yeast, statistical analysis reveals that their combined datasets account for less than 30% of the entire yeast interactome [@pone.0017701-Yu1]. Furthermore, there is surprisingly little overlap of PPIs between each of the four aforementioned studies and with the literature-curated (LC) interaction dataset. The LC data, which are derived from small scale Y2H studies (otherwise known as the "community" dataset) displays a narrow focus on a few proteins or an interactome sub-network. Despite recent reports to the contrary [@pone.0017701-Venkatesan1], [@pone.0017701-Cusick1]--[@pone.0017701-Dreze1], the LC dataset is commonly believed to be of higher quality than the HT-Y2H interactions due to its narrow focus on the PPIs of a few well-characterized proteins [@pone.0017701-Salwinski1]--[@pone.0017701-Mrowka1]. Furthermore, LC studies often report reciprocal interactions (bidirectional interactions where proteins A and B interact as either bait or prey), recapitulate their results via multiple independent orthogonal methods and integrate their findings with other forms of biochemical and genetic data [@pone.0017701-Boulon1]--[@pone.0017701-Charette1]. The poor PPI overlap among the large-scale screens and with the LC dataset has led to the suggestion that the current HT-Y2H studies were not done to saturation, and therefore must be missing additional interactions [@pone.0017701-Koegl1]. This may be due to a number of reasons. First, most genome-wide HT-Y2H studies do not include all of the protein-coding genes in the yeast genome. The absence of even a few proteins from HT-Y2H screens can significantly reduce interactome coverage [@pone.0017701-Yu1]. Also, the enormous scope of genome-wide HT-Y2H screens often necessitates a pooling strategy in which up to 96 or more baits or preys are pooled then tested for interaction. However, when pooled, proteins that are toxic when expressed at high levels may display a dominant negative phenotype and interactions involving weakly expressed proteins may be under-reported [@pone.0017701-Koegl1]. Similarly, certain proteins may be inefficiently imported into the nucleus, the site of the Y2H assay. Furthermore, PPIs that are not physiologically relevant (the so called "biological false-positives") may be obtained for proteins normally residing in different cellular compartments, expressed at different stages of the cell cycle or in different tissues. These confounding factors are believed to result in pooled HT-Y2H screening strategies being less sensitive than array-based one-by-one screens, while potentially containing a higher number of false positive interactions [@pone.0017701-Koegl1], [@pone.0017701-Rajagopala2]. We focused on mapping the PPIs of the small subunit (SSU) processome, a very large ribonucleoprotein complex comprised of ∼72 proteins and the U3 small nucleolar RNA (snoRNA). This biochemically well defined complex guides the endonucleolytic processing events at sites A~0~, A~1~ and A~2~ that liberate the mature 18S rRNA from the pre-rRNA transcript [@pone.0017701-Dragon1]--[@pone.0017701-Phipps1]. The SSU processome is also believed to chaperone the folding of the pre-18S rRNA and its assembly with ribosomal proteins into the mature SSU of the ribosome. The SSU processome was originally identified by tandem affinity purification followed by mass spectrometry (TAP/MS) studies [@pone.0017701-Dragon1]--[@pone.0017701-Bernstein1], [@pone.0017701-Grandi1]. Subsequent TAP/MS studies expanded the list of SSU processome protein components and provided some of the first data on the presence of sub-complexes [@pone.0017701-Dosil1]--[@pone.0017701-Rudra1]. In all, nearly 70% of all SSU processome proteins have been identified by TAP/MS studies [@pone.0017701-Dragon1]--[@pone.0017701-Bernstein1], [@pone.0017701-Dosil1]--[@pone.0017701-Rudra1], with the remaining proteins being identified by other biochemical or genetic methods. Thus, TAP/MS studies have significantly contributed to our current, nearly complete list of the protein constituents of the SSU processome [@pone.0017701-Dragon1]--[@pone.0017701-Bernstein1], [@pone.0017701-Dosil1]--[@pone.0017701-Rudra1]. Typically, SSU processome protein components meet the following criteria: *i*) they reside in the nucleolus, the site of ribosome biogenesis, *ii*) their genetic depletion results in an 18S rRNA processing defect and *iii*) they co-immunoprecipitate the U3 snoRNA and/or another SSU processome protein component. There are currently 46 confirmed SSU processome proteins and 26 potential candidates suggested from partial data ([Table S1](#pone.0017701.s001){ref-type="supplementary-material"}). Some of these proteins have been categorized into the t-Utp/UtpA, UtpB, UtpC, Mpp10, Rcl1/Bms1 and U3 snoRNP sub-complexes by TAP tag co-complex purifications and small-scale Y2H studies [@pone.0017701-Champion1]--[@pone.0017701-Freed1], [@pone.0017701-Lee1], [@pone.0017701-Wegierski1], [@pone.0017701-Dosil1]--[@pone.0017701-Rudra1]. However, the majority of SSU processome proteins remain unassigned to a specific subcomplex due to a lack of interaction data. Some proteins may even be components of subcomplexes yet to be identified ([Table S1](#pone.0017701.s001){ref-type="supplementary-material"}). Identifying the protein-protein interactions of the SSU processome thus becomes the next step in elucidating its assembly, mechanism of function and regulation in pre-rRNA processing. Considering the SSU processome\'s well characterized and nearly complete component list, we sought to generate an up-to-date, comprehensive yeast SSU processome PIN by extracting and pooling protein interaction data from existing datasets. After retrieving both high-throughput and literature-curated binary protein interaction data, an interaction map was drawn using Cytoscape. The result is the most current protein interactome map of the yeast SSU processome to date, from which we identify additional interactions within the subcomplexes and some of the first potential interactions linking the various subcomplexes. Materials and Methods {#s2} ===================== Mining databases for known PPIs {#s2a} ------------------------------- For each SSU processome component, both IntAct (<http://www.ebi.ac.uk/intact/>) [@pone.0017701-Aranda1] and BioGRID (<http://thebiogrid.org/>) [@pone.0017701-Breitkreutz1] databases were queried for protein-protein interaction data. These repositories were chosen because they: *i*) provide downloadable data in a tab delimited format for every queried protein, *ii*) each contain PPIs from a different subset of genome-wide high-throughput studies, *iii*) each include PPIs from a different subset of LC studies, *iv*) pool interaction data from various organism-specific databases and *v*) are updated on a monthly basis to include novel interactions. We downloaded a total of 72 files from both IntAct and BioGRID databases, one for each of the 72 SSU processome proteins, totaling 144 spreadsheets by November 5, 2010. These files contained all known interactors --- both binary and co-complex --- for the query protein, the experimental method used to detect the interaction and the publication reference. Organizing the data {#s2b} ------------------- All 144 spreadsheets underwent five editing stages to remove information unnecessary to this study and were streamlined into six columns: Bait, Prey, Experimental System (Y2H, Y2H array, Y2H pooling approach, PCA), Literature Code (Uetz *et al.* [@pone.0017701-Uetz1], Ito *et al.* [@pone.0017701-Ito1], Yu *et al.* [@pone.0017701-Yu1], Hazbun *et al.* [@pone.0017701-Hazbun1], PCA [@pone.0017701-Tarassov1] or LC [@pone.0017701-Boulon1]--[@pone.0017701-Charette1]), Organism (yeast, *Drosophila* and *C. elegans*) and Reference. ### Edit Stage 1 {#s2b1} Data were sorted by experimental methods; non-Y2H and non-PCA derived PPIs were removed. For IntAct files, deleted examples include "tandem affinity purification" and "inferred by author" methods, and for BioGRID, they include "Affinity Capture-MS", "Phenotypic Enhancement" and "Synthetic Lethality". Interactions where neither the bait nor the prey represented the query protein were also removed. The IntAct files also included PPI for non-yeast organisms. These data were extracted and edited separately. ### Edit Stage 2 {#s2b2} Proteins with missing names were labeled with the "Standard Name" [@pone.0017701-IsselTarver1], and all names were kept congruent between IntAct and BioGRID files. Proteins with multiple aliases were labeled with the name most commonly used in literature (*e.g.*, Sas10 was re-named Utp3 and Sik1was re-named Nop56). ### Edit Stage 3 {#s2b3} Columns with information irrelevant to our study were deleted from both sets of data files. For IntAct, 32 data columns were reduced to five columns: bait ID, prey ID, interaction detection method, source (author) and PubMed ID. We also removed the extra columns from BioGRID, cutting nine columns down to the same five of the IntAct files. ### Edit Stage 4 {#s2b4} The 72 BioGRID and 72 IntAct files were merged into one large spreadsheet and duplicates entries were removed. These included identical interactions with the same experimental method and authors, a consequence of some, but not all interactions being reported in both BioGRID and IntAct. However, duplicate interactions identified via different experimental methods or by different research groups were kept. ### Edit Stage 5 {#s2b5} All interactions involving only one SSU processome component (*i.e.*, interactions between an SSU processome component and a non-SSU processome protein) were removed as a function of the SSU processome protein components having been relatively well catalogued biochemically. A "Literature Code" column was added to separate the data into Uetz *et al.* [@pone.0017701-Uetz1], Ito *et al.* [@pone.0017701-Ito1], Yu *et al.* [@pone.0017701-Yu1], Hazbun *et al.* [@pone.0017701-Hazbun1], PCA [@pone.0017701-Tarassov1] and LC [@pone.0017701-Boulon1]--[@pone.0017701-Charette1] categories. Completion of all edit stages resulted in one master spreadsheet containing all the query proteins (bait), their interactors (prey), the experimental system used, the literature code, the source organism and the reference ([Table S2](#pone.0017701.s002){ref-type="supplementary-material"}). Interologues -- conserved SSU processome PPIs in other species {#s2c} -------------------------------------------------------------- All downloaded IntAct files also included protein-protein interactions for *C. elegans*, *D. melanogaster*, *H. sapiens*, *S. pombe*, *P. falciparum* and *M. musculus*. Y2H interactions from organisms other than *S. cerevisiae* (non-yeast) were quarantined during *Edit Stage 1* and underwent the remaining editing stages separately. BioGRID pre-categorizes interactions by organism; PPIs for non-yeast organisms were downloaded separately and edited as described above. In *Edit Stage 5* following the IntAct and BioGRID merge, an "Organism" column was added to the master spreadsheet to enable sorting of yeast and non-yeast data. Protein nomenclature specific to the source organism was queried in Homologene (<http://www.ncbi.nlm.nih.gov/sites/homologene>) [@pone.0017701-Sayers1] to determine the *S. cerevisiae* homologue. Proteins with available Homologene data were renamed as the *S. cerevisiae* homolog (*e.g.*, *D. melanogaste*r CG13097 renamed Mpp10). BLAST analysis [@pone.0017701-Altschul1] was used to identify the yeast homologues of non-yeast proteins not annotated in Homologene [@pone.0017701-Sayers1]. As with the yeast datasets, only PPIs both involving SSU processome components were kept. Visualizing the interactome {#s2d} --------------------------- We used Cytoscape [@pone.0017701-Killcoyne1], a bioinformatics software used to visualize molecular interaction networks, to convert the spreadsheet files to interactome maps. Nodes refer to proteins and are labeled with the protein\'s commonly used name. Edges connect two nodes, illustrating a protein-protein interaction. We distinguished in different colored nodes the various known subcomplexes of the SSU processome (see [Table S1](#pone.0017701.s001){ref-type="supplementary-material"}; green for the t-Utp/UtpA subcomplex, blue for UtpB, yellow for UtpC, gray for the U3 snoRNP proteins, brown for the Bms1/Rcl1 subcomplex and red for Mpp10 subcomplex) and labeled the proteins unassigned to a subcomplex in pink. The numerous RNA helicases of the SSU processome are depicted as diamonds. Cytoscape maps were generated for the SSU processome protein interactions from the Uetz *et al.* [@pone.0017701-Uetz1], Ito *et al.* [@pone.0017701-Ito1], Yu *et al.* [@pone.0017701-Yu1], Hazbun *et al.* [@pone.0017701-Hazbun1], Tarassov *et al.* [@pone.0017701-Tarassov1] and literature-curated datasets [@pone.0017701-Boulon1]--[@pone.0017701-Charette1]. An additional Cytoscape map was drawn for the merged dataset and included SSU processome interologues. Protein motif and domain identification {#s2e} --------------------------------------- The motifs and domains present in the SSU processome proteins were identified using the SCOP Superfamily (<http://supfam.org/SUPERFAMILY/index.html>) [@pone.0017701-Gough1], the MIPS Comprehensive Yeast Genome Database (<http://mips.helmholtz-muenchen.de/genre/proj/yeast/>) [@pone.0017701-Guldener1], Pfam domains (<http://pfam.sanger.ac.uk/>) [@pone.0017701-Finn1], PROSITE (<http://ca.expasy.org/prosite/>) [@pone.0017701-Sigrist1], SMART (<http://smart.embl-heidelberg.de/>) [@pone.0017701-Letunic1] and the Conserved Domain Database at NCBI (<http://www.ncbi.nlm.nih.gov/Structure/cdd/wrpsb.cgi>) [@pone.0017701-MarchlerBauer1]. Results {#s3} ======= Mining databases for known SSU processome protein-protein interactions {#s3a} ---------------------------------------------------------------------- We aimed to assemble a protein-protein interaction map of the yeast SSU processome from existing datasets. Three HT-Y2H studies [@pone.0017701-Uetz1]--[@pone.0017701-Yu1], one PCA dataset [@pone.0017701-Tarassov1] and many small-scale LC studies [@pone.0017701-Boulon1]--[@pone.0017701-Charette1] were queried for PPIs involving the 72 SSU processome proteins. For each protein, one set of data from BioGRID [@pone.0017701-Breitkreutz1] and one from IntAct [@pone.0017701-Aranda1] were downloaded, totaling 144 spreadsheets for the 72 processome proteins. The files were curated to remove interaction detection methods that were neither Y2H nor PCA, such as TAP-Tag, mass spectrometry and genetic interactions. Furthermore, since the list of protein components of the SSU processome has been well characterized [@pone.0017701-Dragon1]--[@pone.0017701-Bernstein1], [@pone.0017701-Grandi1]--[@pone.0017701-Rudra1], and is believed to be nearly complete, we also discarded interactions involving non-SSU processome proteins. Most of the PPIs involving non-SSU processome components were with proteins that are poorly characterized, not nucleolar or with no known role in ribosome biogenesis. While deleting these proteins from our analyses may have resulted in the loss of important interactions or potentially novel SSU processome members, we limited our study to nucleolar proteins involved in ribosome biogenesis or known to co-immunoprecipitate other SSU processome constituents such as the U3 snoRNP. The spreadsheets for each SSU processome protein were merged into a master file and duplicate entries originating from PPIs listed in both BioGRID and IntAct databases were removed ([Table S2](#pone.0017701.s002){ref-type="supplementary-material"}). The master spreadsheet was sorted by study (Literature Code) to determine how many of the protein interactions for the 72 SSU processome proteins are attributed to each of the three HT-Y2H studies [@pone.0017701-Uetz1]--[@pone.0017701-Yu1], the PCA dataset [@pone.0017701-Tarassov1] and the small-scale LC studies [@pone.0017701-Boulon1]--[@pone.0017701-Charette1]. An interactome map was drawn using Cytoscape [@pone.0017701-Killcoyne1] for each dataset to show the extent of SSU processome coverage per study. Finally, the merged master spreadsheet was converted to a Cytoscape map to illustrate the most up-to-date interactome of the 72 SSU processome proteins. Expert curation of protein-protein interaction datasets is often required {#s3b} ------------------------------------------------------------------------- We initially explored a variety of different PPI databases, including BioGRID [@pone.0017701-Breitkreutz1], IntAct [@pone.0017701-Aranda1], MIPS Mpact [@pone.0017701-Guldener2], DIP [@pone.0017701-Salwinski2], STRING [@pone.0017701-Jensen1] and SPIDer [@pone.0017701-Wu1]. Our survey found that BioGRID and IntAct contained the most complete and up-to-date PPIs, with the other databases containing non-overlapping subsets of the HT-Y2H, PCA and LC datasets. We did, however, identify a number of problems with both the BioGRID and IntAct datasets. Although BioGRID is continuously updated, some published Y2H interactions have yet to be included in the database (as of January 2011), such as the Y2H interactions of the UtpB subcomplex published by Champion *et al.* [@pone.0017701-Champion1] in November 2008. Thus, BioGRID does not contain a complete inventory of all currently known PPIs. In some instances, the IntAct database had difficulty filtering and reporting interactions involving only the queried protein due to nomenclature conflicts. For example, a query of the proteins Imp3 ("Interacts with Mpp10 \#3") or Imp4 ("Interacts with Mpp10 \#4") retrieved the appropriate PPIs and erroneous included additional PPIs between Mpp10 and other proteins. Furthermore, a few PPIs from one database were absent in the other, such as the interaction between Utp20 and Sof1 reported by Tarassov *et al.* [@pone.0017701-Tarassov1], which is included in the IntAct database, but not found in BioGRID. Thus, assembling an interactome from current datasets without expert curation is likely to result in an incorrect protein-protein interaction map. Sparse coverage of SSU processome proteins from the three genome-wide HT-Y2H studies {#s3c} ------------------------------------------------------------------------------------ Mining the three genome-wide HT-Y2H datasets for PPIs among SSU processome components revealed disappointingly sparse coverage. The Uetz *et al.* study (2000) [@pone.0017701-Uetz1], which was the first comprehensive HT-Y2H, screened DNA binding domain fusion clones (baits) against both an array and a pool of activation domain fusion clones (preys). For the SSU processome, this yielded five interactions among six of the 72 proteins, as well as one self-interaction for Ckb2 ([Fig. 1A](#pone-0017701-g001){ref-type="fig"} and [Table 1](#pone-0017701-t001){ref-type="table"}) [@pone.0017701-Uetz1]. The Ito *et al.* study [@pone.0017701-Ito1], published in 2001, assembled a yeast interactome by assaying for interactions between the approximately 6,000 proteins of yeast. Sixty-two mating crosses of bait and prey pools were performed with each pool containing 96 different clones as either bait or prey. Their interactions were divided into higher quality "Core" and lower quality "Full" datasets: the former included only the interactions observed 3+ times, while the latter included interactions observed two times. The Ito *et al.* study [@pone.0017701-Ito1] identified four interactions among six of the 72 SSU processome proteins, all from the lower quality "Full" dataset ([Fig. 1B](#pone-0017701-g001){ref-type="fig"} and [Table 1](#pone-0017701-t001){ref-type="table"}). The most recent and third genome-wide HT-Y2H assay, the Yu *et al.* study (October 2008) [@pone.0017701-Yu1], screened individual baits against pools of 188 different preys. Their dataset revealed only one PPI between two of the 72 SSU processome proteins, Utp18 and Utp21 ([Fig. 1C](#pone-0017701-g001){ref-type="fig"} and [Table 1](#pone-0017701-t001){ref-type="table"}). This interaction had previously been identified in the Ito *et al.* dataset ([Fig. 1B](#pone-0017701-g001){ref-type="fig"}) [@pone.0017701-Ito1]. Thus, among the three HT-Y2H datasets, the Uetz *et al.* [@pone.0017701-Uetz1] and Ito *et al.* [@pone.0017701-Ito1] studies provide the highest coverage of PPIs for SSU processome proteins ([Fig. 1A, B, C](#pone-0017701-g001){ref-type="fig"} and [Table 1](#pone-0017701-t001){ref-type="table"}). In all, the three genome-wide HT-Y2H studies account for interactions among only 12 of the 72 SSU processome components (16.7%) and show minimal overlap with the exception of the Utp18-Utp21 interaction reported by Ito *et al.* [@pone.0017701-Ito1] and Yu *et al.* [@pone.0017701-Yu1]. ::: {#pone-0017701-g001 .fig} 10.1371/journal.pone.0017701.g001 Figure 1 ::: {.caption} ###### Interaction maps of the SSU processome proteins from existing HT-Y2H datasets. Proteins are colored as described in the [Materials and Methods](#s2){ref-type="sec"}; green nodes refer to proteins of the t-Utp/UtpA subcomplex, blue for UtpB, yellow for UtpC, gray for the U3 snoRNP proteins, brown for Bms1/Rcl1 and red for the Mpp10 subcomplex. Pink nodes refer to proteins that have yet to be assigned to a subcomplex. RNA helicases are depicted as diamonds. Multiple edges, or interactions, linking the proteins represent interactions identified in different studies or reciprocally identified as both bait and prey. Self-interactions are shown as looped edges. **A**) Results from the Uetz *et al.* dataset [@pone.0017701-Uetz1]. **B**) Results from Ito *et al.* dataset [@pone.0017701-Ito1]. **C**) Results from the Hazbun *et al.* dataset [@pone.0017701-Hazbun1]. **D**) Results from the Yu *et al.* dataset [@pone.0017701-Yu1]. :::  ::: ::: {#pone-0017701-t001 .table-wrap} 10.1371/journal.pone.0017701.t001 Table 1 ::: {.caption} ###### Number of SSU processome proteins (nodes) and the interactions between them (edges) identified in the HT-Y2H, PCA and LC datasets. ::: {#pone-0017701-t001-1} Screen \# of nodes \% of nodes \# of edges \% of predicted edges ----------------------------------------------------------------------------- ------------- ------------- ------------- ----------------------- Uetz *et al.* [@pone.0017701-Uetz1] (Y2H) 6 8.3 6 2.8 Ito *et al.* [@pone.0017701-Ito1] (Y2H) 6 8.3 4 1.9 Yu *et al.* [@pone.0017701-Yu1] (Y2H) 2 2.8 1 0.5 Hazbun *et al.* [@pone.0017701-Hazbun1] (Y2H) 3 4.2 2 2.7 Tarassov *et al.* [@pone.0017701-Tarassov1] (PCA) 25 34.7 27 12.5 Literature-curated [@pone.0017701-Boulon1]--[@pone.0017701-Charette1] (Y2H) 34 47.2 44 20.4 All merged (not including standalone proteins) 46 63.9 67 31.0 Predicted total 72 100 216 100 Redundant edges were not counted twice. Self-interactions, shown as looped edges in [Figs. 1](#pone-0017701-g001){ref-type="fig"} to [](#pone-0017701-g002){ref-type="fig"} [](#pone-0017701-g003){ref-type="fig"} [4](#pone-0017701-g004){ref-type="fig"}, were included in the tabulation. The predicted total number of edges is derived by estimating 3 interactions per protein [@pone.0017701-Yu1] for each of the 72 SSU processome proteins (72×3 = 216). ::: A systems biology study by Hazbun *et al.* (2003) [@pone.0017701-Hazbun1] used the Y2H methodology to help assign roles to yeast proteins of unknown function. This study individually screened each of 100 essential ORFs of unknown function as baits against an array of approximately 6,000 prey ORFs. From this dataset, we identified three of the 72 SSU processome proteins and two PPIs among them ([Fig. 1D](#pone-0017701-g001){ref-type="fig"} and [Table 1](#pone-0017701-t001){ref-type="table"}), with no data overlap with any of the three HT-Y2H studies. The genome-wide PCA study contains the best coverage of SSU processome PPIs {#s3d} --------------------------------------------------------------------------- The protein fragment complementation assay is an alternative method for identifying direct, physical PPIs. This strategy was used by Tarassov *et al.* in 2008 [@pone.0017701-Tarassov1] to compile a forth genome-wide yeast interactome. Unlike the three HT-Y2H studies, the PCA dataset was derived from individual one-by-one matings between haploid yeast strains each carrying bait and prey ORFs. The PCA dataset accounts for 25 of the 72 SSU processome proteins and 27 interactions among them --- the highest coverage among the genome-wide studies ([Fig. 2](#pone-0017701-g002){ref-type="fig"} and [Table 1](#pone-0017701-t001){ref-type="table"}) and shows some overlap of PPIs with the Uetz *et al.* [@pone.0017701-Uetz1] dataset. ::: {#pone-0017701-g002 .fig} 10.1371/journal.pone.0017701.g002 Figure 2 ::: {.caption} ###### Interaction map of the SSU processome proteins from the PCA dataset. Nodes are colored as in [Fig. 1](#pone-0017701-g001){ref-type="fig"}. :::  ::: The literature-curated dataset contains the best SSU processome coverage overall {#s3e} -------------------------------------------------------------------------------- The SSU processome protein coverage of the aforementioned datasets was compared to coverage from literature-curated (LC) sources [@pone.0017701-Boulon1]--[@pone.0017701-Charette1]. These small-scale interaction studies cooperatively account for more SSU processome proteins than any of the individual high-throughput genome-wide datasets [@pone.0017701-Uetz1]--[@pone.0017701-Tarassov1]. In all, the LC dataset accounts for 34 of the 72 proteins and 44 interactions ([Fig. 3](#pone-0017701-g003){ref-type="fig"} and [Table 1](#pone-0017701-t001){ref-type="table"}) and displays some overlap with the HT-Y2H [@pone.0017701-Uetz1]--[@pone.0017701-Yu1] and PCA studies [@pone.0017701-Tarassov1]. ::: {#pone-0017701-g003 .fig} 10.1371/journal.pone.0017701.g003 Figure 3 ::: {.caption} ###### Interaction map of the SSU processome proteins from the LC dataset. Nodes are depicted as in [Fig. 1](#pone-0017701-g001){ref-type="fig"}. :::  ::: Mining for SSU processome interologues {#s3f} -------------------------------------- Conserved protein-protein interactions -- or interologues -- found in multiple organisms, as well as PPIs replicated by multiple studies or distinct experimental methods, carry a higher confidence value and are more likely to represent true interactions [@pone.0017701-Uetz2]--[@pone.0017701-Wiles1]. To determine which interactions have been identified in other organisms, we extracted PPI data for the 72 SSU processome proteins from BioGRID and IntAct for *C. elegans*, *D. melanogaster*, *H. sapiens*, *S. pombe*, *P. falciparum* and *M. musculus*. The Cytoscape map of the interologue dataset disappointingly showed only two interactions between Mpp10 and Imp3, and Mpp10 and Imp4 orthologues in *D. melanogaster* [@pone.0017701-Giot1] and one interaction between Mpp10 and Utp3 orthologues in *C. elegans* ([Fig. 4](#pone-0017701-g004){ref-type="fig"}) [@pone.0017701-Simonis1]. These interactions overlap completely with the yeast dataset, thereby further increasing their likelihood. No interactions within the components of the SSU processome were identified in *S. pombe*, *Plasmodium*, human and mouse PPI datasets. ::: {#pone-0017701-g004 .fig} 10.1371/journal.pone.0017701.g004 Figure 4 ::: {.caption} ###### The current, merged SSU processome interactome map from the three HT-Y2H, PCA, LC and interologue datasets. Interologues identified in *Drosophila* (D) [@pone.0017701-Giot1] and *C. elegans* (C) [@pone.0017701-Simonis1] are also shown, with red and blue edges, respectively. The PPI redundancy (same interactions identified by different studies, methods or reciprocally) was removed from the figure to highlight the interacting partners. Nodes are depicted as in [Fig. 1](#pone-0017701-g001){ref-type="fig"}. Standalone nodes depict proteins without interaction data from any of the compiled datasets. :::  ::: The first partial protein interaction map of the SSU processome {#s3g} --------------------------------------------------------------- Merging all the collected yeast and non-yeast PPI datasets [@pone.0017701-Uetz1]--[@pone.0017701-Giot1], [@pone.0017701-Simonis1], [@pone.0017701-Boulon1]--[@pone.0017701-Charette1] for the 72 SSU processome proteins provides the first partial protein interaction map of the SSU processome. The Cytoscape map of the merged dataset includes 67 distinct edges, corresponding to 67 different interaction pairs among the 72 queried SSU processome proteins ([Fig. 4](#pone-0017701-g004){ref-type="fig"}, [Table 1](#pone-0017701-t001){ref-type="table"} and [S2](#pone.0017701.s002){ref-type="supplementary-material"}). Twenty-six out of the 72 proteins (36.1%) did not have any known interacting partners. The LC data ([Fig. 3](#pone-0017701-g003){ref-type="fig"}) contributed the largest number of interactions of any dataset (47.2% coverage of the 72 queried nodes and 65.7% of the 67 known edges) followed by the PCA data (34.7% of the 72 nodes, 40.3% of the 67 known edges). The other studies each account for less than 10% of the 67 currently known PPIs among the 72 SSU processome proteins ([Table 1](#pone-0017701-t001){ref-type="table"}). A poor overlap for the HT-Y2H, PCA and LC datasets {#s3h} -------------------------------------------------- Interactions identified by different studies or using independent methods carry a higher confidence value [@pone.0017701-Uetz2]--[@pone.0017701-Wiles1]. Therefore, we examined the level of overlap between the genome-wide HT-Y2H studies, the PCA and LC datasets. Minimal congruence was found among the HT-Y2H datasets, with Uetz *et al.* [@pone.0017701-Uetz1] and Ito *et al.* [@pone.0017701-Ito1] not sharing any reported interactions ([Figs. 1](#pone-0017701-g001){ref-type="fig"} and [5](#pone-0017701-g005){ref-type="fig"}). The SSU processome interactions reported by Yu *et al.* [@pone.0017701-Yu1] overlap completely with those of Ito *et al.* [@pone.0017701-Ito1] and were thus already known. The interactions reported in the systems biology study of Hazbun *et al.* [@pone.0017701-Hazbun1] do not overlap with any of the HT-Y2H datasets [@pone.0017701-Uetz1]--[@pone.0017701-Yu1]. Some overlap was found between the HT-Y2H studies [@pone.0017701-Uetz1]--[@pone.0017701-Yu1] and the PCA dataset [@pone.0017701-Tarassov1] (nine proteins and four PPIs; [Figs. 1](#pone-0017701-g001){ref-type="fig"}, [2](#pone-0017701-g002){ref-type="fig"} and [5](#pone-0017701-g005){ref-type="fig"}). Overlap was also found between the HT-Y2H studies [@pone.0017701-Uetz1]--[@pone.0017701-Yu1], the PCA dataset [@pone.0017701-Tarassov1] and the LC dataset ([Figs. 1](#pone-0017701-g001){ref-type="fig"}, [2](#pone-0017701-g002){ref-type="fig"}, [3](#pone-0017701-g003){ref-type="fig"} and [5](#pone-0017701-g005){ref-type="fig"}) [@pone.0017701-Boulon1]--[@pone.0017701-Charette1]. However, 18 of the 34 proteins in the LC dataset did not overlap with any of the HT-Y2H [@pone.0017701-Uetz1]--[@pone.0017701-Yu1] or PCA [@pone.0017701-Tarassov1] studies. ::: {#pone-0017701-g005 .fig} 10.1371/journal.pone.0017701.g005 Figure 5 ::: {.caption} ###### Comparison of the overlap between the HT-Y2H, PCA and LC datasets for the PPIs of the SSU processome. Numbers within the Venn diagram refer to the number of SSU processome proteins present and overlapping in the HT-Y2H, PCA and LC datasets. :::  ::: Discussion {#s4} ========== Large-scale, genome-wide yeast binary protein interaction networks contain thousands of PPIs suggesting comprehensive and complete investigations of the yeast interactome. We mined the existing databases, containing PPIs from all HT-Y2H [@pone.0017701-Uetz1]--[@pone.0017701-Yu1], [@pone.0017701-Hazbun1], PCA [@pone.0017701-Tarassov1] and LC [@pone.0017701-Boulon1]--[@pone.0017701-Charette1] yeast interactome studies to date for interactions among the 72 SSU processome proteins. Individual datasets were analyzed for the extent of PPI coverage and overlap and were merged to generate one comprehensive interaction dataset. Individual datasets and their amalgamation were each drawn into interactome maps using Cytoscape. Our results show that filtering the current HT-Y2H [@pone.0017701-Uetz1]--[@pone.0017701-Yu1], [@pone.0017701-Hazbun1], PCA [@pone.0017701-Tarassov1] and LC [@pone.0017701-Boulon1]--[@pone.0017701-Charette1] datasets for SSU processome PPIs provided sparse data, with as many as 36.1% (26 of 72 SSU processome proteins) of the protein components having no currently known interaction partner. A strategy similar to ours has successfully been used to draw an interaction map of promyelocytic leukaemia protein nuclear bodies (PML-NBs) [@pone.0017701-VanDamme1]. How many protein-protein interactions are expected? {#s4a} --------------------------------------------------- There are approximately 6,000 proteins and a conservative estimate of 18,000+/−4500 PPIs in the entire yeast interactome [@pone.0017701-Yu1], [@pone.0017701-Grigoriev1]--[@pone.0017701-Bader1], equaling an average of 3 to 3.5 interactions per protein (though this number may be as high as five interactions per protein [@pone.0017701-Blow1]). By this calculation, for 72 SSU processome proteins, we expected roughly 216 to 252 PPIs in total ([Table 1](#pone-0017701-t001){ref-type="table"}). Based on the lower end of the theoretical number of expected PPIs, the 67 PPIs that we obtained from the merged datasets represent at most 31.0% of the predicted interactions in the SSU processome ([Table 1](#pone-0017701-t001){ref-type="table"}). This number is in line with similar estimates from merged HT-Y2H datasets suggesting ∼20% coverage of the entire yeast interactome [@pone.0017701-Yu1]. From these values, it is clear that we do not yet have an interactome of the SSU processome that is nearly complete. Comparing the HT-Y2H, PCA and LC datasets {#s4b} ----------------------------------------- Among the genome-wide studies, the PCA dataset of Tarassov *et al.* [@pone.0017701-Tarassov1] reports the highest PPI coverage when compared to the three HT-Y2H-based approaches [@pone.0017701-Uetz1]--[@pone.0017701-Yu1], accounting for 25 SSU processome proteins and 12.5 percent of the predicted edges ([Table 1](#pone-0017701-t001){ref-type="table"}). This might be attributed to the distinctiveness of the PCA method [@pone.0017701-Jensen2] and to the screening strategy, which involved a one-by-one matrix array where each bait-containing strain was individually mated to each prey-containing strain [@pone.0017701-Tarassov1]. In contrast, the prey pooling approach used in the Uetz *et al.* [@pone.0017701-Uetz1], Ito *et al.* [@pone.0017701-Ito1] and Yu *et al.* [@pone.0017701-Yu1] HT-Y2H studies has potentially lower quality data and coverage, possibly because: *i*) some prey plasmids may replicate faster due to their smaller size, and can overtake the population in the pool by outcompeting larger prey plasmids that take longer or are more difficult to replicate, *ii*) some proteins, when over-expressed, may be toxic to the cell resulting in a dominant negative phenotype, while other proteins can enhance cell growth (cells with improved growth can outcompete other cells, while those with a dominant negative phenotype will be eliminated from the pool) and *iii*) there may be transformation and mating differences among different prey fusion protein plasmids [@pone.0017701-Koegl1], [@pone.0017701-Rajagopala2]. Furthermore, array-based screened may be more sensitive and more easily screened to saturation [@pone.0017701-Koegl1], [@pone.0017701-Rajagopala2]. Thus, the individualized mating process used by Tarassov *et al.* [@pone.0017701-Tarassov1], which avoids many of the potential problems associated with the pooling approach, could explain their higher coverage of the SSU processome protein interactome. Protein interactions reported by more than one study, replicated via distinct methods or reported in different organisms are more likely to be authentic [@pone.0017701-Uetz2]--[@pone.0017701-Wiles1]. As has been found in other studies [@pone.0017701-Tarassov1], [@pone.0017701-Mrowka1], [@pone.0017701-Jensen2]--[@pone.0017701-vonMering1], inspection and comparison among the compiled HT-Y2H, PCA and LC datasets, however, revealed poor overlap, especially among the genome-wide HT-Y2H datasets [@pone.0017701-Uetz1]--[@pone.0017701-Yu1] which contained very few overlapping PPIs. Due to the large contributions of the LC [@pone.0017701-Boulon1]--[@pone.0017701-Charette1] and PCA [@pone.0017701-Tarassov1] datasets to the interaction map of the SSU processome, most of the overlaps occurred between the LC and PCA datasets ([Figs. 2](#pone-0017701-g002){ref-type="fig"}, [3](#pone-0017701-g003){ref-type="fig"} and [5](#pone-0017701-g005){ref-type="fig"}). The poor overlap among the comprehensive HT-Y2H interactomes brings into question their proposed completeness and suggests that these screens were not exhaustive nor done to saturation. The high quality of the LC dataset {#s4c} ---------------------------------- Smaller-scale LC datasets provided the highest coverage of the SSU processome proteins, reporting 34 proteins and 44 interactions (47.2% and 20.4% of the predicted totals, respectively). While conventional wisdom supports LC datasets to be accurate and of high-quality, some have remained skeptical, pointing to the poor overlap among the literature-curated studies, as well as protein name and species classification errors [@pone.0017701-Rual1], [@pone.0017701-Venkatesan1], [@pone.0017701-Cusick1]--[@pone.0017701-Dreze1]. Surveys to assess the reliability of literature-curated data by re-curation revealed roughly half of LC derived data to lack validation via alternative, independent methods [@pone.0017701-Rual1], [@pone.0017701-Venkatesan1], [@pone.0017701-Cusick1]. In contrast to these claims, our analysis revealed the LC data to be the most comprehensive. Furthermore, many of the SSU processome PPIs from the mined LC dataset were found to be validated by independent methods such as *E. coli* pull-downs and biochemical and biophysical assays ([Table 2](#pone-0017701-t002){ref-type="table"}). ::: {#pone-0017701-t002 .table-wrap} 10.1371/journal.pone.0017701.t002 Table 2 ::: {.caption} ###### Y2H-derived PPI data confirmed by alternative and supplementary experimental methods. ::: {#pone-0017701-t002-2} Interactor A Interactor B Y2H GST/His-Tag pull-down Biochemical activation Motif mapping Surface plasmon resonance -------------- -------------- ------------------------------------------------------------- ---------------------------------- ---------------------------------- ------------------------------------------------------------- --------------------------------- Imp3 Mpp10 **✓** [@pone.0017701-Lee1] **✓** [@pone.0017701-Lee1] **✓** [@pone.0017701-Lee1] Imp4 Mpp10 **✓** [@pone.0017701-Lee1] **✓** [@pone.0017701-Lee1] **✓** [@pone.0017701-Lee1] Esf2 Dbp8 **✓** [@pone.0017701-Granneman1] **✓** [@pone.0017701-Granneman1] **✓** [@pone.0017701-Granneman1] Pfa1 Prp43 **✓** [@pone.0017701-Pandit1], [@pone.0017701-Lebaron2] **✓** [@pone.0017701-Lebaron2] **✓** [@pone.0017701-Lebaron2] **✓** [@pone.0017701-Lebaron2] t-Utp8 t-Utp9 **✓** [@pone.0017701-Freed1], [@pone.0017701-Huang1] **✓** [@pone.0017701-Huang1] **✓** [@pone.0017701-Huang1] **✓** [@pone.0017701-Huang1] Utp6 Utp21 **✓** [@pone.0017701-Champion1] **✓** [@pone.0017701-Champion1] **✓** [@pone.0017701-Champion1] Utp6 Utp18 **✓** [@pone.0017701-Champion1] **✓** [@pone.0017701-Champion1] Utp25 Utp3 **✓** [@pone.0017701-Goldfeder1], [@pone.0017701-Charette1] **✓** [@pone.0017701-Goldfeder1] **✓** [@pone.0017701-Goldfeder1], [@pone.0017701-Charette1] Many PPIs from the LC data have alternative forms of supporting evidence from experiments that test for binary interactions, including pull-downs, activation of enzymatic activities, motif mapping by truncations and surface plasmon resonance. This list of protein-protein interactions identified by Y2H and validated by independent methods is not exhaustive. ::: Sparse interologue data for SSU processome components {#s4d} ----------------------------------------------------- The use of interologues in protein-protein interaction maps is rapidly increasing and constitutes a valid strategy for augmenting interactome coverage [@pone.0017701-Wiles1]. Some of the PPIs identified by multiple studies, such as between Imp3 and Mpp10, and Imp4 and Mpp10, were also reported in different organisms such as *Drosophila* [@pone.0017701-Giot1]. Although all 72 SSU processome components were queried in six additional organisms other than *S. cerevisiae*, the majority of retrieved PPIs were with non-SSU processome proteins or with proteins with no known yeast orthologues. Once the SSU processome components of various model organisms are better characterized, and their yeast orthologues determined, additional conserved interactions may be identified. However, our analysis suggests that the interactome coverage of *C. elegans*, *D. melanogaster*, *S. pombe*, *P. falciparum*, human and mouse may be even less than that of yeast. This is in line with a recent report suggesting that low interactome coverage, and not evolutionary divergence and loss of interologues, as the main obstacle to interactome network alignment [@pone.0017701-Ali1]. What does this tell us about the SSU processome protein-protein interaction map? {#s4e} -------------------------------------------------------------------------------- A few novel interactions previously undetected by HT-Y2H and LC studies surfaced in the PCA dataset: between t-Utp4 and t-Utp10, t-Utp5 and t-Utp8, t-Utp5 and t-Utp9, and t-Utp8 and t-Utp15 of the UtpA/t-Utp subcomplex and between Utp1 and Utp12 of the UtpB subcomplex (compare [Figs. 2](#pone-0017701-g002){ref-type="fig"} and [3](#pone-0017701-g003){ref-type="fig"}). The identification of these interactions in the PCA dataset [@pone.0017701-Tarassov1] but not in the HT-Y2H or LC datasets [@pone.0017701-Champion1]--[@pone.0017701-Freed1] may be due to differences between the Y2H and PCA methodologies [@pone.0017701-Jensen2] or to differences resulting from the use of different fusion tags in Y2H and PCA screening strategies. Indeed, the N- versus C-terminal placement of fusion tags in Y2H assays has been shown to influence the outcome of screens [@pone.0017701-Stellberger1]. Regardless, validating these PCA derived interactions will further clarify the assembly of the t-Utp/UtpA and UtpB subcomplexes of the SSU processome. Novel interactions were also reported between t-Utp4 of the UtpA/t-Utp and Utp18 of the UtpB subcomplexes. This interaction may suggest one of the first PPIs linking the various subcomplexes of the SSU processome, and is also a candidate for future validation studies. Interestingly, all genome-wide HT-Y2H screens [@pone.0017701-Uetz1]--[@pone.0017701-Yu1] are missing these interactions, potentially due to these findings being either an artifact of the PCA approach, or a false negative of the Y2H methodology. False negatives in Y2H screens may arise from bait and prey proteins that normally interact via their N-terminus, since the DNA binding or activation domains, which are typically attached to the N-terminus of the proteins, may mask these interaction surfaces. A truly comprehensive interactome map of the SSU processome will provide us with insight into the complexities of the assembly, function and regulation of this large ribonucleoprotein complex. Since the SSU processome is required for the production of ribosomes in all eukaryotes, understanding its assembly is essential to elucidating its function in ribosome biogenesis. Our analyses of the existing databases indicates that ∼70% of the PPIs in the SSU processome have yet to be determined, and because of this we do not yet have an accurate picture of how this complex is assembled. The current lack of data includes both proteins with no known interactors, and missing PPIs between other connected proteins. Enhancing the experimental approaches to both the classic methods --- such as the Y2H --- and new methods --- such as the PCA --- are likely to be crucial for not only deriving an interactome map of the SSU processome, but a comprehensive and exhaustively screened yeast PPI map that covers the entire yeast proteome. This quantitative survey of existing databases for PPIs from HT-Y2H [@pone.0017701-Uetz1]--[@pone.0017701-Yu1], PCA [@pone.0017701-Tarassov1] and LC [@pone.0017701-Boulon1]--[@pone.0017701-Charette1] studies reveals a remarkably sparse coverage of the SSU processome proteins, albeit having drawn data from interactomes purporting to be highly comprehensive. Nevertheless, the absence of a truly comprehensive, genome-wide interactome is apparent. The LC dataset, which provided the highest coverage of the SSU processome proteins, contained PPIs that were confirmed by alternative methods, such as *E. coli* pull-downs and biochemical and biophysical methods that also test for direct binary interactions. This confirms that PPIs from LC sources, despite previously proposed skepticism, are largely credible. Although lacking many proteins and interactions, the up-to-date SSU processome interaction map compiled in this study can be applied to generate new hypotheses of subcomplex interactions, assembly and function. Additionally, approaches to experimentally determine the domain-domain interactions of the known PPIs [@pone.0017701-Pang1] can be applied to better understand the biology of the SSU processome. Supporting Information {#s5} ====================== Table S1 ::: {.caption} ###### **The protein components of the SSU processome.** The catalogued proteins are listed based on their membership in the known subcomplexes of the yeast SSU processome. Confirmed SSU processome components which have not been assigned to a specific subcomplex are listed as unclassified. Candidate SSU processome proteins are listed as unknown. The yeast SSU r-proteins (Rps4, Rps6, Rps7, Rps9 and Rps14) that are known components of the SSU processome [@pone.0017701-Bernstein1] are not listed. (?) denotes uncertain membership in an SSU processome sub-complex. Motif and domain abbreviations include: glycine/arginine-rich (GAR); coiled-coil (CC); middle domain of eIF4G (MIF4G); MA3 domain (similar to MIF4G domains/MI domain); helicase conserved C-terminal domain (HELICc); helicase associated domain (HA2); glycine-rich nucleic binding domain (G-patch); RxxxH ssRNA binding motif (R3H); Pumilio homology RNA binding domain (PUM/PUF); RNA recognition motif (RRM, RBD or RNP domain); low-temperature viability protein domain (LTV1); fungal-specific family of rRNA processing proteins (rRNA processing domain); small domain in a novel nucleolar family (NUC153); beta-transducin repeats (WD40); S1 RNA-binding motifs; Half-A-TPR (HAT) repeats; K homology RNA-binding domain (KH); Down-Regulated In Metastasis (DRIM); Armadillo (ARM) protein-protein interaction repeat; CBF/Mak21 family; nucleolar complex (NOC) associated protein domain. Table modified from Phipps *et al.* [@pone.0017701-Phipps1]. (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S2 ::: {.caption} ###### **The SSU processome PPIs derived from the HT-Y2H, PCA and LC datasets.** (XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: The authors wish to thank all members of the Baserga laboratory for their support and insightful discussions. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was supported by a National Institutes of Health ([www.nih.gov](http://www.nih.gov)) grant GM052581 (to SJB), the Science, Technology and Research Scholars II (STARS II) program (<http://yalecollege.yale.edu/content/stars-ii-program>) at Yale University (to YHL) and a post-doctoral NIH-Ruth L. Kirschstein National Research Service Award, Institutional Research Training Grant: Radiation Therapy, Biology, Physics (<http://grants.nih.gov/training/nrsa.htm>) T32 CA009259 (to JMC). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: JMC SJB. Performed the experiments: YHL JMC. Analyzed the data: YHL JMC SJB. Contributed reagents/materials/analysis tools: YHL JMC SJB. Wrote the paper: YHL JMC SJB.

PubMed Central

2024-06-05T04:04:19.843590

2011-3-10

PMC3053387

Introduction {#s1} ============ Previous reports indicated that bacterial genomic GC content ranges from 25 to 75% [@pone.0017677-Barbu1], [@pone.0017677-Belozersky1], [@pone.0017677-Sueoka1]. Based on this observation, Sueoka [@pone.0017677-Sueoka1],[@pone.0017677-Sueoka2] predicted that differences in GC content would affect protein amino acid sequence even though the genetic code had not been elucidated at the time. Once DNA sequence technology was developed, Muto and Osawa [@pone.0017677-Muto1] and later Knight et al. [@pone.0017677-Knight1] analyzed the available nucleotide sequences from the genes of the available species of bacteria and showed that the GC content at all three codon positions increased with increasing genomic GC content. Similar results were obtained when 23 complete cyanobacterial genomes were analyzed [@pone.0017677-Fryxell1]. In all three datasets, the biggest change in GC content was observed in the third codon position where most nucleotide changes do not change the amino acid sequence of the encoded proteins due to the redundancy of the genetic code. However, significant increases in GC content were observed in the first and second codon positions as well. Since nucleotide changes in the first and second codon positions usually result in changes in the amino acid sequence, these results suggested that genomic GC content has a significant impact protein amino acid sequence. A more in depth analysis [@pone.0017677-Lobry1] showed that the use of amino acids encoded by GC-rich codons increased with increasing genomic GC content while the use of amino acids coded by GC-poor codons decreased with increasing GC content. Subsequently, Gu et al. [@pone.0017677-Gu1] examined 15 individual proteins in 15 bacteria with a range of genomic GC content and showed that the amino acid composition of each protein was affected by changes in genomic GC content. A year later, Wilquet and Van de Casteele [@pone.0017677-Wilquet1] performed an analysis of 22 protein families from 30 species of bacteria and archaea and concluded that the first codon position dominated the relationship between GC content and amino acid composition. Similar results were obtained when complete bacterial genomes were analyzed [@pone.0017677-Singer1], [@pone.0017677-Chen1]. Thus, bacterial codon usage patterns appear to reflect genomic GC content rather than phylogenetic relationships. This conclusion was reinforced by a subsequent study of 100 bacterial and archaeal genomes which showed that codon usage could be predicted by an analysis of the non-coding regions of the genomes [@pone.0017677-Chen2]. Since an inspection of the completely sequenced bacterial genomes available in GenBank June 2010) suggested that the diversification of bacterial genomic GC content occurred independently in most bacterial phyla and classes, we decided to investigate the impact of genomic GC content on codon usage in more detail than what is described in the studies discussed above. An inspection of the bacterial genome sequences available within GenBank revealed four classes of the phylum *Proteobacteria* and four additional phyla that contained completed genome sequences from at least 15 different species. Therefore, we chose 8--12 species that represented the range of genomic GC content of each class or phylum and performed separate analyses on each group. We hypothesized that if genomic GC content was the primary determinant of codon and amino acid usage patterns, then similar usage patterns would have arisen independently in each of the eight classes or phyla. However, if codon usage patterns were the principal determinant of genomic GC content, then different codon usage patterns might have arisen in different bacterial phyla. Since similar patterns of codon usage were observed in all eight comparisons, we concluded that shared codon usage patterns have arisen independently in multiple bacterial phyla. In addition, we were able to quantify the impact of genomic GC content on predicted protein amino acid content. Results {#s2} ======= Diversification of genomic GC content {#s2a} ------------------------------------- An inspection of the GC content of the completely sequenced bacterial genomes currently available in GenBank (June 2, 2010) revealed that many bacterial phyla, and also the individual classes of *Proteobacteria*, contain broad ranges of genomic GC content. Since bacterial phyla and classes are considered to be well established taxonomic groups, the diversification of genomic GC content must have occurred after the evolution of bacterial phlya and classes. However, since the bacterial species in the same genus usually have a similar genomic GC content, genomic GC content diversification must have occurred prior to, or as part of, the process of the evolution of bacterial genera. When individual phyla and classes of bacteria were examined, eight were represented by at least 15 species ([Tables 1](#pone-0017677-t001){ref-type="table"} and [2](#pone-0017677-t002){ref-type="table"}). Within each group, a broad range of genomic GC content was observed ([Fig. 1](#pone-0017677-g001){ref-type="fig"}), indicating that a wide range genomic GC content has evolved independently in each of these eight taxonomic groups. The distribution of genomic GC content appears to differ among these phyla and classes. However, the differing patterns of GC content presented in the bar graphs may not reflect the true distribution within each group since the available sequenced genomes do not reflect a random sampling of the members of the group. Nevertheless, we were able to take advantage of the range of genomic GC content to choose a collection of 8 to 12 individual species that represent the range of diversity of genomic GC content observed within each group. In contrast to the widely quoted range of 25 to 75% for bacterial genomic GC content, the current list of completed bacterial genome sequences includes 12 bacterial species in three different phyla with GC contents that range from 16.6% to 24% ([Fig. 1](#pone-0017677-g001){ref-type="fig"}). Thus, the range of genomic GC content (16.6% to 74.9%) is broader than previously realized, with GC base pairs comprising only one sixth of the genome at the lower extreme. Clearly, a reduction in the genomic GC content of this magnitude would require a severe reduction in the use of amino acids such as proline, alanine, and glycine that are encoded by GC-rich codons. ::: {#pone-0017677-g001 .fig} 10.1371/journal.pone.0017677.g001 Figure 1 ::: {.caption} ###### Distribution of genomic GC content within bacterial phyla or classes. Within each class or phylum, the genomic GC contents from all individual genomes available on June 16, 2010 were binned in five percent increments with the number on the X-axis representing the top range of the bin. A) *Actinobacteria*; B) *Alphaproteobacteria*; C) *Betaproteobacteria*; D) *Bacterioides/Chlorbi*; E) *Cyanobacteria*; F) *Deltaproteobacteria*; G) *Firmicutes*; H) *Gammaproteobacteria*. :::  ::: ::: {#pone-0017677-t001 .table-wrap} 10.1371/journal.pone.0017677.t001 Table 1 ::: {.caption} ###### Bacterial species included within each phylum analyzed in this study. ::: {#pone-0017677-t001-1} Phylum Species GC% -------------------------- -------------------------------------------------------- ------ *Actinobacteria* *Gardnerella vaginalis* 409-05 42 *Atopobium parvulum* DSM 20469 45.7 *Cryptobacterium curtum* DSM 15641 50.9 *Corynebacterium diphtheriae* NCTC 13129 53.5 *Renibacterium salmoninarum* ATCC 33209 56.3 *Propionibacterium acnes* KPA171202 60 *Corynebacterium urealyticum* DSM 7109 64.2 *Arthrobacter chlorophenolicus* A6 66 *Frankia sp.* EAN1pec 71.2 *Kineococcus radiotolerans* SRS30216 74.2 *Bacteroidetes/Chlorobi* *Candidatus Sulcia* muelleri GWSS 22.4 *Blattabacterium sp.* (*Blattella germanica*) str. Bge 27.1 *Flavobacterium psychrophilum* JIP02/86 32.5 *Candidatus Amoebophilus asiaticus* 5a2 35 *Pedobacter heparinus* DSM 2366 42 *Chloroherpeton thalassium* ATCC 35110 45 *Prosthecochloris aestuarii* DSM 271 50.1 *Robiginitalea biformata* HTCC2501 55.3 *Rhodothermus marinus* DSM 4252 64.3 *Salinibacter ruber* DSM 13855 66.1 *Cyanobacteria* *Prochlorococcus marinus* str. MIT 9515 30.8 *Prochlorococcus marinus* str. NATL1A 35 *Nostoc sp.* PCC 7120 41.3 *Acaryochloris marina* MBIC11017 47 *Prochlorococcus marinus* str. MIT 9303 50 *Synechococcus elongatus* PCC 7942 55.4 *Synechococcus sp.* JA-3-3Ab 60.2 *Gloeobacter violaceus* PCC 7421 62 *Firmicutes* *Mycoplasma capricolum subsp. capricolum* ATCC 27343 23.8 *Mycoplasma mobile* 163K 25 *Mycoplasma arthritidis* 158L3-1 30.7 *Bacillus cereus* G9842 35 *Mycoplasma pneumoniae* M129 40 *Lactobacillus brevis* ATCC 367 46.1 *Paenibacillus sp.* JDR-2 50.3 *Acidaminococcus fermentans* DSM 20731 55.8 *Candidatus Desulforudis audaxviator* MP104C 60.8 *Symbiobacterium thermophilum* IAM 14863 68.7 ::: ::: {#pone-0017677-t002 .table-wrap} 10.1371/journal.pone.0017677.t002 Table 2 ::: {.caption} ###### Bacterial species included within the classes analyzed in this study. ::: {#pone-0017677-t002-2} Class Species GC% ----------------------- ----------------------------------------------------------------- ------ *Alphaproteobacteria* *Ehrlichia ruminantium* str. Gardel 27.5 *Ehrlichia chaffeensis* str. Arkansas 30.1 *Wolbachia* endosymbiont of *Drosophila melanogaster* 35.2 *Neorickettsia sennetsu* str. Miyayama 41.1 *Hirschia baltica* ATCC 49814 45.2 *Anaplasma centrale* str. Israel 50 *Ochrobactrum anthropi* ATCC 49188 56.1 *Gluconobacter oxydans* 621H 60.8 *Rhodopseudomonas palustris* BisB18 65 *Betaproteobacteria* *Polynucleobacter necessarius* subsp. asymbioticus QLW-P1DMWA-1 44.8 *Methylotenera mobilis* JLW8 45.5 *Nitrosomonas eutropha* C71 48.5 *Nitrosomonas europaea* ATCC 19718 50.7 *Nitrosospira multiformis* ATCC 25196 53.9 *Methylobacillus flagellatus* KT 55.7 *Dechloromonas aromatica* RCB 59.2 *Comamonas testosteroni* CNB-2 61.4 *Ralstonia pickettii* 12D 63.3 *Verminephrobacter eiseniae* EF01-2 65.2 *Leptothrix cholodnii* SP-6 68.9 *Deltaproteobacteria* *Lawsonia intracellularis* PHE/MN1-00 33.1 *Desulfotalea psychrophila* LSv54 46.6 *Bdellovibrio bacteriovorus* HD100 50.6 *Pelobacter carbinolicus* DSM 2380 55.1 *Geobacter bemidjiensis* Bem 60.3 *Desulfovibrio vulgaris* str. 'Miyazaki F' 67.1 *Sorangium cellulosum* 'So ce 56' 71.4 *Anaeromyxobacter dehalogenans* 2CP-C 74.9 *Gammaproteobacteria* *Candidatus Carsonella ruddii* PV 16.6 *Buchnera aphidicola str.* Cc *(Cinara cedri)* 20.2 *Buchnera aphidicola str.* Bp *(Baizongia pistaciae)* 25.3 *Candidatus Vesicomyosocius okutanii* HA 31.6 *Haemophilus somnus (histophilus somni)* 129PT 37.2 *Haemophilus parasuis* SH0165 40 *Shewanella denitrificans* OS217 45.1 *Escherichia coli* APEC O1 50.3 *Dickeya dadantii* Ech703 55 *Pseudomonas fluorescens* SBW25 60.1 *Xanthomonas campestris pv. campestris str.* 8004 65 *Halorhodospira halophila* SL1 68 ::: Codon and amino acid usage {#s2b} -------------------------- To determine how the observed variation in genomic GC content impacted codon usage, the GC content of each codon position in the coding region of the selected genomes was plotted against the genomic GC content using the *Alphaproteobacteria* as an example. Consistent with previous observations of bacterial genomes in general [@pone.0017677-Muto1], [@pone.0017677-Fryxell1], the GC content of the third codon position of the *Alphaproteobacteria* genomes increased rapidly with increasing genomic GC content ([Fig. 2](#pone-0017677-g002){ref-type="fig"}). The GC content of the first and second codon positions increased as well, but to a lesser extent. Thus, changes in all three codon positions contribute to the observed differences in genomic GC content in the *Alphaproteobacteria*. Similar patterns of changes in all three codon positions were observed among the representatives of each of the other seven phyla and classes included in this study (data not shown). Although the increases in third codon position GC content could be accomplished without affecting the protein amino acid sequences due to the redundancy of the genetic code, the observed increases in the GC content of the first or second codon positions would require changes in the distribution of amino acids incorporated into cellular proteins. ::: {#pone-0017677-g002 .fig} 10.1371/journal.pone.0017677.g002 Figure 2 ::: {.caption} ###### The GC content at each of the three codon positions was plotted against the genomic GC content of the representative *Alphaproteobacteria* listed in [**Table 2**](#pone-0017677-t002){ref-type="table"}. :::  ::: To examine how these changes in genomic GC content impact the distribution of the amino acids used to make proteins, the proportion of amino acids that are encoded by degenerate codons with either the three high (G or C at each of the first two positions) ([Fig. 3](#pone-0017677-g003){ref-type="fig"}) or three of the five low (A or T at each of the first two positions) GC-content codon families ([Fig. 4](#pone-0017677-g004){ref-type="fig"}) was plotted against the genomic GC content of the representative bacteria in the phyla and classes examined. Clear trends were observed. Use of amino acids encoded by the high GC codons increased with increasing genomic GC content while use of amino acids encoded by the low GC codons decreased. For example, the frequency of codons coding for alanine in the *Alphaproteobacteria* increased more than two-fold with increasing genomic GC content ([Fig. 3A](#pone-0017677-g003){ref-type="fig"}) while the frequency of codons coding for isoleucine dropped more than two-fold ([Fig. 3A](#pone-0017677-g003){ref-type="fig"}). To obtain a quantitative measure of the change in amino acid codon use, we calculated the slope of the best fit line for each of the amino acids shown in [Fig. 3](#pone-0017677-g003){ref-type="fig"} and [4](#pone-0017677-g004){ref-type="fig"}. The slope for alanine was 0.0021 in the *Alphaproteobacteria*. Thus, if one *Alphaproteobacteria* genome had a 10% higher GC content than another, the percentage of the total codons that coded for alanine would increase approximately 2.1% (for example, from 6% to 8.1% of the total). Conversely, the best fit line for isoleucine codons had a slope of −0.0014, and a 10% increase in GC content would result in a 1.4% decrease in the number of isoleucine codons. Results similar to those obtained with the *Alphaproteobacteria* were obtained with each of the other bacterial phyla and classes ([Fig. 3](#pone-0017677-g003){ref-type="fig"}). In fact, for the Gammaproteobacteria, only 1.5% of the total codons were alanine codons in the *Candidatus Carsonella ruddii* PV genome (16.6% GC), while alanine codons were 11.9% of the total in the *Halorhodospira halophila* SL1 genome (68% GC). ::: {#pone-0017677-g003 .fig} 10.1371/journal.pone.0017677.g003 Figure 3 ::: {.caption} ###### The predicted percentage of amino acids encoded by the three high-GC codon families, proline, alanine, and glycine, plotted against the genomic GC content of the representative bacteria among the following groups. A\) *Alphaproteobacteria*; B) *Betaproteobacteria*; C) *Gammaproteobacteria*; D) *Deltaproteobacteria*; E) *Actinobacteria*; F) *Bacteriodetes/Chlorobi*; G) *Cyanobacteria*; H) *Firmicutes*. :::  ::: ::: {#pone-0017677-g004 .fig} 10.1371/journal.pone.0017677.g004 Figure 4 ::: {.caption} ###### The predicted percentage of amino acids encoded three low-GC codon families, asparagine, lysine, and isoleucine, plotted against the genomic GC content of the representative bacteria among the following groups. A\) *Alphaproteobacteria*; B) *Betaproteobacteria*; C) *Gammaproteobacteria*; D) *Deltaproteobacteria*; E) *Actinobacteria*; F) *Bacteriodetes/Chlorobi*; G) *Cyanobacteria*; H) *Firmicutes*. :::  ::: To summarize the impact of GC content on amino acid codon use, we plotted the frequency of each amino acid codon family in the coding region of the genomes of the representatives of each bacterial phylum or class analyzed in this study versus the genomic GC content for each species. The six-fold degenerate codon families were sub-divided into two codon families based on the first two codon positions since the use of these sub-divided families differed with genomic GC content (see below). We then calculated the slopes of the best fit lines for each amino acid plot ([Tables S1](#pone.0017677.s001){ref-type="supplementary-material"}--[S3](#pone.0017677.s003){ref-type="supplementary-material"}). These slopes were then averaged for each amino acid codon family across all eight phyla or classes of bacteria investigated in this study and plotted versus the average GC content of each codon family ([Fig. 5](#pone-0017677-g005){ref-type="fig"}). The results showed that the high GC amino acid codon families (upper right corner) that code for glycine, proline, and alanine had average slopes that ranged from 0.0008 to 0.0019 indicating an increase of 0.8 to 1.9% for each 10% increase in GC content for all eight phyla or classes of bacteria examined. In contrast, the low GC amino acid codon families (lower left corner) that code for asparagine, lysine, and isoleucine had average slopes that ranged from −0.0011 to −0.0017 indicating a decrease of 1.1 to 1.7% for each 10% increase in GC content. In both of these data sets, standard deviations were less than 20% of the average value. The low GC amino acid codon families that code for phenylalanine and tyrosine showed a similar, but smaller effect with average slopes of −0.0005 and −0.0004, respectively. However, these two amino acids generally occur in proteins at lower frequencies. As expected, the slopes for most codon families with 50% average GC content ranged from −0.0002 and 0.0003 indicating little or no change in the use of these amino acids with increasing genomic GC content. In contrast, the slope for the leucine-C subgroup (leucine codons with a C in the first position) was 0.0017 even though the average GC content of this set of leucine codons is also 0.5. This anomaly is easily explained when the slope of the alternative subgroup of leucine codons (leucine T) is considered (−0.0015). Even though the average GC content of the leucine-C codons is moderate, the average GC content of the alternative leucine-T codons is much lower (0.1667). Thus, genomes with a low GC content preferentially use codons that begin with T, and genomes with a high GC content preferentially use leucine codons that start with C ([Fig. 6](#pone-0017677-g006){ref-type="fig"}). A similar result was obtained when the use of the six arginine codons was examined. The arginine-A codon family exhibited an average slope of −0.00045, while the slope of the arginine-C codon family was 0.0018 ([Fig. 7](#pone-0017677-g007){ref-type="fig"}). In both of these six-codon families, the synonymous first position substitutions behave like synonymous third position substitutions. Thus, there is a consistent trend in all eight phyla and classes that GC-rich codons are used preferentially in all high GC genomes, resulting in both synonymous and nonsynonymous changes relative to genomes with lower GC content. Similarly, AT-rich codons are used preferentially in low GC genomes. ::: {#pone-0017677-g005 .fig} 10.1371/journal.pone.0017677.g005 Figure 5 ::: {.caption} ###### Average of the slopes for all 20 amino acid codon families (with the six-fold degenerate amino acid codon families being divided into two groups based on the first position nucleotide) for all eight groups plotted against the average GC content of their respective codons. For each class or phylum, the frequency of each amino acid in the coding portion of each genome was plotted against the genomic GC content of the genome. The slopes of these eight plots were then averaged and plotted versus the average GC content of the codon family for each amino acid. :::  ::: ::: {#pone-0017677-g006 .fig} 10.1371/journal.pone.0017677.g006 Figure 6 ::: {.caption} ###### Arginine codon use plotted against the genomic GC content of the representative bacteria among the following groups. A\) *Alphaproteobacteria*; B) *Betaproteobacteria*; C) *Gammaproteobacteria*; D) *Deltaproteobacteria*; E) *Actinobacteria*; F) *Bacteriodetes/Chlorobi*; G) *Cyanobacteria*; H) *Firmicutes*. Arg-A and Arg-C refer to arginine codon families with either an A or a C in the first codon position, respectively. :::  ::: ::: {#pone-0017677-g007 .fig} 10.1371/journal.pone.0017677.g007 Figure 7 ::: {.caption} ###### Leucine codon use plotted against the genomic GC content of the representative bacteria among the following groups. A\) *Alphaproteobacteria*; B) *Betaproteobacteria*; C) *Gammaproteobacteria*; D) *Deltaproteobacteria*; E) *Actinobacteria*; F) *Bacteriodetes/Chlorobi*; G) *Cyanobacteria*; H) *Firmicutes*. Leu-T and Leu-C refer to leucine codon families with either a T or a C in the first codon position, respectively. :::  ::: Discussion {#s3} ========== It is well known that the genomic GC content of bacteria varies widely and that divergent genomic GC content is correlated with altered codon and amino acid usage. In contrast to previous studies, our analysis of individual bacterial classes and phyla was designed to determine whether similar patterns of codon and amino acid usage variation with respect to genomic GC content have evolved independently in each of the eight groups. The divergence of genomic GC content within disparate Gram negative and Gram positive bacterial phyla indicate that GC content divergence was not produced by a single event that resulted in individual phyla or classes filling a certain GC content niche. Instead, each of the eight phyla and classes in this study contained species with widely disparate genomic GC concentrations ([Fig. 1](#pone-0017677-g001){ref-type="fig"}). Since the current classification of bacteria into phyla and classes is generally consistent with the available nucleotide sequence information, the observed distribution of genomic GC content must have evolved independently in each of the phyla and classes analyzed in this study. In addition, since closely related bacterial species usually have very similar genomic GC content, the observed divergence in genomic GC content must have occurred prior to the divergence of these contemporary species. The presence of closely-related species with similar genomic GC content also implies that current genomic GC content is relatively stable on an evolutionary time scale. Further, since random mutation would eventually lead to a 50% GC content, this stability must reflect the presence of some mechanism(s) for preserving GC content. Taken together, these data indicate that divergent bacterial genomic GC content has evolved repeatedly and is actively maintained by contemporary bacterial species. The observed variation in genomic GC content has a strong impact on synonymous codon usage with most of the variation occurring in the third codon position due to the redundancy of the genetic code [@pone.0017677-Muto1], [@pone.0017677-Knight1], [@pone.0017677-Fryxell1]. However, reproducible changes in the GC content of the first and second codon positions occur as well, as illustrated in [Fig. 2](#pone-0017677-g002){ref-type="fig"}. Since more than 1000 complete bacterial genomes are now available, we decided to explore this phenomenon further to determine if the patterns of codon usage previously observed in bacteria with GC-rich genomes [@pone.0017677-Chen1] were replicated across a broad range of bacterial taxa. Our analyses show that, within each bacterial phylum and class analyzed in this study, the use of GC-rich codons increases as genomic GC content increases ([Fig. 5](#pone-0017677-g005){ref-type="fig"}). Furthermore, the data indicate that for leucine and arginine, each of which are encoded by two codon families that differ in GC content, the codons with the higher GC content are found preferentially in genomes with high GC content ([Fig. 6](#pone-0017677-g006){ref-type="fig"} and [7](#pone-0017677-g007){ref-type="fig"}). For leucine, the synonymous codons range from no GC content (UUA) to two thirds GC (CUC, CUG) allowing for a greater shift in GC content than is possible with four-fold degenerate codons. Similarly, arginine codons range from one third GC (AGA) to 100% GC (CGG and CGC). In fact, these first position synonymous substitutions account for most of the difference in the slopes of plots of the first position GC content versus total genomic GC content ([Fig. 6](#pone-0017677-g006){ref-type="fig"} and [7](#pone-0017677-g007){ref-type="fig"}). Again, this consistency across the eight classes and phyla of Gram-negative and Gram-positive bacteria included in this study indicates that the observed patterns are a result of shared GC content rather than shared ancestry. As shown previously [@pone.0017677-Lobry1], we demonstrated that for each class and phylum analyzed, the proportion of amino acids that are encoded by high GC codons increases as genomic GC content increases ([Fig. 4](#pone-0017677-g004){ref-type="fig"}). Thus, increased use of alanine, arginine, glycine, and proline is paired with decreased use of asparagine, lysine, and isoleucine. Furthermore, our analyses show that this pattern is consistent across all eight phyla and classes of bacteria analyzed in this study ([Fig. 3](#pone-0017677-g003){ref-type="fig"}--[](#pone-0017677-g004){ref-type="fig"} [](#pone-0017677-g005){ref-type="fig"} [](#pone-0017677-g006){ref-type="fig"} [7](#pone-0017677-g007){ref-type="fig"}). Thus, in addition to causing changes in synonymous codon use, changes in genomic GC content have a strong effect on the amino acid composition of the encoded proteins. These results indicate that considerable variation in amino acid content can be tolerated in most bacterial proteins without causing a significant impact on protein function. This consistency of the impact of genomic GC content on patterns of protein amino composition across bacterial phyla and classes provides additional support for the idea that genomic GC content is a driving force in genome evolution. GC content not only affects the proportions of amino acids encoded by a genome, but also the sequence complexity of the encoded proteins. Of the 10 two-fold and three-fold degenerate amino acid codon families, five have an A or T at both of the first two positions and five are GC neutral at the first two positions. In the four-fold degenerate codon families, three have either G or C at both of the first two positions and two are GC neutral. Since genomes with high GC content have increased GC content in the first and second codon positions, there is increased use of the three GC-rich codon families and decreased use of the five GC-poor codon families in protein coding regions. The reverse is true in low GC genomes. Thus, both high GC and low GC genomes have a limited amino acid vocabulary compared to that of mid-GC genomes. Furthermore, the lower number of GC-rich codon families causes this effect to be greater in high-GC genomes, resulting in greater genome homogeneity as previously observed by Bohlin and Skjerve [@pone.0017677-Bohlin1]. Bohlin and Skjerve\'s work [@pone.0017677-Bohlin1] also showed a correlation between GC content and aerobiosis which supports the hypothesis that increased genomic GC concentration evolved in response to rates of mutation associated with the use of oxygen in metabolism. Previously, Naya et al. [@pone.0017677-Naya1] proposed that increased GC content would be advantageous in aerobic organisms because those amino acids that are preferentially oxidized (Cys, Met, Trp, Tyr, Phe, and His) have reduced frequencies in aerobic bacteria. The codons for four of these amino acids (Cys, Met, Tyr, and Phe) have an A or T in both of the first two positions, and reduction in their use would increase genomic GC content. Naya et al. [@pone.0017677-Naya1] also noted that increased GC content makes the genome less susceptible to aerobiosis-related deleterious mutations either by guanine scavenging of oxidizing agents that protects other bases, or by withstanding mutations more easily in the third position because of the increased use of 4-fold degenerate amino acids. However, if aerobiosis was the driving force in the evolution of high-GC genomes, we would not expect to find anaerobic bacteria with high-GC genomes. Nevertheless, anaerobes such as *Desulfovibrio vulgaris, Halorhodospira halophila*,and *Anaeromyxobacter dehalogenans* have extremely high GC genomes and their patterns of codon usage are identical to those of the aerobic bacteria whose genomes have similar GC content. There are also examples of low GC aerobic bacteria such as *Ehrlichia chaffeensis*, *Buchnera aphidicola,* and *Candidatus carsonella rudii*. These bacteria, however, are intracellular parasites and, in the cases of *B. aphidicola and C. carsonella rudii,* have extremely small genomes. Since small genomes are correlated with low GC content [@pone.0017677-Musto1], it is not clear which factor might be a driving force. While these exceptions do not rule out the possibility that aerobiosis may be a factor that influences bacterial genomic GC content, they suggest that it is not the only factor. Genomic GC content also has been correlated with optimal growth temperature. Musto et al. [@pone.0017677-Musto2], [@pone.0017677-Musto3] used an approach similar to ours and showed that within families a significant correlation between growth temperature and GC content was observed in 9 out of 20 families. Thus, optimal growth temperature may influence genomic GC content in some bacterial families. Genome size also correlated with genomic GC content except with anaerobic bacteria [@pone.0017677-Musto1]. Thus, multiple factors have been correlated with genomic GC content for various subsets of bacteria. However, no cause and effect relationships have been established so it not clear whether genomic GC content influences the probability of the success of a particular lifestyle, or whether a particular lifestyle influences genomic GC content. Since life itself is complex, and most bacteria are constantly subjected to changing environments, it is likely that the apparent status quo is maintained by the interplay of conflicting forces. Despite the uncertainty about the factors responsible for the origin and maintenance of the observed disparate bacterial genome GC content, it seems clear that genomic GC content determines bacterial codon usage since similar patterns of codon use were observed in each of the eight phyla and classes included in this study. Specifically, in every group, codons with higher GC content were used with increasing frequency as genomic GC content increased ([Fig. 5](#pone-0017677-g005){ref-type="fig"}). In addition, Chen et al. [@pone.0017677-Chen2] showed that codon usage patterns could be predicted from an analysis of the noncoding regions of the genome. In the alternative scenario where amino acid content is the driving force, changes in codon usage would have caused changes in genomic GC content, and therefore, selection for changes in GC content of the noncoding regions would not be expected. Furthermore, if individual genera were evolving the preferential use of specific codons for specific amino acids, we might expect to find genomes that use GC-rich codons for some amino acids and AT-rich codons for others. This pattern was not observed for any of 78 species studied. Instead, the codons preferred for individual amino acids always reflect the genomic GC content suggesting that in bacteria, codon usage has had to adapt to a particular genomic GC content. Thus, genomic GC content appears to be the primary determinant of both codon and amino acid usage patterns. As a result, the observed selection for translational efficiency of highly expressed genes within a given genome must be constrained by the established parameters of that genome as suggested by the analyses of Chen et al. [@pone.0017677-Chen2]. Materials and Methods {#s4} ===================== Five bacteria phyla (*Actinobacteria, Bacteriodetes/Chlorobi, Cyanobacteria, Firmicutes,* and *Proteobacteria*) with a sufficiently large number of sequenced genomes (\>15 species) were identified in the NCBI Genome Database (<http://www.ncbi.nlm.nih.gov/sites/entrez?db=genome>). In addition, genome sequences for more than 15 species were available in four classes of the *Proteobacteria*, *alpha, beta, gamma*, and *delta.* To choose species for further analyses, the available species in each of these phyla and classes were ordered according genomic GC content, and the species with the lowest genomic GC content was selected from each 5% GC interval except in the last interval where two species, one with the highest and one with the lowest GC% content within the interval, were selected. The one exception was the *Betaproteobacteria* where two species were selected from each 5% interval because of the narrower range of GC% content (45--63%) in the available genomes. A list of the chosen bacterial species and their genomic GC content is shown in [Tables 1](#pone-0017677-t001){ref-type="table"} and [2](#pone-0017677-t002){ref-type="table"}. The coding regions of each of the 78 genomes were obtained from GenBank (<ftp://ftp.ncbi.nih.gov>) on June 16, 2010. Codon usage for each coding sequence was calculated using an original program (available upon request) written in Strawberry Perl (<http://strawberryperl.com/>), and the resulting codon usage tables were used to determine the amino acid composition of the translated genome. Artemis [@pone.0017677-Rutherford1] was used to compare the GC content of each reading frame within a species. Supporting Information {#s5} ====================== Table S1 ::: {.caption} ###### **Slope of a plot of codon use versus genomic GC content for codon families with non-neutral average GC content.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S2 ::: {.caption} ###### **Slope of a plot of codon use versus genomic GC content for codon families with neutral average GC content.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S3 ::: {.caption} ###### **Slope of a plot of codon use versus genomic GC content for six member codon families sub-divided by the first position base.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: The authors thank Jeremy Dietrick for expert technical assistance. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was funded in part by National Science Foundation (nsf.gov) grants DBI-0451403 and EF-0826792 to BE. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: JL NRF BE. Performed the experiments: JL. Analyzed the data: JL BE. Contributed reagents/materials/analysis tools: NRF. Wrote the paper: JL BE. [^2]: ¤a Current address: Ripon College, Ripon, Wisconsin, United States of America [^3]: ¤b Current address: Vanderbilt University, Nashville, Tennessee, United States of America

PubMed Central

2024-06-05T04:04:19.847884

2011-3-10

PMC3053388

Introduction {#s1} ============ The incidence of type 2 diabetes increases with age and one in five individuals are affected by the age of 60 [@pone.0017858-Prevention1]. The natural history of type 2 diabetes is progressive. In its early stages, type 2 diabetes manifests as impaired glucose tolerance and defective insulin secretion which occur in the presence of an intact β-cell mass [@pone.0017858-Prentki1]. This is followed by a frank loss of β-cells and a concomitant need for insulin therapy [@pone.0017858-Prentki1], [@pone.0017858-Butler1], [@pone.0017858-Yoon1]. Recent genomic studies have underscored the influence of inherited factors that affect β-cell integrity and function in age-related diabetes [@pone.0017858-Saxena1], [@pone.0017858-Scott1], [@pone.0017858-Zeggini1], [@pone.0017858-Lyssenko1]. While these germline factors are hypothesized to play a role in the pathogenesis of diabetes, the biology that underlies their increasing penetrance with age is not understood. Telomeres shorten progressively with cell division, and short telomeres activate a DNA damage response that leads to apoptosis and senescence [@pone.0017858-Harley1], [@pone.0017858-Lee1]. Telomerase synthesizes new telomere repeats onto chromosome ends to offset in part this telomere shortening [@pone.0017858-Greider1], [@pone.0017858-Greider2]. Mutations in *TERT*, the telomerase reverse transcriptase, and *TR*, the telomerase RNA, cause telomere shortening and a degenerative organ failure syndrome that manifests prominently in tissues of rapid turnover: the skin, mucosa and bone marrow [@pone.0017858-Armanios1], [@pone.0017858-Vulliamy1]. This disease complex is often recognized in dyskeratosis congenita (DC), a premature aging syndrome defined by classic mucocutaneous features [@pone.0017858-Armanios2]. In DC and related disorders, short telomeres cause stem cell loss and progressive organ dysfunction which leads to premature mortality due to bone marrow failure and pulmonary fibrosis (reviewed in [@pone.0017858-Armanios2]). Even when the telomerase locus is wild-type, short telomeres are sufficient to cause age-associated degenerative disease similar to the DC phenotype [@pone.0017858-Hao1], [@pone.0017858-Armanios3]. This observation has pointed to short telomere length as the relevant lesion in the setting of mutant telomerase genes, and more commonly, because telomere length is polymorphic, as an important genetic determinant of age-related disease [@pone.0017858-Armanios3]. Because β-cell function declines with age, we hypothesized that short telomeres might contribute to β-cell failure and to the increasing incidence of diabetes with age. We show, in a genetically defined model, that short telomeres are sufficient to impair glucose homeostasis. Mice with short telomeres have defects in insulin release. In *ex vivo* studies, we show short telomeres limit insulin exocytosis because of signaling defects. These include impaired mitochondrial membrane hyperpolarizaion and Ca^2+^ handling which are essential for intact insulin exocytosis. In the setting of ER stress, short telomeres compromise β-cell mass and worsen diabetes severity by inducing apoptosis. We also find a relatively increased incidence of diabetes in patients with DC who have short telomeres. Our data implicate telomere length as a critical determinant of β-cell function and diabetes risk. Results {#s2} ======= Mice with short telomeres have impaired glucose tolerance and glucose-stimulated insulin release {#s2a} ------------------------------------------------------------------------------------------------ To test whether short telomeres impair glucose homeostasis, we studied late generation CAST/EiJ mice that are heterozygous null for telomerase RNA, mTR^+/−^, and have short telomeres ([Figure S1A](#pone.0017858.s001){ref-type="supplementary-material"}). Although they were lean, mTR^+/−^ mice with short telomeres had fasting hyperglycemia compared with wild-type mice ([Figure 1A](#pone-0017858-g001){ref-type="fig"} and [Figure S1B](#pone.0017858.s001){ref-type="supplementary-material"}). When challenged in a 2 hour glucose tolerance test, mTR^+/−^ mice with short telomeres had relatively higher serum glucose levels compared with controls ([Figure 1B](#pone-0017858-g001){ref-type="fig"}). To determine whether the glucose intolerance was due to islet-intrinsic or extrinsic factors, we measured fasting insulin levels and found mTR^+/−^ mice with short telomeres had lower levels ([Figure 1C](#pone-0017858-g001){ref-type="fig"}). The insulin level was inappropriately low even after correcting for the serum glucose ([Figure 1D](#pone-0017858-g001){ref-type="fig"}). We next examined insulin release in response to a glucose stimulus, and found that mice with short telomeres had impaired insulin secretion ([Figure 1E](#pone-0017858-g001){ref-type="fig"}). There were no impairments in glucose uptake in an insulin tolerance test indicating mTR^+/−^ mice with short telomeres had no defects in peripheral insulin sensitivity ([Figure 1F](#pone-0017858-g001){ref-type="fig"}). To exclude a developmental defect, we examined islet architecture and histology but did not detect any abnormalities ([Figure S1C](#pone.0017858.s001){ref-type="supplementary-material"}--[1D](#pone.0017858.s001){ref-type="supplementary-material"}). Moreover, when we examined islets for insulitis, we found no infiltrates. β-cell mass quantifies the total insulin producing capacity accounting for the insulin positive area, the total pancreas section area, as well as the pancreas weight [@pone.0017858-Xu1]. When we measured β-cell mass, we found that it was intact (Figure 1SE). The insulin content and the percent of β-cells per islet were also intact in mice with short telomeres compared with controls, and there was no change in individual β-cell size (Figure 1SF--H). The glucose intolerance, lower insulin levels, and defective glucose-stimulated insulin release were also present in a second strain of mice with short telomeres \[mTR^−/−^ fourth generation (G~4~) mice\] on the C57BL/6 background. In this strain, telomere-mediated phenotypes are seen only when telomerase is null, and after successive generations of breeding [@pone.0017858-Lee1], [@pone.0017858-Blasco1] ([Figure S2A](#pone.0017858.s002){ref-type="supplementary-material"}--[2D](#pone.0017858.s002){ref-type="supplementary-material"}). The lower basal insulin levels and glucose-stimulated insulin release were not seen in early generation mTR^−/−^ G~1~ mice which have long telomeres ([Figure S2E--G](#pone.0017858.s002){ref-type="supplementary-material"}), indicating that the absence of telomerase alone was not sufficient to cause these defects. These data, in two independent genetic backgrounds, indicated that short telomeres impair glucose tolerance because of defective insulin release. This defect is independent of β-cell mass, size and insulin content. ::: {#pone-0017858-g001 .fig} 10.1371/journal.pone.0017858.g001 Figure 1 ::: {.caption} ###### Mice with short telomeres have defective insulin secretion. **A**. mTR^+/−^ with short telomeres have fasting hyperglycemia compared with wild-type mice (n = 25/group). **B**. Higher mean serum glucose during an intraperitoneal 2 hour glucose tolerance test in mice with short telomeres (n = 25/group). **C**. mTR^+/−^ mice with short telomeres have lower fasting insulin levels. **D**. The fasting insulin is lower even when corrected for serum glucose. **E**. When challenged with glucose, mTR^+/−^ with short telomeres have lower insulin levels. **F**. Insulin tolerance test shows mTR^+/−^ mice with short telomeres have intact peripheral insulin sensitivity. Serum glucose was measured after insulin injection at the timepoints shown. Mice were 3--4 months of age. For **C--F**, n = 10--14/group. Error bars represent SEM. \* indicates two-sided P-value\<0.05. :::  ::: Short telomeres cause mitochondrial dysfunction and impair Ca^2+^ handling in islets {#s2b} ------------------------------------------------------------------------------------ To probe the mechanisms underlying the telomere-mediated β-cell dysfunction, we isolated islets and examined the kinetics of insulin secretion *ex vivo*. In response to glucose stimulus, islets from short telomere mice had defective insulin release *in vitro* ([Figure 2A](#pone-0017858-g002){ref-type="fig"}). The defects were noted in both the first and second phases of insulin release, as well as at maximum release induced by arginine ([Figure 2A](#pone-0017858-g002){ref-type="fig"}). Insulin secretion was also impaired when exocytosis was directly stimulated by KCl depolarization ([Figure 2A](#pone-0017858-g002){ref-type="fig"}). These data established that short telomeres significantly impair insulin release. ::: {#pone-0017858-g002 .fig} 10.1371/journal.pone.0017858.g002 Figure 2 ::: {.caption} ###### Impaired insulin release, mitochondrial function and Ca^2+^ handling in islets with short telomeres. **A**. Dynamic insulin secretion in islets from mTR^−/−^G~4~ mice compared with wild-type mice. Bars above the traces indicate the duration of stimulation. 3G, 11G, 16.7G+Arg and 25KCl indicate 3 mM and 11 mM glucose, 16.7 mM glucose plus 20 mM Arginine, and 25 mM KCl, respectively. CCh indicates carbachol. Insulin was measured every 2 minutes. Data are mean insulin level (ng/ng DNA). \* Indicates P-value\<0.05 and \*\* ≤0.01 (one-sided). **B**. Measurements of mitochondrial membrane potential in response to 11 mM glucose in islets from mTR^−/−^G~4~ mice. The rhodamine 123 fluorescence decreases when the mitochondrial membrane polarizes. **C**. Data are means of decrease in % rhodamine 123 fluorescence normalized to carbonyl cyanide 4-(trifluoro-methoxy) phenylhydrazone (FCCP) depolarization. **D**. Decreased Δ peak \[Ca^2+^\]~i~ values in islets from mTR^−/−^G~4~ and control mice after stimulation with 11 mM glucose. (**E**) and (**F**) Fura-2 fluorescence ratio is shown. Bars above the traces indicate the duration of stimulation. 3G and 11G indicate 3 mM and 11 mM glucose, respectively. The traces are representatives of 12--13 experiments from four islet preparations. **G**. Glucose stimulated fast \[Ca^2+^\]~i~ oscillation frequency in isolated islets from mTR^−/−^G~4~ and control mice. **H**. Effects of adding 2.56 mM CaCl~2~ in the perfusion chamber on changes in \[Ca^2+^\]~i~, indicating Ca^2+^ influx over the plasma membrane. Example tracings of Fura-2 fluorescence ratio are shown. Bars above the traces indicate the duration of stimulation. 0 Ca^2+^ and 2.56 Ca^2+^ indicate 0 mM CaCl~2~ plus 2 mM EGTA and 2.56 mM CaCl~2~, respectively. The traces are representatives of 10 experiments from four islet preparations. **I**. Means in Δfura-2 fluorescence ratio is shown for (**H**). Mice were 10 months old (wild-type n = 5, mTR^−/−^G~4~ n = 6). Error bars represent SEM. For **B--I**, \*, \*\* and \*\*\* indicate two-sided P-value\<0.05, 0.01 and 0.001, respectively. :::  ::: Glucose-stimulated insulin release is signaled by an increase in the ATP/ADP ratio, closure of ATP-dependent K^+^-channels, depolarization of the β-cell plasma membrane, opening of voltage dependent L-type Ca^2+^-channels and an increase in cytoplasmic free Ca^2+^ concentration (\[Ca^2+^\]~i~), resulting in the release of insulin containing secretory granules [@pone.0017858-Yang1]. To define the mechanisms underlying the telomere-mediated defects in insulin secretion, we first measured mitochondrial membrane potential in isolated islets after a glucose stimulus. When the mitochondrial membrane polarizes, a change in rhodamine 123 fluorescence reflects the extent of mitochondrial membrane hyperpolarization [@pone.0017858-Yang1]. In response to glucose, we found a decrease in membrane hyperpolarization in islets with short telomeres, consistent with a defect in the respiratory chain ([Figure 2B--2C](#pone-0017858-g002){ref-type="fig"}). Since Ca^2+^ is the main trigger of exocytosis, we next examined its influx in response to glucose stimulation and found an impairment in \[Ca^2+^\]~i~ rise in islets with short telomeres ([Figure 2D](#pone-0017858-g002){ref-type="fig"}). It is well-established that subsequent to glucose stimulation, \[Ca^2+^\]~i~ increases and decreases in an oscillatory manner, reflecting an intricate balance between Ca^2+^ influx over the plasma membrane and Ca^2+^ mobilization from intracellular stores [@pone.0017858-Yang1]. These \[Ca^2+^\]~i~ oscillations are essential for appropriate β-cell function. Although islets from mice with short telomeres had intact slow \[Ca^2+^\]~i~ oscillations, there was a decrease in the frequency of fast \[Ca^2+^\]~i~ oscillations, indicating a functional defect in β-cell Ca^2+^ handling ([Figure 2E--2G](#pone-0017858-g002){ref-type="fig"}). When we examined Ca^2+^ influx over the plasma membrane in the absence of glucose, we also found impairments in mutant islets ([Figure 2H--2I](#pone-0017858-g002){ref-type="fig"}). Hence, the insulin secretion defect caused by short telomeres is multi-factorial and mediated by β-cell autonomous defects in both mitochondrial function as well as deteriorated Ca^2+^ handling. Short telomere islets have the hallmarks of senescence {#s2c} ------------------------------------------------------ The defects in insulin exocytosis suggested that DNA damage associated with short telomeres might provoke global β-cell dysfunction even when β-cell mass is preserved. To test whether in β-cells with short telomeres there is evidence of DNA damage, we examined 53BP1 foci and found an increase compared with controls ([Figure 3A](#pone-0017858-g003){ref-type="fig"}). Short telomeres are a potent inducer of senescence, we therefore examined the hallmarks of senescence and tested whether there was evidence of impaired proliferation and accumulation of cyclin dependent kinase inhibitors. At 6 weeks of age, β-cells in mTR^+/−^ mice with short telomeres had a lower proliferative fraction as measured by Ki-67 immunostaining of insulin positive cells ([Figure 3B](#pone-0017858-g003){ref-type="fig"}). The slower β-cell proliferation was also detected in C57BL/6 mTR^−/−^G~4~ mice ([Figure 3C](#pone-0017858-g003){ref-type="fig"}). We next isolated islets and quantitated p16*^INK4a^* transcript levels by real time RT-PCR. p16*^INK4a^* levels increased in wild-type mice with age, as shown previously [@pone.0017858-Krishnamurthy1] ([Figure 3D--3E](#pone-0017858-g003){ref-type="fig"}). However, at every age group we examined, short telomeres induced p16*^INK4a^* prematurely and, by 8 months of age, there was a 3-fold up-regulation compared to wild-type mice ([Figure 3D--3E](#pone-0017858-g003){ref-type="fig"}). The accumulation of p16*^INK4a^* was a direct consequence of the short telomeres, not the absence of telomerase, as telomerase null mice with long telomeres did not show this increase ([Figure 3E](#pone-0017858-g003){ref-type="fig"}). The accumulation was also specific to p16*^INK4a^* and not to other genes near the *INK4a* locus (e.g. p15) or other cyclin-dependent kinase inhibitors including the p53 target p21 ([Figure S3A](#pone.0017858.s003){ref-type="supplementary-material"}--[3C](#pone.0017858.s003){ref-type="supplementary-material"}). There was a trend toward accumulation of Arf, the other *INK4a* transcript, in islets from mice with short telomeres ([Figure S3D](#pone.0017858.s003){ref-type="supplementary-material"}). These data indicated that short telomeres impair the cell cycle progression of β-cells and cause premature accumulation of the senescence marker p16*^INK4a^* in pancreatic islets. ::: {#pone-0017858-g003 .fig} 10.1371/journal.pone.0017858.g003 Figure 3 ::: {.caption} ###### Islets from mice with short telomeres have hallmarks of senescence. **A**. DNA damage foci in β-cells from mTR^+/−^ late generation mice were detected by immunofluorescence against 53BP1 (green) in nuclei (blue). Analysis was limited to insulin positive cells and cells were deemed positive if they had at least one focus (n = 5 mice/group, 4 months old). **B**. Immunofluorescence images from wild-type and mTR^+/−^ mice with short telomeres. Proliferating β-cells are positive for both insulin (green) and Ki-67 (red) (n = 5 mice/group, 50 islets/mouse, 6 weeks old). **C**. The percent of β-cells with incorporated EdU was lower in mTR^−/−^G~4~ C57BL/6 mice after a 14 day pulse (n = 5/group, 50 islets/mouse, 6 months old). **D**. Relative expression of p16*^INK4a^* in islets shows a gradual increase with age in wild-type mice. mTR^−/−^G~4~ mice have higher levels at the age groups shown. **E**. p16*^INK4a^* expression in islets was not increased in mTR^−/−^G1 mice. For **D** and **E**, n = 3--6 mice/timepoint. **F**. Heat map of mRNA microarray data shows differential expression profiles in wild-type compared with mTR^−/−^G~4~ mice. The red color expresses genes that are up-regulated and green down-regulated genes. The fold-change based on color is shown in the key below **F**. Heat map is based on genes with 1.5-fold expression change (n = 3 mice/group, 15 months old). Values were normalized by subtracting the mean of all samples. Fold change indicated in the key is log~2~-based. **G**. β-cell relevant pathways that are altered in the gene ontology analysis plotted relative to the significance of the P-value. Only pathways with P-value\<0.01 were analyzed. The number of genes altered relative to the total retrieved is shown to the right of the bar graph. **H**. qRT-PCR verification of Reg gene family expression shows significant increases in mTR^−/−^G~4~ mice (n = 4 mice/group, 15 months old). Error bars represent SEM. \* indicates two-sided P-value\<0.05 and \*\* P-value\<0.01. :::  ::: Short telomeres cause gene-expression changes in signaling pathways {#s2d} ------------------------------------------------------------------- We next examined the mechanism by which short telomeres may affect β-cell function in the absence of β-cell loss. Senescence is associated with altered gene expression signatures that are cell- and context-dependent [@pone.0017858-Zhang1]. We tested whether transcriptional changes in islets may explain the functional defects due to short telomeres. Microarray analysis on purified islets from wild-type and mTR^−/−^G~4~ mice showed differential expression of 1,935 genes: n = 1,153 decreased (60%) and n = 782 up-regulated. A unique transcriptional signature distinguished islets with short telomeres from controls by clustering analysis ([Figure 3F](#pone-0017858-g003){ref-type="fig"}). Gene ontology revealed that multiple pathways essential for insulin secretion were affected including signal transduction, Ca^2+^-mediated exocytosis, Ca^2+^ homeostasis, and K^+^ ion transport ([Figure 3G](#pone-0017858-g003){ref-type="fig"} and [Table S1](#pone.0017858.s006){ref-type="supplementary-material"}). The ontology also implicated genes involved in cell cycle control and stress response processes ([Figure 3G](#pone-0017858-g003){ref-type="fig"} and [Table S1](#pone.0017858.s006){ref-type="supplementary-material"}). The genes altered in the stress response included the Reg gene family which had the highest fold up-regulation ([Figure 3H](#pone-0017858-g003){ref-type="fig"} and [Table S1](#pone.0017858.s006){ref-type="supplementary-material"}). The Reg genes were initially identified in the regenerating pancreas, and are known to accumulate in islets from patients with type 2 diabetes [@pone.0017858-Terazono1], [@pone.0017858-Marselli1]. These data indicated that short telomeres alter islet transcriptional programs; this signature affects multiple cellular processes which are essential for insulin secretion. Decreased β-cell survival due to short telomeres in the setting of ER stress {#s2e} ---------------------------------------------------------------------------- Because short telomeres cause β-cell dysfunction, we reasoned that telomere length will be a modifier of severity in monogenic forms of diabetes that affect β-cell integrity. We crossed the mTR null allele in C57BL/6 mice onto the diabetic Akita mouse that carries a mutation in the insulin gene, Ins2^C96Y/WT^, and generated double mutant mice that have short telomeres ([Figure S4A](#pone.0017858.s004){ref-type="supplementary-material"}). Mutations in the insulin gene cause diabetes in humans and in the Akita mouse where insulin misfolding leads to ER stress and apoptosis and clinically manifests as irreversible β-cell failure [@pone.0017858-Stoy1], [@pone.0017858-Wang1], [@pone.0017858-Oyadomari1]. In Ins2^C96Y/WT^ mice, diabetes developed as expected, and was severe in males [@pone.0017858-Oyadomari1]. We therefore studied double mutant female mice and hypothesized that short telomeres would modulate disease severity. By 8 months, we found greater impairments in glucose tolerance in Ins2^C96Y/WT^ mice that had short telomeres ([Figure 4A](#pone-0017858-g004){ref-type="fig"}). The glucose intolerance was associated with a loss of β-cell mass which did not occur in Ins2^C96Y/WT^ mutant mice with long telomeres ([Figure 4B](#pone-0017858-g004){ref-type="fig"}). Ins2^C96Y/WT^ mice with short telomeres also had lower serum insulin levels ([Figure 4C](#pone-0017858-g004){ref-type="fig"}). Consistent with a β-cell intrinsic defect, these phenotypes were not due to abnormalities in insulin resistance ([Figure S4B](#pone.0017858.s004){ref-type="supplementary-material"}). ::: {#pone-0017858-g004 .fig} 10.1371/journal.pone.0017858.g004 Figure 4 ::: {.caption} ###### Short telomeres worsen diabetes severity in Akita mice and cause β-cell loss. **A**. Two hour glucose tolerance test of Ins2^C96Y/WT^ mTR^−/−^iG~4~ with short telomeres shows more severe impairments compared with Ins2^C96Y/WT^ mice. **B**. Ins2^C96Y/WT^mTR^−/−^iG~4~ mice have decreased β-cell mass compared with Ins2^C96Y/WT^ mice. This decrease is associated with lower basal serum insulin levels (**C**). **D**. Representative images of TUNEL (red) co-staining with insulin (green) shows an increase in β-cell apoptosis as quantitated in the bar graph. Data shown are from 8 month old females (n = 5--10 mice/group). Error bars represent SEM. \* indicates two-sided P-value\<0.05. :::  ::: We next examined whether short telomeres affected the rate of β-cell apoptosis due to ER stress in Akita mice. In the absence of ER stress, the baseline cell death rate of β-cells was low, and there was no difference between wild-type and short telomere mice ([Figure 4D](#pone-0017858-g004){ref-type="fig"}). However, in the setting of ER stress, we identified a two-fold increase in TUNEL positive β-cells in Ins2^C96Y/WT^ mice with short telomeres compared with Ins2^C96Y/WT^ ([Figure 4D](#pone-0017858-g004){ref-type="fig"}). These data indicated that short telomeres cause additive β-cell dysfunction in the setting of ER stress, decreasing the threshold for apoptosis. This effect clinically manifests as more severe diabetes associated with loss of β-cell mass. Increased incidence of diabetes in DC {#s2f} ------------------------------------- Our data indicated that short telomeres cause β-cell dysfunction in mice. To address whether telomere length contributes to disease in humans, we asked whether patients with DC might be at risk for developing diabetes. We found several published case reports that have incidentally reported on individuals who carried the diagnosis of DC and who had impaired glucose homeostasis [@pone.0017858-Steier1], [@pone.0017858-Reichel1], [@pone.0017858-Walne1]. These individuals were diagnosed with glucose intolerance or diabetes at 20 months, 21, 22 and 50 years of age. In our registry of families with DC and related telomere syndromes, we queried 20 consecutive individuals from 7 families, and identified 3 individuals who developed insulin-dependent diabetes prior to the age of 30. The clinical details of 2 of the 3 individuals are listed in [Figure S5](#pone.0017858.s005){ref-type="supplementary-material"}. Given the expected incidence of diabetes in this age group (3 per 1000), the likelihood that diabetes would occur by chance alone is low (P\<0.001, chi-square test). Discussion {#s3} ========== The study of the telomerase knockout mouse has particular significance for understanding the genetics of type 2 diabetes since the short telomere defect that is present in this model is acquired universally in human aging. Over the last decade, telomere-mediated phenotypes identified in the mouse have consistently converged with the human age-related disease phenotypes [@pone.0017858-Armanios2]. Here we show that short telomeres are sufficient to recapitulate the glucose intolerance and insulin secretion defects which occur in the early stages of human age-related diabetes. Telomere length is a polymorphic trait, and short telomeres are a uniquely inherited genotype that can cause degenerative disease even when telomerase is wild-type [@pone.0017858-Armanios3]. In recent cross-sectional studies, short telomeres have been associated with type 2 diabetes [@pone.0017858-Adaikalakoteswari1], [@pone.0017858-Salpea1], [@pone.0017858-Olivieri1], [@pone.0017858-Sampson1]. Our data indicate that short telomeres are not simply associated with, but are a relevant modifier of diabetes risk and severity, and may be a valuable biomarker that identifies individuals at greatest risk in clinical settings. In our study of mice with short telomeres, we find both *in vivo* and *ex vivo* defects in β-cell insulin secretion. The insulin secretion impairments surprisingly occur in the presence of intact insulin content and β-cell mass. The defective insulin secretion is multi-factorial and appears to be due to simultaneous, independent impairments in key exocytosis pathways including glucose-dependent mitochondrial membrane hyperpolarization, as well as glucose-independent defects in Ca^2+^ handling. Our data implicate gene expression changes in the setting of senescence as underlying these defects. The transcriptional changes we identify affect pathways essential for β-cell function including Ca^2+^-mediated exocytosis. These findings are in contrast to a recent study which reported that insulin secretion defects in late generation mTR^−/−^ mice are due to a decrease in insulin positive cells [@pone.0017858-Kuhlow1]. Our *ex vivo* studies indicate that short telomeres are sufficient to cause insulin secretion defects even when β-cell number is intact. Our data support a model where short telomeres induce senescence-associated gene expression changes in β-cells; this program contributes to defective signaling and clinically manifests as impaired glucose homeostasis. Short telomeres are canonically known to cause degenerative disease by inducing a loss of stem cells [@pone.0017858-Lee1], [@pone.0017858-Hao1]. Our findings suggest that, in some settings, even when cell mass is preserved, short telomeres compromise organ function. Our data therefore provide evidence for a novel mechanism of telomere-mediated disease where short telomeres can compromise organ homeostasis in the absence of overt cell loss. In the Akita mouse we show that in the setting of ER stress, short telomeres cause additive injury that results in loss of β-cell mass. Short telomeres are a potent inducer of cell death [@pone.0017858-Lee1], and in the presence of ER stress, the additive damage signals from both telomere dysfunction and the misfolded protein response evoke an apoptotic program that is otherwise not apparent in the presence of short telomeres alone. In contrast to the glucose intolerance we observed in male mTR^−/−^G~4~ mice, we did not identify baseline impaired glucose tolerance in female mTR^−/−^iG~4~. This observation is consistent with the known increased penetrance of β-cell mediated diabetes in male mice [@pone.0017858-LeMay1]. ER stress and loss of β-cell mass occur in the late stages of diabetes [@pone.0017858-Prentki1]. Our data indicate that short telomeres cooperate with ER stress to worsen diabetes severity and, because they accumulate with age, may contribute to the progressive natural history of diabetes which culminates in β-cell loss. Therefore both the senescence and apoptotic downstream consequences of telomere shortening contribute to the β-cell degenerative phenotype. We identified an increased incidence of diabetes in a small cohort of DC patients with short telomeres. Idiopathic pulmonary fibrosis patients, who also have short telomeres, have been reported to be more prone to develop diabetes [@pone.0017858-Alder1], [@pone.0017858-Gribbin1], [@pone.0017858-Cronkhite1]. Diabetes incidence is known to be increased in other syndromes associated with DNA damage including Fanconi anemia where 10% of patients are affected [@pone.0017858-Elder1]. Its association with DC, if replicated in larger cohorts, would suggest that, in addition to bone marrow failure and fibrotic disorders, diabetes may be a manifestation of telomere-mediated disease in syndromes of telomere shortening, albeit at lower penetrance. In contrast to the bone marrow failure phenotype which is prominent in the setting of short telomeres, our data indicate that telomere reserves limit the function of tissues of relatively slow turnover such as β-cells. How do we explain that short telomeres limit the function of tissues of slow turnover? One model is that although β-cell have slow turnover, they still undergo several fold expansion by adulthood [@pone.0017858-Dor1]. Thus although the basal proliferation rate is slow, cell turnover superimposed on genetically determined shorter telomere length may be sufficient to precipitate organ dysfunction. It is also possible that, in addition, short telomere length may be a first insult that decreases the threshold for sustaining additional injury that accumulates with age such as ER stress, as our data indicate. The genetic evidence we show therefore supports that telomere function is essential for the integrity of tissues of slow turnover, and indicates that telomere shortening plays a role in a broader spectrum of specific age-related disorders than previously appreciated. In summary, short telomeres limit insulin secretion by affecting global β-cell signaling in the setting of senescence. With ER stress, short telomeres cause increased β-cell apoptosis and loss of β-cell mass. Our data identify impaired signaling and exocytosis as a novel mechanism of telomere-mediated disease, and establish telomere shortening as a potential genetic risk factor for diabetes contributing to both its onset and pathogenesis. Methods {#s4} ======= Mouse studies {#s4a} ------------- Mice were housed at the Johns Hopkins University and all procedures were approved by its Animal Care and Use Committee (approval ID MO07M492). We studied mTR^+/−^ mice on the Cast/EiJ background that were offspring of heterozygous breeding for 8--10 generations [@pone.0017858-Hao1]. C57BL/6 mTR^−/−^ mice were bred for four generations [@pone.0017858-Blasco1]. Glucose levels were measured using a home glucometer. For glucose tolerance tests, mice were fasted overnight then injected with glucose 2.5 g/kg intraperitoneally. Serum insulin levels were measured by ELISA (Mercodia). For insulin tolerance tests, mice were injected with insulin 0.5 U/kg. To quantitate the insulin positive area, we performed immunohistochemistry on paraffin-embedded sections, 140 microns apart, and quantitated the area using Nikon Elements software. β-cell mass was determined by calculating the ratio of the insulin positive area to total pancreas sectional area, and multiplying by the pancreas weight [@pone.0017858-Xu1]. For insulin content, whole pancreas homogenate was extracted in acid/ethanol (0.18M HCl/70% ethanol) overnight at 4°C and insulin levels were determined by ELISA after normalizing for total protein. We obtained C57BL/6 Ins2^C96Y/WT^ mice from the Jackson Laboratory. We generated Ins2^C96Y/WT^ with short telomeres using an intergenerational strategy [@pone.0017858-Hemann1]. All experiments included age-matched male mice, except in the Akita studies where we studied the more attenuated phenotype in females. Islet in vitro studies {#s4b} ---------------------- Dynamic insulin release from islets was analyzed [@pone.0017858-Healy1]. Insulin measurements were performed by a microsphere-based two-photon excitation fluorometer (TPX-technology; ArcDia Diagnostics). Insulin data were normalized to DNA content using PicoGreen (Invitrogen). Mitochondrial membrane potential was analyzed using rhodamine 123 [@pone.0017858-Silva1]. To measure \[Ca^2+^\]~i~. islets were incubated with 2 µM fura-2 AM, and changes in fluorescence ratio 340/380 were analyzed [@pone.0017858-Berggren1]. Experiments were performed on a Zeiss Axiovert 200 M with a fluorescence imaging system. \[Ca^2+^\]~i~ oscillations were analyzed using power spectral analysis in Matlab (The Mathworks, Inc.) [@pone.0017858-Uhlen1]. The amplitudes of fast and slow oscillations were calculated as the square root of the total power of periods from 6 to 60 s (FastOsc), and 60 to 600 s (SlowOsc), respectively [@pone.0017858-Stein1]. Power spectral density for fast oscillations was calculated by the Welsh method [@pone.0017858-Berggren1], and standard fast Fourier transform power spectrum was used for slow oscillations. The dominant fast and slow periods were obtained from peaks in respective power spectrum. Immunofluorescence, qRT-PCR and microarray {#s4c} ------------------------------------------ We measured telomere length using quantitative fluorescence in situ hybridization (FISH) [@pone.0017858-Alder1] and detected DNA damage using 53BP1 immunuofluorescence (Novus) [@pone.0017858-dAddadiFagagna1]. Antibodies for insulin, glucagon, somatostatin and Ki-67 were obtained from Dako. We delivered EdU (2 mg, Invitrogen) over a 14 day period via an implanted subcutaneous pump (Alzet) and used reagents in Click-iT EdU (Invitrogen) for detection. For apoptosis studies, we performed the TUNEL assay (Roche). All histology studies were quantitated blinded to genotype. For expression studies, we purified islets [@pone.0017858-Li1], and extracted total RNA using RNeasy (Qiagen). To obtain sufficient RNA, we isolated islets from mTR^−/−^ mice on the C57BL/6 background which have larger islets. We performed quantitative real time RT-PCR using iQ SYBR Green Supermix (BioRad, Hercules, CA. Primer sequences are listed in [Table S2](#pone.0017858.s007){ref-type="supplementary-material"}. The expression of each gene was normalized to Hprt levels. Transcriptional profiling of purified islets was performed using Mouse Exon 1.0 ST Arrays (Affymetrix) at the Johns Hopkins Microarray Facility. CEL file data were extracted and normalized with Partek® Genomics Suite™ software using the Robust Multichip Analysis algorithm [@pone.0017858-Irizarry1]. To explore the broadest range of transcripts, Affymetrix extended probes were imported, yielding transcripts of which 112,207 have annotation with 50,601 at the gene or mRNA-level. Genes with greater than 1.5 fold change and P-value\<0.05 were considered significant. We used the Spotfire (TIBCO) platform and Gene Ontology (August 2010) to analyze gene expression. Microarray data is MIAME compliant and that the raw data has been deposited in GEO (GEO Series GSE25040). Statistical analyses were performed using GraphPad Prism software (La Jolla), and means were compared using Student\'s *t*-test. Subjects and ethics statement {#s4d} ----------------------------- Probands and family members were evaluated through the Johns Hopkins Hospital, and written informed consent was obtained from all participants. The Johns Hopkins Medicine Institutional Review Board approved the study. Individuals were eligible if they had classic mucocutaneous features of DC, or carried a mutant telomerase gene. We performed telomere length analysis using flow-cytometry and FISH, and sequenced the known DC genes from genomic DNA [@pone.0017858-Alder1]. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### **Mice with short telomeres have intact islet morphology and β-cell mass.** **A**. Late generation mTR^+/−^ mice have short telomeres in islets as measured by quantitative fluorescence in situ hybridization. Representative images in the left panels show decreased intensity of telomere signals (red). Quantitation is shown in the bar graph (n = 5 mice/group, 15 islets/mouse, 3--4 months). **B**. mTR^+/−^ mice with short telomeres are lean (n = 21--24 mice/group, 3--4 months). **C**. Representative immunofluorescence images from wild-type and mTR^+/−^ mice with short telomeres show intact β- and α-cell histology as shown by the co-staining of insulin (green) and glucagon (red). **D**. β-cells have normal appearance and relationship to δ-cells as shown by the co-staining of insulin (green) with somatostatin (red). **E**. β-cell mass is intact in mTR^+/−^ mice with short telomeres. **F**. Total insulin content, the percentage of β-cells relative to total islet cells (**G**), and individual β-cell size (**H**) are similar in mTR^+/−^ mice with short telomeres and wild-type mice. For **G&H**, more than 1000 cells were analyzed per mouse. For **E--H**, n = 5--8 mice/group, 3--4 months of age. Error bars represent SEM. \* indicates two-sided P-value\<0.05 and \*\* P-value\<0.01. (EPS) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### **C57BL/6 mTR^−/−^ mice with short telomeres have impaired glucose homeostasis due to defective insulin release.** **A**. Two hour intraperitoneal glucose tolerance test shows mTR^−/−^ G~4~ mice were more glucose intolerant (n = 17--20/group, 6--8 months). **B**. mTR^−/−^G~4~ mice tended to have lower fasting insulin levels, and significantly lower fasting insulin/glucose ratios (**C**). **D**. At 30 minutes after glucose injection, mTR^−/−^G~4~ mice released less insulin in response to an intraperitoneal glucose load. **E&F**. Early generation mTR^−/−^ mice (G~1~) have fasting insulin levels and fasting insulin/glucose ratio similar to wild-type controls. **G**. At 30 minutes after intraperitoneal glucose injection, mTR-/-G~1~ mice had similar insulin levels compared with controls. For **E--F**, n = 5--8/group, 10--11 months of age. Error bars represent SEM and \* indicates two-sided P-value\<0.05. (EPS) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S3 ::: {.caption} ###### **Accumulation of cyclin-dependent kinase inhibitors in short telomere islets is specific to p16** ***^INK4a^*** **.** **A**. The relative expression of p15*^INK4b^* does not increase with age or with telomere shortening. **B&C**. Expression of p27 and p21 does not change drastically in pancreatic islets with age nor with short telomeres. **D**. There is a trend towards accumulation of the Arf transcript with age and with telomere shortening. 3--6 mice were analyzed for each genotype and timepoint. (EPS) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S4 ::: {.caption} ###### **Breeding scheme for introducing the mTR^−/−^ allele and short telomeres onto the Akita (Ins2^C96Y/WT^) mouse.** **A**. Intergenerational cross of C57BL/6 Ins2^C96Y/WT^ mice with C57BL/6 mTR^+/−^ mice to generate double mutant mice was first performed. The mice were then crossed with mTR^−/−^G~3~ mouse to introduce the short telomere background. **B**. Double mutant mice have no defect in peripheral glucose uptake in this insulin tolerance test (n = 5 mice/group, 8 months old females). (EPS) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S5 ::: {.caption} ###### **Clinical features of probands with dyskeratosis congenita and diabetes.** **A**. Clinical details of 2 patients with diabetes and dyskeratosis congenita. Case 1 had classic dyskeratosis congenita features and, in Case 2, we identified a C→G heterozygous substitution at nucleotide 204 in *hTR* which impairs telomerase activity (P\<0.001, not shown). **B**. Case 1 and Case 2 have lymphocyte telomere length below the 1^st^ percentile compared with age-matched controls. Percentile lines were generated based on data from 400 controls. This range is highly predictive of a germline defect in telomerase (Alder et al *PNAS* 2008). **C**. X-linked pattern of dyskeratosis congenita inheritance in the family for index case 1. **D**. The family for index case 2 displays autosomal dominant inheritance. Both probands have relatives with a history of pulmonary and liver disease along with aplastic anemia, consistent with the diagnosis of an inherited telomere syndrome. (EPS) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S1 ::: {.caption} ###### **Biological processes and associated genes altered in microarray expression analysis of pancreatic islets from mice with short telomeres.** \*Genes listed have greater than 1.5 fold expression change. (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S2 ::: {.caption} ###### **Primers used to measure expression of cyclin-dependent kinase inhibitors and Reg gene family members by qRT PCR.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: We are grateful to the subjects who participated in this study. We are thankful to Dr. Carol Greider, Dr. David Valle and Dr. Brendan Cormack for critical reading of the manuscript. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was supported by grants from the National Institutes of Health: A Ruth L. Kirschstein Award, GM007309, CA118416, AG27406 and awards from the Maryland Stem Cell, Sidney Kimmel, and Doris Duke Charitable Foundations. The in vitro islet work was supported by the Swedish Research Council and the Family Erling-Persson Foundation. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: MA NG P-OB. Performed the experiments: NG EP L-SL FK NL. Analyzed the data: NG P-OB MA. Contributed reagents/materials/analysis tools: L-SL MH P-OB. Wrote the paper: NG MA.

PubMed Central

2024-06-05T04:04:19.852148

2011-3-10

PMC3053389

Introduction {#s1} ============ The transfer of genetic information across normal mating barriers, also known as horizontal or lateral gene transfer (HGT), has long been recognized as one of the major forces driving prokaryote evolution, but has generally been seen as more limited in eukaryotic genomes [@pone.0017872-Keeling1], [@pone.0017872-MarcetHouben1]. The increasing number of available genomic sequences has recently transformed this vision, as numerous cases of HGT have emerged in a wide variety of eukaryotic lineages (reviewed in [@pone.0017872-Keeling1], [@pone.0017872-Keeling2]). Several studies have reported the HGT of sequences of bacterial origin to the genome of *S. cerevisiae* [@pone.0017872-Dujon1], [@pone.0017872-Fitzpatrick1], [@pone.0017872-Hall1], [@pone.0017872-Rolland1]. Introgressions between closely related yeast species [@pone.0017872-Liti1], [@pone.0017872-Muller1] or between varieties of the basidiomycete yeast *Cryptococcus neoformans* [@pone.0017872-Kavanaugh1], [@pone.0017872-Richards1] have also been described, suggesting that this mechanism is more widespread than previously thought in yeast genomes. We recently reported the occurrence of HGT between more distantly related yeast species. The genome sequence of the commercial wine yeast strain EC1118 contains three gene clusters resulting from horizontal transfers [@pone.0017872-Novo1], two of which are widespread among wine yeasts. The genes in these regions encode proteins involved in key metabolic functions during winemaking [@pone.0017872-Novo1], [@pone.0017872-Galeote1], strongly suggesting that HGT is one of the mechanisms by which wine yeast strains adapt to their high-sugar, low-nitrogen environment. The donor of one of these introgressions (region B, 17 kb) was identified as *Zygosaccharomyces bailii*, a major contaminant of wine fermentations, supporting the view that genetic exchange is favored by ecological proximity. The sequences of region B and of the homologous region in *Z. bailii* are almost identical (99.7%), suggesting that this introgression event is recent. However, despite the high degree of sequence similarity, differences in gene organization were found between the *Z. bailii* and EC1118 sequences, suggesting that some reorganization occurred after the transfer [@pone.0017872-Novo1]. The mechanisms underlying the acquisition and reorganization of this DNA fragment have yet to be elucidated. We report here the presence of multiple copies of the *Z. bailii* DNA segment in the diploid genome of EC1118, inserted into different chromosomes and displaying changes in structural organization. A broader survey of this region in the European yeast population revealed the presence of up to eight different forms, mostly in wine yeast strains. The structural organization of this sequence, the presence of a functional ARS and an analysis of the insertion breakpoints strongly suggested an expansion mechanism involving the formation of an extrachromosomal circle DNA (eccDNA) molecule and its integration into the yeast genome by nonhomologous recombination. Results {#s2} ======= Identification of three copies of region B in the genome of EC1118 {#s2a} ------------------------------------------------------------------ In the genome sequence of strain EC1118, we previously identified three large chromosomal segments not found in the S288C reference genome, called regions A, B and C, on chromosomes VI, XIV and XV, respectively [@pone.0017872-Novo1]. We looked for these three regions in the genome sequence of 59A, a haploid derivative of EC1118. These three regions are present in the 59A strain, but the 17 kb region B, donated by *Z. bailii*, was found on chromosome X, conflicting with its location on chromosome XIV in the EC1118 diploid genome assembly. An analysis of the number of reads obtained during sequencing of the EC1118 genome was consistent with the presence of one copy of region A and C but suggested that there were three copies of region B ([Figure 1](#pone-0017872-g001){ref-type="fig"}). Southern blot hybridization on the chromosomes of EC1118 confirmed that three copies of this region were present in EC1118: the copy initially described on chromosome XIV and two additional copies found in chromosomal bands attributed to chromosomes X and XII ([Figure 2](#pone-0017872-g002){ref-type="fig"}). This analysis also confirmed the presence of this region in *Z. bailii* and showed that the 59A strain inherited the copy located on chromosome X. ::: {#pone-0017872-g001 .fig} 10.1371/journal.pone.0017872.g001 Figure 1 ::: {.caption} ###### Estimation of the number of copies of region B in the diploid genome of strain EC1118. The number of sequencing reads covering a nucleotide position was used to identify changes in sequence copy number. The diploid chromosomal region, present in two copies (left side arrow), has 2n = 18 reads, whereas region B was found in three copies (arrow on the right) with 3n = 27 reads. :::  ::: ::: {#pone-0017872-g002 .fig} 10.1371/journal.pone.0017872.g002 Figure 2 ::: {.caption} ###### Chromosomal location of region B in different strains by Southern blot hybridization of PFGE gels. A. Chromosome separation by PFGE. Chromosomes were numbered according to the *Saccharomyces cerevisiae* nomenclature. B. Hybridization of a PFGE blot with a region. B. specific probe (gene 0023g). Positive signals were detected on chromosomes XII, XIV and X for the diploid strain EC1118 and only on chromosome X for the meiotic spore 59A :::  ::: By carrying out a BLAST similarity search on the sequencing reads of the EC1118 genome, we identified the insertion point on chromosome XII (EC1118\_XII form). Region B is located between the *YLR379W* (726 bp downstream from the stop codon) and *YLR380W* (*CSR1,* 261 bp upstream from the start codon) genes, at position 878,021 on S288C chromosome XII ([Figure 3](#pone-0017872-g003){ref-type="fig"}). We also characterized the insertion point on chromosome X (EC1118\_X form), by analyzing the 59A genome. As previously described for EC1118, the right arm of chromosome X was greatly rearranged [@pone.0017872-Novo1] Indeed, a 5 kb region from the left arm of chromosome VI encompassing *YFL058W* (*THI5*) and *YFL062W* (*COS4*) was found inserted into the right telomeric end of chromosome X. This rearrangement may have occurred by homologous recombination between two almost identical (98%) *THI* genes (*YJR156C* and *YFL058W*). We found that region B was integrated into the *COS4* gene. Further confirmation of the breakpoint positions on chromosomes XII and X was obtained by PCR amplification with junction-specific primers for EC1118 and 59A. ::: {#pone-0017872-g003 .fig} 10.1371/journal.pone.0017872.g003 Figure 3 ::: {.caption} ###### Localization and organization of region B in various strains. Colored arrows represent syntenic ORFs, numbered according to Novo *et al*. [@pone.0017872-Novo1]. The points of insertion of region B are numbered according to position in the putative circular form proposed in [Figure 4](#pone-0017872-g004){ref-type="fig"}. The chromosomal insertion positions are indicated according to the chromosome sequences of S288C. :::  ::: Localization and structural organization of B regions in other *S. cerevisiae* strains {#s2b} -------------------------------------------------------------------------------------- We previously showed that region B is present in many *S. cerevisiae* strains, mostly of wine origin [@pone.0017872-Novo1]. We carried out BLAST similarity searches of the yeast genome sequences available to date. We found that 24 of the 69 yeast genomes checked contained region B ([Table S1](#pone.0017872.s002){ref-type="supplementary-material"}). As most of these genomes are available as draft assemblies, we could precisely locate the insertion in only five strains ([Figure 3](#pone-0017872-g003){ref-type="fig"}). In the vineyard-derived RM11-1a strain, the region is present on the left arm of chromosome XIV, between the *YNL249C* and *YNL248C* genes. The vineyard isolate M22 and the bioethanol production strain derivative JAY291 carry a copy on the right arm of chromosome XIII and on the left arm of chromosome XI, respectively. The T73 strain, a haploid derivative of a commercial wine yeast strain, and YJM280, a clinical isolate, had insertions points similar to those of EC1118\_X. We were also able to determine one of the two junctions for four other strains. In the WE372 and CLIB382 strains, the proximal junction is the same as that of M22 on chromosome XIII whereas, for CBS7960 and CLIB324, the proximal junction is identical to that of the EC1118\_X form. All forms except EC1118\_X, which was found in subtelomeric position, were inserted at internal positions on the chromosome. Unexpectedly, we also observed variations in gene organization for all these regions. ([Figure 3](#pone-0017872-g003){ref-type="fig"}). All five genes were consistently present, but their relative positions varied. The EC1118\_XII form exhibits conserved synteny to the *Z. bailii* sequence. For the other forms, the B, C and D genes were syntenic, but the A and E genes varied in position, being located either upstream or downstream from B, C, D group. Genes A and E were sometimes found to have been broken in two, as observed for gene A in the EC1118\_XIV form and gene E in the EC1118\_X form ([Figure 3](#pone-0017872-g003){ref-type="fig"}). All these data, including the structural organization of the B regions in the different strains, suggested that that these regions were integrated into the various genomes as closed circular molecules ([Figure 4](#pone-0017872-g004){ref-type="fig"}). Furthermore, two copies of an 11 bp ARS consensus sequence (ACS, 5′-WTTTAYRTTTW-3′) were identified in the sequence of region B, at position 8524 to 8535, referring to the circular form, (ACS1, ATTTATATTTT) and 13763 to 13774 (ACS2, TTTTATATTTT). ACS sequences have been identified as essential domains of the ARS element of *S. cerevisiae* [@pone.0017872-Broach1], [@pone.0017872-VanHouten1]. These findings suggest that region B may replicate autonomously in *S*. *cerevisiae.* To address this possibility, we inserted the region encompassing ACS2 in a YIp integrative vector that was then used to transform yeast cells. Transformants were obtained with an efficiency similar to that obtained with a replicative vector, providing evidence that this ARS element is functional in *S. cerevisiae*. ::: {#pone-0017872-g004 .fig} 10.1371/journal.pone.0017872.g004 Figure 4 ::: {.caption} ###### Putative circular form of region B. The insertion points found in various *S. cerevisiae* genomes are indicated by a tick. The two ARS consensus sequence (ACS) are represented by yellow circles. :::  ::: Analysis of breakpoint sequences {#s2c} -------------------------------- We investigated the mechanisms by which region B had integrated into the genome, by analyzing the nucleotide sequences surrounding the insertion breakpoints in six different chromosomes ([Figure 5](#pone-0017872-g005){ref-type="fig"}). The breakpoints and their environment were unique for each insertion. No repetitive elements were found around the breakpoints. At four junctions, a two- to three-nucleotide sequence was found to be common to the sequences of the inserted region and the chromosome ([Figure 5](#pone-0017872-g005){ref-type="fig"}). This limited sequence identity suggests the involvement of nonhomologous end joining (NHEJ), a pathway responsible for repairing double-strand breaks in DNA [@pone.0017872-Moore1]. An analysis of the chromosomal sequences at the breakpoints revealed that the integration of region B was generally not accompanied by changes at the junction site, although a loss or gain of 1 or 2 nucleotides was observed in two cases ([Figure 5](#pone-0017872-g005){ref-type="fig"}). We cannot rule out the possibility that these nucleotides are point mutations present on the original chromosome, but they may originate from an addition or deletion event occurring during the repair process, providing further evidence for the role of NHEJ in these insertions. ::: {#pone-0017872-g005 .fig} 10.1371/journal.pone.0017872.g005 Figure 5 ::: {.caption} ###### Analysis of the insertion breakpoints in different genomes. The left panel shows the location of region B on the various chromosomes (green boxes). The centromere is represented by a constriction in chromosomes. The right panel shows the chromosomal sequences at the insertion points with the highlighted inserted regions. Deleted and duplicated nucleotides are shown in parentheses, with a minus or plus sign, respectively. Underlined nucleotides indicate sequence identity between the insertion and the chromosomal moieties. All sequences are shown oriented as in the Watson strand of chromosome. :::  ::: Sequence divergence between B regions {#s2d} ------------------------------------- We examined the 24 available *S. cerevisiae* genome sequences found to contain region B in more detail ([Table S1](#pone.0017872.s002){ref-type="supplementary-material"}). Most of these strains belonged to the wine or European group or had been characterized as strains with mosaic genomes [@pone.0017872-Liti2], [@pone.0017872-Schacherer1]. From these genomes, we obtained 10 sequences with sufficient coverage to infer the phylogeny of region B ([Figure S1](#pone.0017872.s001){ref-type="supplementary-material"}). The resulting dendrogram ([Figure 6A](#pone-0017872-g006){ref-type="fig"}) draws a clear picture of the evolution of this region: the *Z. bailii* sequence is in a basal position and presents the longest branch (mean nucleotide divergence of 1.54 substitutions per kb from the group of *S. cerevisiae* strains). By contrast, the *S. cerevisiae* sequences display a low level of diversity, with a mean estimated nucleotide divergence of 0.3 substitutions per kb ([Figure 6A](#pone-0017872-g006){ref-type="fig"}), much lower than the diversity between *S. cerevisiae* strains, previously estimated to 1.0--1.4 substitution per kb for wine yeasts and up to 7.3 substitutions per kb for the most distantly related *S. cerevisiae* strains ([@pone.0017872-Liti2], [@pone.0017872-Fay1] and [Figure 6B](#pone-0017872-g006){ref-type="fig"}). This estimate suggests that the divergence of the region B is at least three time more recent that wine yeast expansion, which was shown to start very likely during or after the Neolithic era [@pone.0017872-Legras1]. ::: {#pone-0017872-g006 .fig} 10.1371/journal.pone.0017872.g006 Figure 6 ::: {.caption} ###### Evolutionary relationships of region B in *Saccharomyces cerevisiae*. A. Neighbor-joining tree based on single nucleotide polymorphism of the region B sequences obtained from *S. cerevisiae* and *Z. bailii*. The labels refer to *S. cerevisiae* strains, followed by the chromosome on which region B was found, if available. B. Neighbor-joining tree based on single nucleotide polymorphism of the genome sequences of 44 *S. cerevisiae* strains. Strains are shown in color according to their technological or geographical origin: clinical isolates in gray, European wine isolates in green, bread isolates in orange, American bioethanol production isolates in purple, European soil isolates in khaki, American isolates in blue-green, Asian isolates in dark blue and African isolates in brown. :::  ::: The bioethanol and brewery strains (CLIB324, CBS7960, YJM280 and JAY291) share specific mutations in this region ([Figure S1](#pone.0017872.s001){ref-type="supplementary-material"}), and have region B inserted at the same chromosomal site, with the exception of JAY291 ([Figure 3](#pone-0017872-g003){ref-type="fig"}). Similarly, the B region of strain M22 has a nucleotide sequence identical to that inserted into chromosome XII of EC1118 (EC1118\_XII), but the two forms differ in terms of their insertion site. These data suggest that this region can actually move from one locus to another one within the genome. Distribution of B regions in *S. cerevisiae* strains of different origins {#s2e} ------------------------------------------------------------------------- In our previous study [@pone.0017872-Novo1], we carried out PCR analysis to determine the distribution of region B among 53 *S. cerevisiae* strains of different origins. Region B was found in 25 strains, 20 of which were isolated from the wine environment. Here, we studied the copy number and location of region B in these 25 strains, together with three additional wine yeast strains or derivatives (59A, V5 and N96), through a combination of PFGE, Southern blotting and PCR amplification ([Table S1](#pone.0017872.s002){ref-type="supplementary-material"}). We obtained evidence for the presence of eight different B regions in these 28 *S. cerevisiae* strains, with up to four different copies present in a single strain, L1414 ([Figure 7](#pone-0017872-g007){ref-type="fig"}). All strains closely related to EC1118 (line 59A to 3238-32), with the exception of 6bpenciu, Eg25 and T73, had insertion sites similar to those of EC1118. For two strains (L-1374 and AWRI796) we detected PCR amplifications consistent with the insertions characterized for chromosomes XI and XIII respectively ([Figure 3](#pone-0017872-g003){ref-type="fig"} and [Figure 7](#pone-0017872-g007){ref-type="fig"}). However, the successful PCR amplification for the complete set of inter- and intragenic fragments was also consistent with an additional region B. As PFGE and Southern blotting analysis revealed only one chromosomal band, these findings strongly suggest than two copies are present on the same chromosome, separated or in tandem. ::: {#pone-0017872-g007 .fig} 10.1371/journal.pone.0017872.g007 Figure 7 ::: {.caption} ###### Distribution and localization of region B among yeast strains. PCR and southern blot hybridization were performed on genomic DNA from 28 *S. cerevisiae* strains of different origins, with probes designed to bind to region B, as described in [Materials and Methods](#s4){ref-type="sec"}. The presence of the first five regions was determined by PCR amplification, with primers specifically designed to discriminate between the different forms, and by Southern blot hybridization. The presence of the three last forms was determined by Southern blot hybridization. The origin of the strains is indicated by the color of the name: green for wine, purple for bread, brown for soil and gray for clinical. The distribution of the various forms of region B is represented by colored rectangles: blue for EC1118\_XIV, green for EC1118\_XII, purple for EC1118\_X, dark blue for RM11\_XIV, pink for JAY291\_XI, blue-green for XIII, brown for VII/XV and gold for IV. :::  ::: Discussion {#s3} ========== We recently described the striking presence of large introgressions from distantly related yeasts in the genome of wine yeasts [@pone.0017872-Novo1]. We report here that one of these introgressions, the *Z. bailii*--derived 17 kb chromosomal segment, is present in multiple copies in the genome of wine yeasts, mostly at internal positions on various chromosomes ([Figure 5](#pone-0017872-g005){ref-type="fig"}). We propose that the amplification and expansion of this fragment in wine yeast has involved the formation of a circle molecule subsequently integrated into the *S. cerevisiae* genome through nonhomologous recombination. The commercial wine strain EC1118 carries three copies of this region, on three different chromosomes. Similarly, most strains isolated from vineyards or commercial wine yeast strains containing this region carried several copies of it, with up to four copies detected in a given strain. It has been shown that yeast genes transferred from bacteria tend to undergo segmental duplication in their new host [@pone.0017872-Rolland1]. Similarly, the 14-gene chromosomal fragment acquired by intervarietal transfer in the genome of *C. neoformans* is duplicated [@pone.0017872-Kavanaugh1]. The duplication of large DNA segments has occurred repeatedly throughout evolution (see [@pone.0017872-Koszul1]). Intra- and interchromosomal duplications are often mediated by Ty elements, but also occur in the absence of repeated elements, as a result of microhomology/microsatellite-induced replication (MMiR). Other mechanisms involving extrachromosomal amplification [@pone.0017872-Koszul1], [@pone.0017872-Dujon2], [@pone.0017872-Gresham1], [@pone.0017872-Hughes1], [@pone.0017872-Libuda1], [@pone.0017872-Moore2] have been reported in natural or experimental yeast populations. For some known cases of extrachromosomal *S. cerevisiae* DNA amplification [@pone.0017872-Gresham1], [@pone.0017872-Libuda1], the presence of a centromere and origins of replication has been reported in amplified fragments. As no genome sequence for *Z. bailii* is currently available, we were unable to determine directly whether replication elements were present. However, within region B, we identified two sequences corresponding to the *S. cerevisiae* ARS consensus sequences (ACS). *S. cerevisiae* ARS elements consist of two essential functional domains: domain A, containing an 11 bp conserved sequence (ACS), and a broad A+T-rich domain B, which flanks domain A but displays no sequence similarity [@pone.0017872-Campbell1]. Although replication origins are not well conserved among eukaryotes, ARS found on plasmids from *Z. rouxii*, *Z. bisporus* and *Z. bailii* were shown to be effective for autonomous replication in *S. cerevisiae* [@pone.0017872-Araki1], [@pone.0017872-Tohe1], [@pone.0017872-Tohe2], [@pone.0017872-Utatsu1]. Therefore, both the structural organization of the integrated regions with a circular permutation of the genes and the presence of two ACS in region B suggest that the circular molecule exists and replicates autonomously in *S. cerevisiae*. This hypothesis was supported by a direct experimental evidence that at least one ARS element is functional in *S. cerevisiae*. The mechanism leading to multicopy integration of region B is intriguing. Multiple integrations may have required maintenance of the eccDNA molecule through each sequential integration, before being lost. The finding that an ARS in region B supports autonomous replication in *S. cerevisiae* suggests that the circular form may have been stably maintained for quite a while, resulting in integrations at different locations in the various strains. In addition, the eccDNA molecule may have in some instances integrated in tandem array, as suggested by the detection of two integrations in the same chromosome in two strains. This duplicated molecule might then regenerate an eccDNA molecule by homologous recombination, allowing further integrations ([Figure 8](#pone-0017872-g008){ref-type="fig"}). ::: {#pone-0017872-g008 .fig} 10.1371/journal.pone.0017872.g008 Figure 8 ::: {.caption} ###### Model of multicopy integration of region B. We propose that the circular form, which can replicates in *S. cerevisiae*, is able to integrate sequentially at different chromosomal locations by non-homologous recombination. In some instance, integration may also occur in tandem and afterwards regenerate an eccDNA molecule by homologous recombination, allowing further integrations ([Figure 8](#pone-0017872-g008){ref-type="fig"}). :::  ::: Most eukaryotic cells have two DNA double-strand break (DSB) repair pathways: homologous recombination (HR) and nonhomologous end joining (NHEJ). The second of these pathways is rare in *S. cerevisiae*, involves the direct rejoining of two DNA molecules and is closely associated with illegitimate recombination and chromosomal rearrangement [@pone.0017872-Moore1], [@pone.0017872-Daley1]. The extrachromosomal amplifications described to date in *S. cerevisiae* involve repetitive DNA, as shown for Ty or LTR [@pone.0017872-Gresham1], [@pone.0017872-Libuda1], rDNA [@pone.0017872-Hourcade1] and telomeric loci [@pone.0017872-Horowitz1]. An analysis of the sequence at breakpoint junctions identified no repeated elements in the immediate vicinity of the insertion points, but showed that integration might have involved microhomology (2 to 3 bases) for four of the six insertions, possibly in association with base mutations in some cases. In NHEJ, the ends of the DNA are joined with little or no base pairing at the junction, and the end-joining product may include small insertions or deletions [@pone.0017872-Paques1], [@pone.0017872-Wood1]. Both these features are consistent with our analysis of the sequences at insertion junctions. An alternative mechanism to NHEJ is microhomology-mediated end joining (MMEJ) [@pone.0017872-McVey1]. However, since the MMEJ mechanism was found to require at least five homologous nucleotides and always leads to deletion or insertions of intervening sequence between the microhomologies [@pone.0017872-Daley1], [@pone.0017872-McVey1], [@pone.0017872-Ma1], the integration of region B by this mechanism seems unlikely. Random fragments of mitochondrial DNA (NUMTs) can be captured by the nuclear DNA to repair DSB in yeasts [@pone.0017872-Ricchetti1], [@pone.0017872-Sacerdot1], [@pone.0017872-Yu1]. The integration of short fragments of plasmids known as NUPAV has also recently been observed in hemiascomycete yeasts harboring plasmids [@pone.0017872-Frank1]. It has been suggested that NUMT and NUPAV are formed by occasional aberrant DSB repair events in yeast nuclear DNA [@pone.0017872-Ricchetti1], [@pone.0017872-Sacerdot1], [@pone.0017872-Frank1]. The integration of region B displays certain similarities with that of NUMT and NUPAV: (i) seven of the eight insertion sites observed for region B are intergenic, just as mitochondrial sequences are more frequent in noncoding areas than in coding regions [@pone.0017872-Ricchetti1]; (ii) no repeated elements were found in the immediate vicinity of B regions, as for 56% of NUMTs [@pone.0017872-Ricchetti1]. However, one major difference is that no integration of complete molecules has been observed for mitochondrial DNA or for various plasmids. The mechanisms of NUMT formation are unknown, but they result in the presence of random, short (tens to hundreds of nt) mtDNA insertions in the yeast chromosome. By contrast, region B is 17 kb long and is always found intact. Using a combination of phylogeny and syntheny analyses, we previously showed that region B was acquired by HGT from *Z. bailii*, a wine contaminant, to *S. cerevisiae*, this process probably being facilitated by the proximity of these species in the same ecological niche [@pone.0017872-Novo1]. This study confirms the transfer from *Z. bailii* and shows that this event occurred after wine strains had begun to diverge, also accounting for this region not being present in all wine strains today. In addition, our data demonstrate, for the first time, the spontaneous amplification of region B in natural wine yeast populations, potentially accounting for its diffusion in wine yeasts and related *S. cerevisiae*. EccDNA have been described in most eukaryotes [@pone.0017872-Kuttler1], reflecting the plasticity of the genome. High levels of eccDNA molecules are associated with cell stress or aging, and their formation mostly involves repeated elements, although nonhomologous recombination was reported [@pone.0017872-Kuttler1], [@pone.0017872-Cohen1], [@pone.0017872-vanLoon1]. It is tempting to speculate that amplification of genes resulting from HGT has helped fermentative *S. cerevisiae* strains to adapt to a new evolutionary niche by providing new or evolved metabolic functions, although the role of the genes carried by region B, encoding a putative oxoprolinase, nicotinamide transporter, Flo11p and transcription factors [@pone.0017872-Novo1], remains to be determined. We recently identified the function of foreign genes acquired by gene transfer in the genome of EC1118 and of other wine yeast strains [@pone.0017872-Novo1]: one encodes a high-affinity fructose symporter, providing a new function in *S. cerevisiae* that might confer an adaptive advantage during the fermentation of grape must [@pone.0017872-Galeote1] and two other ones encode oligopeptide transporters [@pone.0017872-Damon1], which may help yeast cells to assimilate nitrogen at the end of fermentation or after the main fermentation process has been completed. Similarly, as the foreign genes carried on region B were taken up, maintained and expanded in the genome of wine yeast strains, we can infer that they must contribute in some way to increasing the evolutionary fitness of wine yeast. Materials and Methods {#s4} ===================== Strains {#s4a} ------- EC1118 (Lalvin EC1118) is a diploid heterozygous commercial wine yeast strain isolated in Champagne (France) and produced and sold commercially by Lallemand Inc. (Canada). Strain 59A was generated from a meiotic haploid spore isolated from EC1118 and selected on the basis of its similar fermentation performance and metabolite production. References for the other yeast isolates are detailed in Table SI. Cells were grown in YPD medium (1% yeast extract, 1% peptone, 1% glucose) at 28°C, with shaking. Estimation of the copy number of region B in the EC1118 genome {#s4b} -------------------------------------------------------------- We used EC1118 sequencing reads to identify changes in copy number along a linear genomic coordinate axis. BLAST [@pone.0017872-Altschul1] was used to align the read sequences to the scaffold EC1118\_1N26 (accession no. FN393084). For the sake of clarity, the sequence of the delta Ty2 LTR between positions 21,581 to 21,914 was masked. Each nucleotide position was covered by a mean of 18 reads in the diploid chromosomal region (two copies). Region B was found to be present in three copies, with a mean of 27 reads per nucleotide position. Genome sequencing and data analysis {#s4c} ----------------------------------- The genome sequence of the wine yeast EC1118 is a "pseudohaploid" assembly of 31 supercontigs [@pone.0017872-Novo1]. The genome sequence of strain 59A was determined with Illumina Genome Analyzer II technology with 36 bp paired reads (44 X sequencing depth). Velvet software [@pone.0017872-Zerbino1] version 0.6.05 was used for *de novo* assembly. The best assembly (*i.e* minimum number of contigs with maximum contig size) was obtained with a hash value of 23, resulting in 2,885 contigs with an N50 size of 11,807 bp. BLAST similarity searches were used to identify contig sequences covering region B of EC1118. Region B was found to encompass two contigs separated by a small gap of 35 bp. This gap was filled by the corresponding sequence from EC1118. The nucleotide sequence of region B from strain 59A was deposited in EMBL-GenBank under accession number HQ615872. Southern blot analysis {#s4d} ---------------------- Southern blot hybridization was performed on yeast chromosomes separated by pulsed-field gel electrophoresis (PFGE), as previously described [@pone.0017872-Bidenne1]. Probes were obtained by PCR amplification from EC1118 genomic DNA, using specific primers corresponding to a DNA fragment (gene 0023g) from region B (available upon request). Probes were labeled with the PCR DIG labeling system (Roche Diagnostics), according to the manufacturer\'s instructions. Chemiluminescence was detected with the CSPD alkaline phosphatase substrate and the DIG Luminescent Detection Kit (Roche Diagnostics). Search for region B in *S. cerevisiae* genomes and evolutionary relationships of region B in *Saccharomyces cerevisiae* {#s4e} ----------------------------------------------------------------------------------------------------------------------- We searched for similarity to region B in other *S. cerevisiae* genomes with blastn (no filter). Genome sequences were obtained for 35 strains of the SGRP project of the Sanger Institute [@pone.0017872-Liti2], 26 strains of the sequencing project at Washington University at St Louis (Justin Fay, <http://www.genetics.wustl.edu/jflab/data.html>), a bioethanol production yeast derivative JAY291 [@pone.0017872-Argueso1] and a vineyard isolate derivative RM11-1a (*S. cerevisiae* RM11-1a sequencing project, Broad Institute of Harvard and MIT <http://www.broad.mit.edu/>). We also included in our analysis the sequence of region B from *Z. bailii* [@pone.0017872-Novo1]. When a significant hit was obtained (expected value \<10^−10^, minimum identity 97%), the corresponding sequence was retrieved from contig and low quality regions were clipped if necessary. All sequences were aligned, with Genious software ver. 4.8.4 (Biomatters Ltd, New Zealand) and MUSCLE software ver. 3.8.31 [@pone.0017872-Edgar1]. Evolutionary history was inferred by the neighbor-joining method [@pone.0017872-Saitou1]. Evolutionary distances were calculated by the Tajima-Nei method [@pone.0017872-Tajima1] and are expressed in the units of the number of base substitutions per site. All positions containing alignment gaps and missing data were eliminated only in pairwise sequence comparisons (pairwise deletion option). Dendrograms were generated and phylogenetic analyses were conducted with MEGA4.1 software [@pone.0017872-Tamura1]. Genome sequence alignments {#s4f} -------------------------- Forty four yeast genomes were aligned with MUMmer 3.0 [@pone.0017872-Kurtz1], including 35 yeast genomes of the SGRP project [@pone.0017872-Liti2], JAY291 [@pone.0017872-Argueso1], M22 and YPS163 [@pone.0017872-Doniger1], RM11-1a, YJM789 [@pone.0017872-Wei1], S288C, AWRI1631 [@pone.0017872-Borneman1], EC1118 [@pone.0017872-Novo1] and 59A. Repetitive and low-complexity regions that could not be aligned unambiguously were first screened and masked with RepeatMasker (Smit *et al.,* <http://www.repeatmasker.org>). Polymorphic positions were then extracted, using dedicated Perl scripts to parse the MUMmer output files, with counting of the number of SNPs and indels between the aligned genomes. Experimental validation of the junctions of different B regions {#s4g} --------------------------------------------------------------- Direct experimental support was provided by PCR amplification, with EC1118 or 59A DNA as the template. DNA was isolated as described by Hoffman *et al.* [@pone.0017872-Hoffman1]. PCR primers (available upon request) were designed for the specific amplification of region B insertion junctions and conventional chromosomes, similar to those of S288C. The conventional forms were amplified with primers complementary to chromosomal sequences adjacent to the integration site. The region B insertion junctions were amplified with forward and reverse primers complementary to chromosomal sequences adjacent to the integration site and the region B sequence, respectively. Test of the autonomous replication function of an ARS element found in region B {#s4h} ------------------------------------------------------------------------------- The ARS consensus sequence at position 13763 to 13774 (ACS2) was inserted in YIp352 [@pone.0017872-Hill1], an integrative vector which contains the *S. cerevisiae URA3* gene for selection, at the *Bam*HI site. This ARS amplified fragment was obtained from DNA of 59A strain using the primer pairs: [GGATCC]{.underline} ACAGGTTCGAGTAGTTGAT and [GGATCC]{.underline} TAGTTCAAGAGGACATGA, corresponding to positions 13596 to 13991 of region B. The *Bam*HI sites were underlined. The yeast strain CEN.PK2-1C (*MATa*; *ura3-52*; *trp1-289*; *leu2-3,112*; *his3Δ 1*; *MAL2-8^C^*; *SUC2*) (EUROSCARF) was transformed by the LiAC procedure [@pone.0017872-Guldener1]with YIp352, YIp352-ARS and YEp352 [@pone.0017872-Hill1] as control. Ura^+^ transformants were obtained at a frequency of 4.10^4^ and 1.10^4^ transformants/µg for YIp352-ARS and YEp352 respectively. Extrachromosomal plasmids were recovered from Ura^+^ transformants by transforming *E. coli* to ampicillin resistance. Plasmids were prepared and the ARS region was subsequently sequenced. Distribution of B regions in the *S. cerevisiae* population {#s4i} ----------------------------------------------------------- We tested for the presence of region B variants in various *S. cerevisiae* strains (for a complete list of the strains used, see [Table S1](#pone.0017872.s002){ref-type="supplementary-material"}), by PCR amplification with primers specifically designed to discriminate between the different forms or by Southern blot hybridization (see above). Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### **Polymorphisms found in region B of** ***Saccharomyces cerevisiae***. The sequence of region B from strain 59A was used to query the available yeast genome sequences. For each strain, the matching sequences (black segments) --- often found on different contigs --- were used to identify SNP positions (red dots) and indels (green dots). In most cases, the point of integration of the region into the yeast genome was determined (blue bar). (EPS) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S1 ::: {.caption} ###### **Strains and genomes sequences used in the study.** (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: We thank Brigitte Cambon for technical assistance and Justin Fay for allowing us to use genome sequences and for providing strains from the yeast sequencing project at Washington University (St Louis). We thank Gianni Liti and Marilena Budroni for providing yeast strains. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was supported by the French Agence Nationale de la Recherche (ANR Project GENYEASTRAIT, ANR-07-BLAN-0205) and post doc fellowships of French National Institute for Agricultural Research (INRA) to EB and of the Generalitat de Catalunya to MN, <http://www.international.inra.fr/>. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: SD VG MN FB. Performed the experiments: VG EB FB MN. Analyzed the data: SD VG FB MN EB JLL SC. Wrote the paper: SD VG FB JLL SC.

PubMed Central

2024-06-05T04:04:19.855677

2011-3-10

PMC3053390

Introduction {#s1} ============ Speciation is one of the central processes in evolution. Geographic isolation of previously interbreeding populations and their subsequent evolutionary divergence appear as one of the mechanisms that influence the emergence of reproductive barriers, and then could promote the creation of new species. In this type of speciation, called geographic speciation or allopatric speciation, reproductive isolation may be achieved by prezygotic or postzygotic barriers (reviewed in [@pone.0017726-Coyne1]). The mechanism that allows the establishment of intrinsic postzygotic reproductive barriers was explained by the Bateson, Dobzhansky and Muller (BDM) model of genetic incompatibilities [@pone.0017726-Muller1], [@pone.0017726-Bateson1], [@pone.0017726-Dobzhansky1], as an accumulation of genetic substitutions through the divergent evolution of epistatic genes. These substitutions, which could be either adaptive or neutral in the same population, may be deleterious once confronted in the hybrids. The BDM model was recently supported by the identification of incompatibilities between genes acting as postzygotic barriers [@pone.0017726-Chen1], [@pone.0017726-Lee1], [@pone.0017726-Long1], [@pone.0017726-Brideau1], [@pone.0017726-Masly1], [@pone.0017726-Bikard1], [@pone.0017726-Yamagata1]. In strictly geographic speciation, the complete absence of gene flow between populations suggests a predictable uniform pattern of divergence across genomic regions. By contrast sympatric speciation, that allows a certain level of continuous gene flow between populations, leads to a mosaic genome structure with disparate sequence divergence, as demonstrated in whole genome approaches between closely related animal species [@pone.0017726-Turner1], [@pone.0017726-Kulathinal1]. The regions of high divergence are more likely to be associated with the presence of loci that maintain the genetic isolation of species or populations, since they are influenced by the strong diversifying pressure exerted over these loci [@pone.0017726-Wu1]. Cultivated rice belongs to two distinct species: *Oryza sativa* L. originated from Asia but now cultivated worldwide, and *Oryza glaberrima* Steud originated and restricted to West Africa. Despite the remarkable morphological and agricultural trait differences of Asian and African cultivated rice species [@pone.0017726-Sarla1], their wild relatives diverged recently from a common ancestor, approximately 0.6 to 0.7 million years ago [@pone.0017726-Zhu1], [@pone.0017726-Ma1], [@pone.0017726-Ammiraju1]. An Asian origin and ancestral animal dispersal to Africa of *Oryza* species were proposed to explain the biogeographic pattern of African rice [@pone.0017726-Second1], [@pone.0017726-Vaughan1]. Then, the posterior domestication processes of *O. sativa* in Asia and *O. glaberrima* in Africa took place independently. Despite a complex history, it appears that *japonica* and *indica* subspecies of *O. sativa* were domesticated from each other from pre-differentiated populations of *O. rufipogon* in Asia, approximately 7,000 years ago [@pone.0017726-Sweeney1], [@pone.0017726-Fuller1], *O. glaberrima* was domesticated from its wild relative *O. barthii*, in the Niger River delta in Mali approximately 3,000 years ago [@pone.0017726-Portres1], [@pone.0017726-Murray1]. *O. sativa* and *O. glaberrima* are reproductively isolated, limiting significantly the use of the genetic potential of *O. glaberrima* for the improvement of Asian rice. Therefore, the identification and characterization of the genetic factors that affect fertility in the interspecific hybrids will allow an easier use of *O. glaberrima* in rice breeding programs, and a better understanding of the nature of postzygotic barriers. Reproductive isolation between *O. sativa* and *O. glaberrima* is mediated by a strong postzygotic barrier, which results from the action of several loci over the fertility of the F~1~ hybrids [@pone.0017726-Sano1], [@pone.0017726-Doi1]. Among them, the *S~1~* locus plays a central role. By regular and innovative mapping approaches, the *S~1~* locus was recently fine mapped [@pone.0017726-Garavito1], [@pone.0017726-Koide1]. Additionally, the existence of two linked epistatic loci was inferred from genetic data, and a model based on a BDM incompatibility between the three adjacent loci (*S~1~A*, *S~1~* and *S~1~B*) was proposed to explain the allelic elimination of female gametes [@pone.0017726-Garavito1]. Finally, a gene coding for a putative F-box protein, and a Pack-Mule carrying a segment of an AP2 homolog were inferred as the most likely candidate factors for *S~1~*. Here we study patterns of divergence and evolution in the *S~1~A*, *S~1~* and *S~1~B* loci (called here *S~1~* regions), by genomic comparative approaches between orthologous sequences in *O. glaberrima* cv. CG14 and *O. sativa* sp. *japonica* cv. Nipponbare. Our objectives were (1) to establish the genomic sequence of the *S~1~* regions in *O. glaberrima*, (2) to identify patterns of divergence and evolution in the *S~1~* regions, and (3) to determine whether these patterns are informative to better understand the origin and the mechanism of the *S~1~* postzygotic reproductive barrier. Our results suggest that the *S~1~* regions of *O. sativa* and *O. glaberrima* have undergone no drastic variation in its recent divergence and evolution, implying that the accumulation of small genic changes, following a Bateson-Dobzhansky-Muller (BDM) model, might be the major evolutionary force behind this reproductive barrier. In this context, genetic incompatibilities involving the duplicated F-box genes as putative candidates, and a possible strengthening step involving the chromosomal inversion might participate to the reproductive barrier between Asian and African rice species. Results {#s2} ======= Sequencing of the *S~1~*, *S~1~A* and *S~1~B* loci in *O. glaberrima* {#s2a} --------------------------------------------------------------------- In a previous work we described the fine genetic and physical mapping, sequencing, and comparative analysis of the *S~1~* locus [@pone.0017726-Garavito1]. Additionally we detected the presence of two other loci that interact epistatically with *S~1~* to cause the allelic elimination of female gametes produced by the F~1~ hybrids. In order to study these two additional loci, we sequenced the seven remaining clones from the *O. glaberrima* cv. CG14 BAC library [@pone.0017726-Kim1] that constitute the minimum tiling path (MTP) established around *S~1~* [@pone.0017726-Garavito1]. The eight sequenced clones (including the one from our previous work) were obtained with a coverage ranging between 11 and 14× and an error rate below 1 base per 100 kb. They account altogether for 1,102 kb of sequences which, once assembled, constitute an 813 kb contig, referred here as "the *S~1~* regions". The seven newly obtained sequences are available with the following EMBL accession numbers: FP340539 (OG-BBa0041E07); FP340540 (OG-BBa0056F23); FP340541 (OG-BBa0088O22); FP340542 (OG-BBa0045G15); FP340544 (OG-BBa0017A24); FP340545 (OG-BBa0066E18); and FP340546 (OG-BBa0093E08). Determining the bounds of the *S~1~A* and *S~1~B* loci {#s2b} ------------------------------------------------------ In order to find the bounds of the segment that bears the *S~1~A* and *S~1~B* loci, we designed a set of new polymorphic markers, based on the sequence comparison of the orthologous regions between Nipponbare and CG14. Marker evaluation was carried out in the plants that limited the *S~1~A* and *S~1~B* loci [@pone.0017726-Garavito1], in order to find the approximated recombination sites. Using this approach the *S~1~A* locus was reduced to 171 kb, while no reduction was obtained for *S~1~B* locus, remaining as a 654 kb sequence ([Figure 1](#pone-0017726-g001){ref-type="fig"}). The 813 kb sequence thus covers completely the *S~1~A* and *S~1~* loci and partially the *S~1~B* locus, since the 102 kb proximal segment from the *S~1~B* locus (markers C6\_27332 to RM3805) is not included in the physical map and the sequenced clones. This segment was not found available either in the *O. glaberrima* (cv. GC14) genome project, due to a gap in the obtained physical map (R.A. Wing and P.R. Marri, personal communication). Additionally, microsatellite markers positioned within the gap segment in Nipponbare show an incongruent pattern of amplification and non-amplification in our *O. glaberrima* accessions (Data not shown). These data suggest that the structure of the proximal region of the *S~1~B* segment might be significantly different between *O. sativa* and *O. glaberrima*. Nevertheless, the nature and exact location of these differences remain unknown. ::: {#pone-0017726-g001 .fig} 10.1371/journal.pone.0017726.g001 Figure 1 ::: {.caption} ###### Genetic and physical maps of *S~1~*, *S~1~A* and *S~1~B* loci. Comparison between the genetic and physical maps of the *S~1~* loci, showing the interval of high probability of presence of *S~1~A* and *S~1~B* loci, determined by identifying the recombination breakpoints. The *S~1~A* locus is completely represented in the *O. glaberrima* physical map, while approximately 110 kb are missing from *S~1~B*. A straight line and a gray solid bar represent the *O. sativa* and *O. glaberrima* chromosomes respectively. :::  ::: Sequence annotation and organization of the 813 kb of the CG14 *S~1~* regions {#s2c} ----------------------------------------------------------------------------- The sequences of the CG14 BAC clones were annotated in detail. Gene and transposable element (TE) annotations are indicated in [Figure 2](#pone-0017726-g002){ref-type="fig"}. In total, 143 non-TE related coding regions were annotated, which corresponds to a gene density of about one gene per 6 kb of genomic sequence ([Table S1](#pone.0017726.s007){ref-type="supplementary-material"}). Most of predicted genes were confirmed by identification of protein domains, BLASTX homologies in Swiss-Prot database or BLASTN homologies with nucleotide databases. On the 143 predicted genes, 96.3% showed strong BLASTN homology with *O. sativa* genomic and full-length cDNA sequences. Seven pseudogenes were identified by the presence of fragmented coding regions lacking start codons, or by the presence of stop codons in the frame of exons. Of the 143 predicted genes, 35 (25%) belong to eleven distinct duplicated gene families scattered along the 813 kb analyzed, coding for F-box (gene families I and VII), LRR proteins (II), putative homeobox (III), Early nodulin-like (IV), Pectate lyase (V), Transferase (VI and X), Esterase/Lipase (VIII), Cystein synthase (IX) and Methylase proteins (XI) ([Table 1](#pone-0017726-t001){ref-type="table"} and [Figure 2](#pone-0017726-g002){ref-type="fig"}). Copy numbers of duplicated genes vary from two to seven genes. Most of the duplicated gene families were organized in clusters of relatively adjacent duplicated genes, with the exception of the two genes from family I (F-box family) separated by 154 kb of genomic sequence. The majority of the duplicated gene families were found to be arrayed in tandem while only three families showed duplicated genes in opposite orientations (I, V and X, [Table 1](#pone-0017726-t001){ref-type="table"}). As illustrated by the *S~1~* regions, the rice genome appears to be shaped by a relatively high number of local gene duplications [@pone.0017726-Rizzon1]. ::: {#pone-0017726-g002 .fig} 10.1371/journal.pone.0017726.g002 Figure 2 ::: {.caption} ###### Physical map and annotation of the 813 kb region of the *S~1~*, *S~1~A* and *S~1~B* loci in *O. glaberrima* cv. CG14. Yellow, blue and red boxes represent genes, transposons and retrotransposons, respectively. TEs nested into others TEs or genes are raised above their insertion sites. Markers used in the genetic map are indicated in gray. A black arrow indicates a large sequence inversion, relative to *O. sativa* ssp. *japonica* cv. Nipponbare. Regions spanning the *S~1~*, *S~1~A* and *S~1~B* loci are indicated. Roman numerals indicate duplicated gene families listed in [Table 1](#pone-0017726-t001){ref-type="table"}. :::  ::: ::: {#pone-0017726-t001 .table-wrap} 10.1371/journal.pone.0017726.t001 Table 1 ::: {.caption} ###### List of identified gene families in the *O. glaberrima S~1~* regions. ::: {#pone-0017726-t001-1} Duplicated gene family Gene Name Putative *O. sativa* (Nipponbare) orthologous gene Putative Function Position in the CG14 contig (bp) Orientation ------------------------ -------------------------------------------------------- ---------------------------------------------------- ----------------------------------- ---------------------------------- ------------- **I** **OG-BBa0066E18.4** LOC\_Os06g04690 Putative F-box protein 156499--161044 − **OG-BBa0049I08.11** LOC\_Os06g04980 Putative F-box protein 315435--318859 \+ **II** **OG-BBa0017A24.2** LOC\_Os06g04830 Putative LRR protein 195957--198903 − **OG-BBa0017A24.3** LOC\_Os06g04840 Putative LRR protein 204593--207258 − **III** **OG-BBa0017A24.4** LOC\_Os06g04850 Putative protein 216110--217001 \+ **OG-BBa0017A24.6** LOC\_Os06g04870 Putative protein 229920--230963 \+ **IV** **OG-BBa0049I08.3** LOC\_Os06g04930 Putative ENOD93 protein 267102--267855 − **OG-BBa0049I08.5** LOC\_Os06g04940 Putative ENOD93 protein 273413--274012 − **OG-BBa0049I08.6** LOC\_Os06g04950 Putative ENOD93 protein 277419--277981 − **OG-BBa0049I08.12** LOC\_Os06g04990 Putative ENOD93 protein 320002--320651 − **OG-BBa0049I08.13** LOC\_Os06g05000 Putative ENOD93 protein 325881--326408 − **OG-BBa0049I08.14** LOC\_Os06g05010 Putative ENOD93 protein 328903--329477 − **OG-BBa0049I08.15** LOC\_Os06g05020 Putative ENOD93 protein 332839--333419 − **V** **OG-BBa0088O22.2** LOC\_Os06g05209 Putative Pectate lyase protein 456489--457642 − **OG-BBa0088O22.3** LOC\_Os06g05260 Putative Pectate lyase protein 461242--462685 − **OG-BBa0088O22.7** LOC\_Os06g05272 Putative Pectate lyase protein 503408--504867 \+ **VI** **OG-BBa0088O22.8** LOC\_Os06g05284 Putative Transferase 506199--507791 \+ **OG-BBa0088O22.9** LOC\_Os06g05300 Putative Transferase 509461--511398 \+ **OG-BBa0088O22.10** LOC\_Os06g05310 Putative Transferase 512912--515793 \+ **OG-BBa0088O22.11** LOC\_Os06g05320 Putative Transferase 521812--522540 \+ **VII** **OG-BBa0056F23.12** LOC\_Os06g05560 Putative protein 632773--635500 \+ **OG-BBa0056F23.13** LOC\_Os06g05580 Putative F-box protein 638676--639857 \+ **OG-BBa0056F23.14** [\*](#nt101){ref-type="table-fn"} LOC\_Os06g05590 Putative F-box protein 641606--642834 \+ **OG-BBa0056F23.15** LOC\_Os06g05600 Putative F-box protein 643777--644970 \+ **OG-BBa0056F23.16** LOC\_Os06g05610 Putative F-box protein 646377--647636 \+ **OG-BBa0056F23.17** [\*](#nt101){ref-type="table-fn"} LOC\_Os06g05620 Putative F-box protein 650971--652272 \+ **VIII** **OG-BBa0056F23.11** LOC\_Os06g05550 GDSL esterase/lipase protein 630402--632063 \+ **OG-BBa0056F23.18** LOC\_Os06g05630 GDSL esterase/lipase protein 655875--657747 \+ **IX** **OG-BBa0041E07.3** LOC\_Os06g05690 putative Cystein synthase protein 676063--678199 \+ **OG-BBa0041E07.4** LOC\_Os06g05700 putative Cystein synthase protein 681007--683334 \+ **X** **OG-BBa0041E07.9** LOC\_Os06g05750 Putative Transferase protein 695968--697392 \+ **OG-BBa0041E07.12** LOC\_Os06g05790 Putative Transferase protein 710049--711485 − **XI** **OG-BBa0041E07.21** LOC\_Os06g05900 Putative Methylase protein 759933--764709 \+ **OG-BBa0041E07.22** LOC\_Os06g05910 Putative Methyltransferases 766462--768723 \+ \*Pseudogene. ::: In total 380 known TE were identified representing 18.3% of the genomic sequence. Of the 380 elements, 37 were classified as class I retroelements ([Table S2](#pone.0017726.s008){ref-type="supplementary-material"}). Interestingly, ten annotated transposons overlaped with predicted coding regions. These transposons were classified as pack-MULE since they showed similarities with Mutator-like elements and contained embedded coding sequences [@pone.0017726-Jiang1]. Among enclosed coding regions, six were classified as pseudogene due to the presence of frame-shift mutations or the complete absence of start codons. Only three pack-MULE displayed similarities to proteins with known functions ([Table S3](#pone.0017726.s009){ref-type="supplementary-material"}). Sequence comparisons of orthologous *S~1~* regions {#s2d} -------------------------------------------------- The orthologous regions were identified in the *O. sativa* ssp. *japonica* cv. Nipponbare public sequence as a stretch of 847 kb between coordinates 1,900,806 and 2,748,457 on chromosome 6. The orthologous regions in *O. sativa* ssp. *indica* cv 93-11 sequence was also used for comparative analysis and consist of a stretch of 1,077 kb broken by 152 ambiguous segments (with 'N'), representing gaps, that sometimes did not allow accurate comparative analysis. Pairwise comparisons between CG14 and Nipponbare *S~1~* regions revealed the presence of stretches of highly conserved segments interrupted by a limited number of zones with significant alterations. Most of the sequence variations involved mechanisms of sequence insertions, deletions, duplications, and a large sequence inversion, as illustrated in [Figure 3](#pone-0017726-g003){ref-type="fig"}. Sequence variation was not limited to intergenic regions since they overlap with different segments that include coding genes. Detailed analysis of the collinearity was then performed between the non-TE genes located within the CG14 and Nipponbare orthologous regions. All CG14 predicted genes were used as queries to BLAST them against Nipponbare genes within the *S~1~* regions, to generate a matrix of distance between genes used to draw the relationships between orthologous and paralogous genes ([Figure S1](#pone.0017726.s001){ref-type="supplementary-material"}). Most of the genes (\>90%) were found conserved in the same order and orientation between the two orthologous sequence. Seventeen and 11 predicted genes, respectively in CG14 and Nipponbare, were found to be involved in mechanisms that disrupt the collinearity. First, eleven genes were present in CG14 and absent in Nipponbare, while only three extra genes were found in the Nipponbare segment. Of the eleven extra genes predicted in CG14, seven were enclosed into pack-MULE elements (66E18.45, 17A24.25, 49I08.4, 49I08.45, 49I08.75, 49I08.9 and 49I08.10, [Table S1](#pone.0017726.s007){ref-type="supplementary-material"} and [S3](#pone.0017726.s009){ref-type="supplementary-material"}). In Nipponbare, two of the three extra genes (Os06g04710 and Os06g05470) were classified as expressed proteins. This result suggests that a significant number of collinearity disruptions in the *S~1~* regions may be produced by gene movement mechanisms, such as the transposition of pack-MULE elements. These disruptions were mainly distributed into three large segments along the *S~1~* regions and one region outside the *S~1~* regions ([Figure 3](#pone-0017726-g003){ref-type="fig"}). Detailed comparisons between CG14 and the two sub-species of *O. sativa* (ssp. *japonica* cv. Nipponbare and ssp. *indica* cv. 93-11) were carried out in these sites, in order to investigate the molecular mechanisms involved. ::: {#pone-0017726-g003 .fig} 10.1371/journal.pone.0017726.g003 Figure 3 ::: {.caption} ###### Orthologous sequence comparisons between the 813 kb and 847 kb from the *O. glaberrima* cv. GC14 and *O. sativa* ssp. *japonica* cv. Nipponbare *S~1~* regions. Comparisons were performed using dot plot alignment of the CG14 sequence (horizontal axis) against the Nipponbare sequence (vertical axis; coordinates 1,900,806--2,750,619-bp on chromosome 6). The *S~1~A*, *S~1~* and *S~1~B* regions are indicated along the horizontal axis. Positions and orientation of genes are symbolized by black and colored boxes along X and Y-axes. Colored boxes represent genes that disrupted microcollinearity. Clear boxes in the dot blot underline four large regions showing a strong disruption in the microcollinearity. :::  ::: Around the 93E18.3 and 93E18.5 CG14 genes, located in the flanking region of the genetic interval of *S~1~A*, sequence comparisons showed a large insertion of 13,123 bp of sequence in CG14 relative to Nipponbare ([Figure 3](#pone-0017726-g003){ref-type="fig"}, box A). This extra segment carries a predicted pseudogene (93E18.3) and a gene coding for a putative protein (93E18.5), absent in *O. sativa* (Nipponbare and 93-11; data not shown). It was not possible to identify the mechanism that originated this insertion (or deletion in Nipponbare), since its extremities have no similarity with known TEs, and no traces of short duplications were clearly visible at the insertion site in CG14. Although the insertion (or deletion) of large segments containing genes appears to be common in rice compared to distant species such as *Brachypodium* [@pone.0017726-Bossolini1], such rearrangement hasn\'t been previously reported between closely related rice species. Comparisons between orthologous sequences around 66E18.3 (Putative protein) and 66E18.4 (Putative F-box protein) genes indicated that both CG14 and Nipponbare regions have undergone a multitude of small changes since the loci have diverged from a common ancestor ([Figure 3](#pone-0017726-g003){ref-type="fig"}, box B). Here collinearity is altered by a local gene order alteration involving both genes, compared to the positions of the orthologous genes from *O. sativa* (ssp. *japonica* and *indica*) ([Figure 4](#pone-0017726-g004){ref-type="fig"}). In addition to the order rearrangement, a several TE insertions (two MITEs and two retrotransposons in CG14, and a large helitron and a transposon in Nipponbare and 93-11) were detected. Furthermore, a block of approximately 13 kb comprising genes Os06g04690, Os06g04699 and one helitron was found duplicated in tandem orientation in Nipponbare but not in 93-11 ([Figure 4 B](#pone-0017726-g004){ref-type="fig"}). The mechanisms at the origin of the gene rearrangement between 66E18.3 and 66E18.4 remain unidentified, thus no evolutionary model could be developed. However it seems clear that numerous TEs have been inserted up- and downstream the orthologous genes after the divergence between *O. glaberrima* and *O. sativa*; and that after the divergence between *indica* and *japonica* subspecies, a large duplication occurred in Nipponbare relative to 93-11. Furthermore, comparisons between the F-box duplicated genes in Nipponbare (Os06g04690 and Os06g04710) reveal significant changes. Due to frame-shift mutations, the predicted Os06g04710 gene is shorter than the duplicated Os06g04690 gene, resulting in a predicted protein that lacks the N-terminal region of the F-box domain. Altogether, these events indicate that this region may represent an intense spot of recent divergence between *O. glaberrima* and *O. sativa*, but also within *O. sativa* subspecies. ::: {#pone-0017726-g004 .fig} 10.1371/journal.pone.0017726.g004 Figure 4 ::: {.caption} ###### Comparisons of the orthologous sequences around the 66E18.3/66E18.4 CG14 genes. **A**. Dot plot comparison between the genomic region of genes 66E18.3 and 66E18.4 from CG14 (horizontal axis; coordinates 145--180 kb) against the *O. sativa* ssp. japonica cv. Nipponbare orthologous sequence (vertical axis; coordinates 2,034,806--2,088,806-bp on chromosome 6). **B**. Schematic representation of the comparison between the genomic region of genes 66E18.3 and 66E18.4 and their orthologous genes in *O. sativa* ssp. *japonica* cv. Nipponbare and *O. sativa* ssp. *indica* cv. 93-11. Colored backgrounds link orthologous regions. Boxes symbolize the positions of transposons and helitrons elements (light blue), Retrotransposons (red), Putative protein gene 66E18.3 and its orthologs (green), Putative F-box protein gene 66E18.4 and its orthologs (purple), and other genes (yellow). :::  ::: At the 49I08.7/49I08.11 genes region ([Figure 3](#pone-0017726-g003){ref-type="fig"}, box C), localized within the *S~1~* locus, the collinearity was altered by the presence of a total of five extra genes in *O. glaberrima* compared to *O. sativa* [@pone.0017726-Garavito1]. Most of them (49I08.75; 49I08.9 and 49I08.10) appear to be enclosed within pack-MULE elements, suggesting that massive re-localization of these elements in the *O. glaberrima* region may be here the mechanism for collinearity perturbation ([Table S3](#pone.0017726.s009){ref-type="supplementary-material"}). Dot plot alignment of the CG14 and Nipponbare orthologous sequences around genes 88O22.2/88O22.7 evidenced a paracentric chromosomal inversion of approximately 45 kb ([Figure 3](#pone-0017726-g003){ref-type="fig"}, box D). This inversion involved four different coding genes (88O22.3, 88O22.4, 88O22.5 and 88O22.6) in CG14, perturbing gene orders and orientations. A detailed comparison between orthologous sequences was carried out in order to identify the chromosomal inversion breakpoints and to investigate the process responsible of such rearrangement. Close analysis indicated that the distal and proximal inversion breakpoints contain gene duplications in both species (respectively 88O22.2/88O22.7 genes and Os06g05209/Os06g05272 genes). These genes, coding for Pectate lyase proteins, belonged to a locally duplicated gene family composed of three gene copies (Family V, [Table 1](#pone-0017726-t001){ref-type="table"}). Duplicated genes at the edge of the inversion were nearly identical, with the exception of the first 36 extra-nucleotides at the 5′ end of 88O22.7 and Os06g05209 genes, resulting in twelve extra amino-acids for each gene (purple boxes and arrowheads, [Figure 5 A](#pone-0017726-g005){ref-type="fig"}). All genes located in the inversion breakpoints in CG14 and Nipponbare appear intact and seem putatively functional (even after switching their upstream segment), since their coding regions are identical. A tentative model for the chromosomal inversion process is depicted in [Figure 5 B](#pone-0017726-g005){ref-type="fig"}. In the ancestral fragment---here the structure of the fragment is identical to the Nipponbare one---, two homologous Pectate lyase genes in opposite orientations flanked an internal region of 45 kb (Blue and green boxes, [Figure 5 B](#pone-0017726-g005){ref-type="fig"}). A mechanism of homologous recombination between inverted Pectate lyase genes occurred, leading to an inversion of the internal region and the exchange of upstream regions of Pectate lyase genes. ::: {#pone-0017726-g005 .fig} 10.1371/journal.pone.0017726.g005 Figure 5 ::: {.caption} ###### Comparisons of the orthologous sequences around the 88O22.2/88O22.7 CG14 *g*enes. **A**. Dot plot comparison and structures between the genomic region of genes 88O22.3/88O22.7 from CG14 (horizontal axis; coordinates 445--506 kb) and their orthologous in Nipponbare (vertical axis; coordinates 2,336,995--2,376,853-bp on chromosome 6). **B**. A model for the generation of the chromosomal inversion between *O. glaberrima* and *O. sativa*. **a**. Hypothetical ancestral segment (identical to the organization in the Nipponbare genome); **b**. Breakpoints occur by homologous recombination between two homologous genes coding for Pectate lyase proteins. **c**. Segment in *O. glaberrima*. Boxes symbolize the positions of Pectate lyase duplicated genes (blue, green and orange), transposons elements (light blue), Retrotransposons (red), and other genes (yellow). Purple boxes and arrowheads indicate the positions of twelve extra amino acids between duplicated Pectate lyase genes. :::  ::: Unfortunately the reduced quality of the sequence assembly for 93-11 did not allow us to confirm the presence of the chromosomal inversion between *O. glaberrima* and *O. sativa ssp. indica*. Nevertheless, a mapping analysis based on BLAST alignments of available BAC end sequence (BES) pairs from seven different *Oryza* species [@pone.0017726-Ammiraju2] around and within the inversion breakpoints, clearly indicated that the inversion structure is identical between *O. glaberrima* and four other *Oryza* species (*O. nivara*, *O. officinalis*, *O. alta* and *O. australiensis*). On the contrary, the mapping of BESs from *O. sativa* (ssp. *japonica* cv Nipponbare) and *O. rufipogon*, the wild ancestor of *O. sativa*, suggest a different genomic structure compared to the *O. glaberrima* region ([Figure S2](#pone.0017726.s002){ref-type="supplementary-material"}). Since chromosomal inversions are known to suppress genetic recombination between normal and inverted chromosomal segments, we evaluated the recombination rates in our *O. sativa*×*O. glaberrima* backcross populations [@pone.0017726-Garavito1] around and within the inversion. The genetic map obtained from 779 BC~1~F~1~ plants, after a high marker saturation in the site of the structural variation, showed a complete absence of genetic recombination between the inversion breakpoints, in contrast to the recombination rates found in the rest of the 813 kb contig ([Figure S3](#pone.0017726.s003){ref-type="supplementary-material"}). These data suggest that the chromosomal inversion represents an inter-specific rearrangement between *O. glaberrima* and *O. sativa*, which strictly restricts recombination within its limits. To our knowledge, this is the first report of a chromosomal inversion initiated by duplicated genes in plants. The considerable number of locally duplicated genes in rice may offer potential recombination targets for chromosomal rearrangements mechanisms involving coding regions [@pone.0017726-Rizzon1]. Transposable elements participated to the dynamic evolution of *S~1~* regions {#s2e} ----------------------------------------------------------------------------- Beside the alteration of the order and orientation of genes through re-localization of pack-MULE elements, comparative analysis between the Nipponbare and CG14 *S~1~* regions reveals changes of the genomic structure due to differential insertion or deletion of a multitude of transposable elements. Since the divergence of the two species, more than 117 kb of TE (13% of the segment) were inserted in Nipponbare, against 72 kb (9%) in CG14. The size difference observed at the *S~1~* regions, mainly due to the insertion of long full-length LTR retrotransposons, is in agreement with the genome size difference between *O. sativa* (434 Mb) and *O. glaberrima* (352 Mb) [@pone.0017726-Martinez1]. Most of the TEs appear randomly inserted along the *S~1~* regions, with the apparent exception of the TE accumulation that occurred in the *S~1~* locus ([Figure 3](#pone-0017726-g003){ref-type="fig"}). Here, the CG14 segment has undergone a 1.5× sequence size increase due to the local accumulation of TEs in the neighborhood of the *S~1~* candidate gene (49I08.11) [@pone.0017726-Garavito1]. Beside the *S~1~* locus, successive but isolated TE insertions responsible for the observed interruptions on the collinearity occurred specifically in Nipponbare, downstream the chromosomal inversion ([Figure 3](#pone-0017726-g003){ref-type="fig"}). Gene divergence in *S~1~* regions between *O. sativa* and *O. glaberrima* {#s2f} ------------------------------------------------------------------------- Of the 143 annotated genes in the sequenced regions, 120 were used for pairwise comparisons and divergence analysis with their respective Nipponbare orthologous genes, from which 109 fell into the *S~1~* regions. Twenty-three CG14 genes were not analyzed due to the absence of a Nipponbare orthologous gene, deep annotated gene structure differences between them, or because one of the two was annotated as a pseudogene. Similar analyses were carried out as a control, using two other published genomic regions in CG14 where no reproductive isolation region between the two species has been previously reported: (1) the *ADH* region on the short arm of chromosome 11, containing 13 annotated genes [@pone.0017726-Ammiraju1], and (2) the *MOC1* region on the long arm of chromosome 6, containing 17 annotated genes [@pone.0017726-Lu1]. The mean rate of non-synonymous substitutions (Ka) and synonymous substitutions (Ks) across the *S~1~* regions are respectively 0.009 and 0.035 ([Figure 6](#pone-0017726-g006){ref-type="fig"}). No statistical difference in the mean levels of Ka and Ks (respectively P = 0.771 and P = 0.317) was found when comparing with the *ADH* (Ka = 0.007, Ks = 0.040) and *MOC1* (Ka = 0.004, Ks = 0.020) regions ([Table S4](#pone.0017726.s010){ref-type="supplementary-material"} and [Figure S4](#pone.0017726.s004){ref-type="supplementary-material"}), suggesting that the *S~1~* as well as the *ADH* and *MOC1* regions are globally under identical evolution rates of protein-coding genes. Despite an overall uniformity of Ka and Ks values along the *S~1~* regions, several isolated peak values were higher than both background and mean values ([Figure 6](#pone-0017726-g006){ref-type="fig"}). Functions and Gene Ontology (GO) of annotated genes harboring increased Ks and/or Ka values were investigated. Of the 21 genes showing elevated Ks or Ka (at least two times the Ks or Ka mean values, indicated by symbols on [Figure 6](#pone-0017726-g006){ref-type="fig"}), 14 have known functions and are classified into the following categories of the "biological process" of Gene Ontology: response to biotic stimulus, protein modification, signal transduction and response to endogenous stimulus; and into the "molecular function" categories: kinase activity, nucleotide binding, protein binding, transferase activity, catalytic activity and hydrolase activity. The function and ontology of these genes suggest that high divergence may be a consequence of a local and accelerated evolution possibly driven by adaptation [@pone.0017726-Tang1] ([Figure 6](#pone-0017726-g006){ref-type="fig"}). The Ka/Ks ratio was also calculated to characterize the evolution of protein-coding sequences in the *S~1~* loci ([Table S5](#pone.0017726.s011){ref-type="supplementary-material"}). More than 85% of the genes were found to be under strong purifying selection, six genes to have a neutral evolution, while the 11 genes with the highest Ka/Ks values seemed to be evolving under positive selection or relaxed selective constraint. Finally, we investigated whether high gene divergence is globally associated to structural variations such as transposable element abundance, duplications and chromosomal inversions. No clear association was found between divergent genes and transposable element abundance as illustrated by the *S~1~* locus, where a clear accumulation of TEs was observed in CG14 compared to Nipponbare [@pone.0017726-Garavito1], with no significant effect on gene divergence. Similarly, within the chromosomal inversion, no peak of high divergence was associated to genes at the relative exception of the hypothetical gene 88O22.6. The high Ks values observed for the duplicated Pectate lyase genes might be generated by mosaic gene structures since these genes are located at the inversion breakpoints ([Figure 6 A](#pone-0017726-g006){ref-type="fig"}). ::: {#pone-0017726-g006 .fig} 10.1371/journal.pone.0017726.g006 Figure 6 ::: {.caption} ###### Representation of Ka and Ks values between 109 *O. sativa* and *O. glaberrima* orthologous genes in the *S~1~* regions. **A**. Representation of Ks values. The horizontal line is the mean value of Ks (Ks = 0.035) for the 109 analyzed orthologous genes in the *S~1~* regions. **B**. Representation of Ka values. The horizontal line is the mean value of Ka (Ka = 0.09) for the 109 analyzed orthologous genes in the *S~1~* regions. Lines below graphics represent analyzed genes located into *S~1~A*, *S~1~*, *S~1~B* loci, and those within the chromosomal inversion (i). Symbols represent different divergent genes with similar annotated function or unknown function as follows in *O. glaberrima*: (\*) 66E18.4 and 49I08.11 (F-box proteins), (") 66E18.6 (HAD phosphatase protein), (+) 66E18.8, 17A24.2 and 17A24.3 (LRR proteins), (°) 49I08.16 and 49I08.18 (Putative Serine/threonine protein kinases), (∧) 88O22.2 and 88O22.7 (Pectate lyase located at breakpoint inversion), (§) 41E07.23 (PRR protein) and (\<) 88O22.10, 88O22.11 and 41E07.12 (Putative transferase proteins), (\$) 88O22.6 and 88O22.19 (Hypothetical proteins) and (\#) 93E08.17, 66E18.3, 56F23.3, 56F23.12 and 41E07.25 (Putative proteins). :::  ::: Characterization of duplicated F-box genes in the *S~1~* regions {#s2g} ---------------------------------------------------------------- The gene 49I08.11 coding for a putative F-box protein has been proposed as a putative candidate gene for the *S~1~* locus, on the basis of the protein function of homologous genes and its high degree of divergence between the two species [@pone.0017726-Garavito1]. Sequencing of the 813 kb of the *S~1~* regions in *O. glaberrima* revealed a duplicated copy of the 49I08.11 F-box gene, located 154 kb apart in the *S~1~A* locus (gene 66E18.4) ([Figure 2](#pone-0017726-g002){ref-type="fig"} and [Table S1](#pone.0017726.s007){ref-type="supplementary-material"}). The duplicated genes exhibited high overall sequence similarities (90.7% of nucleotide identities). Both *O. glaberrima* F-box genes were found conserved in the orthologous *O. sativa* region. Altogether these data suggest that the duplicated F-box 66E18.4 gene may also be a valuable candidate gene for *S~1~A*. To study evolution of this F-box gene family, detailed gene comparisons were performed at the *S~1~A* and *S~1~* loci. In the *S~1~* locus, nucleotide and amino acid alignments showed significant sequence variations between orthologous F-box genes in *O. glaberrima* and *O. sativa* ([Figure S5](#pone.0017726.s005){ref-type="supplementary-material"}). The elevated Ka and Ks values and the calculated Ka/Ks ratio ([Table S5](#pone.0017726.s011){ref-type="supplementary-material"}) suggest an accelerated but neutral evolution, while the majority of genes in the *S~1~* regions appears to be under a purifying evolution. Besides coding region evolution, the CG14 49I08.11 F-box gene is embedded in an accumulation of TEs that reshaped its upstream and downstream regions. Furthermore its gene structure also evolved through the insertion of a non-autonomous transposon nested into the fourth intron of the gene [@pone.0017726-Garavito1]. In the *S~1~A* locus, a unique F-box gene is present in CG14 (66E18.4) and in *O. sativa* ssp. *indica* (BGIOSIBCE020566), compared to two tandemly duplicated orthologous genes in *O. sativa ssp. japonica* (Os06g04690 and Os06g04710). While the structure of the Os06g04690, BGIOSIBCE020566 and 66E18.4 genes appears to be the same, the 5′ part of the Os06g04710 gene exhibits several frame-shifts, resulting in a shorter predicted protein that lacks the N-terminal region of the F-box domain. Beside this variation, the alignments between the *S~1~A* F-box genes showed numerous polymorphisms at the amino-acid levels ([Figure S5](#pone.0017726.s005){ref-type="supplementary-material"}), which shaped the corresponding phylogenetic tree ([Figure S6](#pone.0017726.s006){ref-type="supplementary-material"}). Ka/Ks rates, calculated for these coding regions, suggest a similar evolution to the one observed for the *S~1~* F-box genes. Together these results suggest an accelerated evolution of these F-box genes that drives the divergence of the *O. sativa* and *O. glaberrima* orthologous genes, but also between the two *O. sativa* subspecies. These evidences allow considering these duplicated genes as potential candidates for the *S~1~A* and *S~1~* loci. Discussion {#s3} ========== The growing availability of whole genome sequences and comparative analysis of gene divergence have helped evolutionists to deduce the presence of reproductive barriers within highly divergent genomic regions [@pone.0017726-Wu1], and even to infer the possible path of speciation for several related species [@pone.0017726-Turner1], [@pone.0017726-Kulathinal1], [@pone.0017726-Osada1]. Whole genome sequence comparisons between *indica* and *japonica* subspecies of Asian rice have also led to the identification of large regions of high polymorphisms, whose origins have been associated with geographical differentiation, reproductive barriers, subsequent independent domestications, and a more recent admixture possibly mediated by human migration [@pone.0017726-Tang1]. However no direct comparison between experimentally validated postzygotic isolating loci has been performed so far at the sequence level, to directly investigate in detail the genomic evolution of such regions. Between the two cultivated rice species, the *S~1~* locus acts as the strongest postzygotic reproductive barrier, having an important role on their origin and conservation. In a previous work, we described the fine genetic and physical mapping of the *S~1~* locus. Additionally we detected the presence of two other loci (*S~1~A* and *S~1~B*) that interact epistatically with *S~1~* to cause the allelic elimination of female gametes produced by the F~1~ hybrids [@pone.0017726-Garavito1]. Based on available data, we build a genetic model where BDM incompatibilities between the alleles of the *O. sativa* and *O. glaberrima S~1~A*, *S~1~* and *S~1~B* loci are provoking the female gamete elimination and the strong transmission ratio distortion observed in the hybrids [@pone.0017726-Garavito1]. Our genetic model states that the final allelic frequencies and final survival rates of female gametes are associated to the recombination ratio between the three epistatic loci, their segregation during meiosis, and the alleles (*indica* or *japonica*) confronted in a given cross. In order to understand the basis of the evolution of the *S~1~* genomic regions, and to infer possible gene candidates or mechanisms behind this reproductive barrier, we sequenced the seven remaining *O. glaberrima* clones that constitute the physical map of the *S~1~* regions, and compared them with the orthologous regions in *O. sativa*. The comparisons revealed that the *S~1~* regions in both species are strongly conserved in terms of genomic structure and coding sequence divergence. Three isolated regions showing a disturbed collinearity were identified concerning: (1) local invasion of transposable elements (mainly Pack-MULEs carrying remnant of coding genes) around a putative F-box gene, candidate gene for the locus *S~1~*, (2) multiple duplication and subsequent divergence of the same F-box gene, within *S~1~A*, (3) and an interspecific chromosomal inversion in *S~1~B*. Additionally, we showed that most of the genes in the *S~1~* regions undergone a strong purifying selection, with the exception of few isolated divergent genes. These genes belong to functional categories known to confer adaptive advantages, and their highly divergent evolution could be a consequence of local adaptation to the African or Asian environments, or of human selection following the independent domestication processes. The pattern of evolution of a genomic region involved in a reproductive barrier could provide indications on its establishment, specifically, if it occurred under either an active or a restricted gene flow [@pone.0017726-Wu1]. The similar rate of gene divergence between the *S~1~* regions and two other genomic sites not involved in reproductive isolation may suggest a limited gene flow between populations during the establishment of the *S~1~* barrier. In consequence, the geographic localizations of *O. rufipogon* and *O. sativa* in Asia and of *O. barthii* and *O. glaberrima* in West Africa, together with a restricted gene flow could imply that this speciation process is the result of geographical isolation, in agreement with the current hypothesis of a common Asian origin and ancestral migrations to Africa [@pone.0017726-Second1], [@pone.0017726-Vaughan1]. However a precise estimation of gene flow rate is required to test this hypothesis. Under this highly conservative context, the *S~1~* barrier between *O. sativa* and *O. glaberrima* appears to have evolved from the divergent evolution of punctual genes and not from large genomic structural rearrangements, as predicted by the BDM model of incompatibilities. A detailed analysis of genes known to be implicated in BDM incompatibilities could help to identify possible candidates for the *S~1~* locus. Recently several molecular studies in animals and plants (including rice) revealed that gene duplication and divergence could be directly involved in postzygotic reproductive barriers concerning BDM incompatibilities in hybrids [@pone.0017726-Masly1], [@pone.0017726-Bikard1], [@pone.0017726-Yamagata1], [@pone.0017726-Mizuta1]. In our previous work, two putative candidate genes for the locus *S~1~* were identified: an F-box gene and a Pack-MULE transposon carrying a fragment of a AP2 gene [@pone.0017726-Garavito1]. Interestingly, a strongly conserved copy of the F-box gene from the *S~1~* locus is located in *S~1~A*, constituting the only gene family to be present at two different loci along the *S~1~* regions. The presence of these duplicated genes appears to match well the evolutionary model of an ancestral duplication followed by a divergent evolution of the alleles in each population. In terms of divergence, the F-box genes 66E18.4 and 49I08.11 in *O. glaberrima* cv. CG14 and their respective orthologous genes Os06g04690 and Os06g04710, and Os06g04980 in *O. sativa* cv. Nipponbare, exhibit a significant accelerated but neutral evolution ([Table S5](#pone.0017726.s011){ref-type="supplementary-material"}), in contrast to the purifying evolution of the majority of genes along the *S~1~* regions. Additionally, the up- and downstream regions of these F-box genes have undergone a multitude of structural variations since *O. sativa* and *O. glaberrima* diverged (including a second gene duplication and divergence in the *S~1~A* locus, in the *japonica* genome), suggesting that a dynamic evolution may be associated to them. In rice, the implication of F-box proteins in postzygotic barriers has already been reported for the *Sa* intersubspecific male sterility locus. In this case, the selective abortion of microspores is caused by the interaction between the *indica* and *japonica* alleles of a SUMO E3 ligase (*SaM*) and a F-box gene (*SaF*) [@pone.0017726-Long1]. This constitutes another argument for considering the hypothesis that the duplicated F-box genes are involved in the sterility barrier mediated by the *S~1~* locus. Even more, the second gene duplication and divergence in the *S~1~A* locus would allow to explain not only the observed differences in the TRD levels found between the *O. glaberrima*×*O. sativa* ssp. *indica* and the *O. glaberrima*×*O. sativa* ssp. *japonica* hybrids [@pone.0017726-Garavito1], but also the presence of the intersubspecific sterility locus *S~10~*, localized on the same genetic position [@pone.0017726-Zhu1], [@pone.0017726-Sano2]. F-box proteins constitute one of the largest multi-gene families with more than 700 putative genes and pseudogenes in rice [@pone.0017726-Jain1], [@pone.0017726-Xu1], [@pone.0017726-ThibaudNissen1]. F-box proteins and their SCF (Skp1-Cullin-F-box) complexes are known to be involved in regulatory functions on several processes, such as the progression throughout the meiotic [@pone.0017726-Wang1], [@pone.0017726-Pesin1] and mitotic divisions [@pone.0017726-Gusti1] during gametogenesis. In our genetic model for the female sterility caused by *S~1~*, only the cells that inherit a compatible allelic combination are able to pursue their development after each cellular division, to form a functional embryo sac [@pone.0017726-Garavito1]. Taking into account the recognized role of F-box proteins in the cell cycle progression, a BDM incompatibility after a hypothetical divergent subfunctionalization or neofunctionalization involving these genes could thus explain the arrested development of some allelic forms of hybrid gametes. Remarkably, the results from a previous F-box protein microarray analysis during rice panicle development evidenced the expression of genes Os06g04690, Os06g04710 and Os06g04980 (probe Os.3577.1.S1\_x\_at) in whole panicles throughout the meiotic and young microspore stages, and their down-regulation in mutants for the gene *Udh1*, an important transcription factor for meiocyte differentiation [@pone.0017726-Jain1]. These expression data demonstrate that these genes are expressed at the time and in the tissues where the BDM incompatibility is supposed to take place in the hybrids according to our genetic model. Taking into account the ability of F-box genes to closely interact with other proteins, their evolutionary plasticity, their known role in cell cycle progression and reproductive barriers, and their expression in reproductive tissues, the duplicated copies of the F-box gene appear as the best candidate factors for the *S~1~A* and *S~1~* loci. The comparative analysis of gene divergence have helped us to identify two genes possibly involved in the sterility barrier caused by the *S~1~A* and the *S~1~* loci, however no plausible candidate was determined for *S~1~B*, since the available sequence only partially spans the locus, and the structure of its proximal region seems to have a different configuration between the two species. However a striking alteration of the collinearity was observed within the *S~1~B* locus, in the form of a 45 kb chromosomal paracentric inversion between CG14 and Nipponbare. Mapping of BES pairs from seven *Oryza* species suggest a similar structure of the inversion region between *O. glaberrima* and both closely and distantly related species; while a different structure was found in *O. sativa* cv. Nipponbare and *O. rufipogon*. These data suggest that the inversion may have occurred recently in *O. rufipogon*, and has been inherited by *O. sativa* after domestication. In addition to the direct genomic sequence comparison between *O. sativa* and *O. glaberrima*, the genetic analysis showed a complete restriction of recombination between markers spanning the inversion in our interspecific BC~1~F~1~ populations. Besides reducing dramatically recombination between inverted and standard non-inverted chromosomes, inversions appear to play a major role in evolution of species [@pone.0017726-Hoffmann1], [@pone.0017726-Rieseberg1]. Between the close relatives sympatric species *Drosophila pseudoobscura* and *D. persimilis*, inversions were found within regions associated with hybrid sterility, suggesting that they might have contributed to their speciation process [@pone.0017726-Noor1]. Moreover, gene divergence was found higher within inverted regions than in non-inverted regions suggesting the occurrence of gene flow between the two species [@pone.0017726-Kulathinal1]. In contrast to the *Drosophila* example, the genic divergence outside and within the inversion in the rice *S~1~B* locus appears to be quite uniform ([Table S5](#pone.0017726.s011){ref-type="supplementary-material"}), suggesting that the inversion might have occurred after speciation or at least after the complete geographical isolation of the species. With the chromosomal inversion fixed in the *O. rufipogon*-*O. sativa* species group, its role in the reproductive isolation mechanisms would be limited to an increase of the genetic linkage between the loci involved in this sterility barrier. Since recombination between the three loci plays a key role in the final allelic frequencies and survival rates of female gametes produced by the hybrids [@pone.0017726-Garavito1], it is probable that the restriction of the recombination caused by the inversion would have a strengthening effect over the barrier [@pone.0017726-Navarro1]. The effect of the inversion on the recombination is not the only sign that the *S~1~* barrier could have been strengthened over time. *S~1~* has been described as a complex locus, having different effects over male and female fertility of the *O. sativa* and *O. glaberrima* hybrids. Plants that carry only the *S~1~* locus in a heterozygote state are partially male sterile [@pone.0017726-Sano1], [@pone.0017726-Koide1], while heterozygocity at the *S~1~A*, *S~1~* and *S~1~B* loci is necessary to observe partial female sterility [@pone.0017726-Garavito1]. This differential effect over male and female fertility could mean that the barrier has been strengthened over time by sequential accumulations of incompatibilities. Furthermore, the presence of an additional locus (*S~1~C*) in one of the four interspecific populations examined, which has a supplementary deleterious effect over female gamete elimination [@pone.0017726-Garavito1], seems to indicate that an auxiliary strengthening step may be currently under fixation. Conclusions {#s3a} ----------- In this work, we have studied the structural and genic divergence of the *S~1~* regions between *O. sativa* and *O. glaberrima*, as a method to understand the basis of their evolution and to infer possible gene candidates or mechanisms working behind this reproductive barrier. The comparisons showed that the *S~1~* regions have undergone no drastic variation in their recent divergence and evolution, suggesting that a small accumulation of genic changes, following a Bateson-Dobzhansky-Muller (BDM) model, might be involved in the establishment of the sterility barrier. In this context, genetic incompatibilities involving the duplicated F-box genes as putative candidates, and a possible strengthening step involving a chromosomal inversion that increases the genetic linkage between the factors involved in the epistatic interaction are suspected to participate in the reproductive barrier between Asian and African rice species. The knowledge generated by these comparative approaches contributes to a better understanding of the general evolution of postzygotic reproductive barriers in plants. Additionally, it allows considering new breeding strategies aiming unlocking the genetic potential of *O. glaberrima* for the improvement of the Asian rice. Additional efforts still remain necessary to confirm the candidate genes and to identify the molecular mechanism that controls the *S~1~* postzygotic barrier. Materials and Methods {#s4} ===================== Sequence analysis and gene annotation method {#s4a} -------------------------------------------- *O. glaberrima* cv. CG14 BAC sequencing was done by the Sanger method. Sequence analysis was done as previously described [@pone.0017726-Garavito1]. Briefly, coding regions were predicted *ab initio* using the FGENESH program [@pone.0017726-Salamov1] and then confirmed by comparative analysis with annotated genes models and proteins in *O. sativa* cv. Nipponbare, downloaded from the TIGR database [@pone.0017726-Ouyang1]. Predicted gene structures were manually evaluated by alignment with rice EST and full-length cDNA (FLcDNA) public sequences [@pone.0017726-Kikuchi1]. Detailed analysis was performed with the EMBOSS Analysis software [@pone.0017726-Rice1] and the physical map diagram was drawn using gff2ps software [@pone.0017726-Abril1]. Putative transposable elements (TEs) were first identified and annotated by RepeatMasker searches (<http://www.repeatmasker.org>) against local databases of rice TEs downloaded from the REPBASE [@pone.0017726-Jurka1], from the TIGR repeat database [@pone.0017726-Ouyang2], and RetrOryza [@pone.0017726-Chaparro1], and finally manually corrected. *De novo* prediction of TEs was performed according to structure of the different classes of TEs. The final annotation of the BAC sequences was performed using the Artemis tool [@pone.0017726-Rutherford1], and the comparison with the Nipponbare genome was accomplished using dot-plot alignments of the Dotter software [@pone.0017726-Sonnhammer1]. Nucleotide and amino-acid alignments were carried out using ClustalX [@pone.0017726-Thompson1]. Molecular marker analysis {#s4b} ------------------------- Genetic markers were designed from the comparison of the Nipponbare sequence with its orthologous CG14 sequence as previously described [@pone.0017726-Garavito1], and evaluated in four *O. sativa*×*O. glaberrima* BC~1~F~1~ populations developed from our previous work [@pone.0017726-Garavito1]. PCR reactions were carried as described [@pone.0017726-Orjuela1], with an annealing temperature and magnesium concentration optimized for each primer pair ([Table S6](#pone.0017726.s012){ref-type="supplementary-material"}). Separation of the PCR products was carried in 4% agarose and revealed with Ethidium Bromide for polymorphisms greater than 12 bp, and in a Li-Cor sequencer (Li-Cor Biosciences) for smaller polymorphisms, using a M13 tail tag (IRD700 and IRD800). Detection of a chromosomal inversion in *Oryza* species by mapping BAC end sequence pairs {#s4c} ----------------------------------------------------------------------------------------- Public BESs from 9 *Oryza* species (*O. sativa*, *O. rufipogon*, *O. glaberrima*, *O. nivara*, *O. punctata*, *O. minuta*, *O. officinalis*, *O. alta* and *O. australiensis*) developed in the frame of the Oryza Map Alignment Project (OMAP, <http://www.omap.org>) were downloaded from AGI web site (<http://www.genome.arizona.edu/stc/rice/>) [@pone.0017726-Ammiraju2]. BACs were mapped onto the *O. glaberrima S~1~* region by aligning BES pairs using BLASTN. BACs overlapping the chromosomal inversion breakpoints (as indicated by the alignment of the two BES of each BAC, inside and outside the inverted region, within a distance \<300,000 bp) were filtered, and the orientation of both BESs relative to the *O. glaberrima S~1~* genomic region was analyzed. Orthologous sequence comparisons {#s4d} -------------------------------- The orthologous CG14 *S~1~* regions were identified by BLASTN against the *O. sativa* ssp. *japonica* cv Nipponbare pseudomolecules (release v. 6.1) downloaded from the MSU Rice Genome Annotation Project web site [@pone.0017726-Ouyang1], and against the *O. sativa* ssp. *indica* 93-11 downloaded from the Beijing Genomic Institute web site (<http://rice.genomics.org.cn/rice2/link/download.jsp>). Sequence comparisons were carried out using the Dotter program [@pone.0017726-Sonnhammer1], the Artemis Comparison Tool [@pone.0017726-Carver1], and the EMBOSS package. The downloaded *O. sativa* sequences were re-annotated for genes and TEs with similar approaches used to annotate the *O. glaberrima* segment. To study microcollinearity between orthologous *O. glaberrima* and *O. sativa* sequences, the nucleotide sequences of non-TE coding genes were extracted for each segment and used as queries for BLAST alignments between each other to generate a distance matrix. Microcollinearity relationships were displayed using GenomePixelizer software (<http://www.atgc.org/GenomePixelizer/>). Calculation of nonsynonymous and synonymous nucleotide substitution rates {#s4e} ------------------------------------------------------------------------- Orthologous *O. sativa* and *O. glaberrima* annotated coding regions were aligned using the Needle tools [@pone.0017726-Rice1] to estimate the degree of gene structure conservations. Orthologous genes with clear distinct annotated gene structure were removed from further analysis. Calculations for nonsynonymous and synonymous nucleotides substitution rate were done as previously described [@pone.0017726-Garavito1]. Identical analyses were carried out with two control loci recently sequenced in *O. glaberrima*: *ADH* [@pone.0017726-Ammiraju1] from chromosome 11 (positions 5.598--5.750 Mbp) and *MOC1* from chromosome 6 (positions 24.25--24,40 Mbp) [@pone.0017726-Lu1]. A non-parametric statistical test (Kruskal--Wallis analysis of variance) was used to analyze the homogeneity of Ka and Ks data between the *S~1~* region and the *ADH* and *MOC1* loci. *P*\<0.05 was considered to be statistically significant to report non homogenous data. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### **Schematic representation of microcollinearity relationships between** ***O. glaberrima*** **cv. GC14 and** ***O. sativa*** **cv. Nipponbare** ***S~1~*** **regions.** Colored boxes indicate positions and orientation of non-TE genes along axes representing the CG14 (upper segment) and Nipponbare (lower segment) *S~1~* regions. Colored lines linking boxes symbolize high identity relationships between one, or several genes from CG14 and Nipponbare. Red boxes indicate genes lacking orthologs. Blue and green boxes represent the positions of duplicated genes in Nipponbare compared to CG14. Orange boxes indicate the positions of genes contained in the inversion. The positions of *S~1~A*, *S~1~* and *S~1~B* loci are indicated along the horizontal axis of CG14. Identified gene families in CG14 as classified in [Table 1](#pone-0017726-t001){ref-type="table"} are indicated below the diagram. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### **Mapping of pairs of BAC end sequences (BES) from seven** ***Oryza*** **species spanning the** ***O. glaberrima*** **inversion breakpoints.** Pairs of BAC end sequences (BES) were mapped on the *O. glaberrima S~1~* regions. BACs spanning the inversion breakpoints were symbolized on the *O. glaberrima* physical map as horizontal lines limited by colored arrows, representing orientations of BES (blue arrows for sense orientations and red arrows for antisense orientations). The name of each mapped BAC is indicated below each horizontal line. BACs limited by two BES in opposite orientation indicate similar organization of the inversion compared to *O. glaberrima*, while BACs limited by two BES in identical orientation suggest a different orientation compared to *O. glaberrima*. A tree in the left of the figure symbolizes the evolutionary relationships of the *Oryza* species used in this analysis as described in Ge *et al.*, 1999 (Proc Natl Acad Sci U S A, 96:14400--14405). (EPS) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S3 ::: {.caption} ###### **Comparison between structural variations identified between orthologous** ***S~1~*** **regions in CG14 and Nipponbare, and the genetic map of the** ***S~1~*** **locus between** ***O. sativa*** **and** ***O. glaberrima*** **.** Structural variations between orthologous *S~1~* regions (left) compared with the genetic map obtained from the *O. sativa*×*O. glaberrima* 779 BC~1~F~1~ plants previously described [@pone.0017726-Garavito1]. Marker positions in the CG14 physical map are shown. Blue frames on the physical maps of *O. glaberrima* cv. CG 14 and *O. sativa* ssp. *japonica* cv. Nipponbare indicate the position of the chromosomal inversion. (EPS) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S4 ::: {.caption} ###### **Frequencies of Ka and Ks values between orthologous coding sequences in the** ***O. glaberrima*** **and** ***O. sativa S~1~*** **,** ***ADH*** **and** ***MOC1*** **regions.** Distribution of the Ka and Ks values of the *S~1~* (**A**), *ADH* (**B**) and *MOC1* (**C**) orthologous regions from *O. sativa* ssp. *japonica* cv. Nipponbare and *O. glaberrima* cv. CG14. (EPS) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S5 ::: {.caption} ###### **Amino-acid alignment between duplicated F-box proteins in the** ***S~1~*** **orthologous regions.** Alignment of predicted amino-acid sequences of F-box genes in the *S~1~* and *S~1~A* loci from *O. sativa* ssp. *japonica* cv. Nipponbare (OSj), *O. sativa* ss. *indica* cv. 93-11 (OSi) et *O. glaberrima* cv. CG14 (OG). Blue box indicates the identified F-box domain. The conserved Leucine amino acids in the Leucine rich repeat regions are underlined in grey. *S~1~A*-OSj^p^ designates the duplicated and partial F-box protein in the Nipponbare *S~1~A* locus. (EPS) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S6 ::: {.caption} ###### **Phylogenetic relationships among the** ***S~1~*** **and** ***S~1~A*** **orthologous F-box proteins.** Phylogenetic tree of the *S~1~* and *S~1~A* orthologous F-box genes from *O. sativa* ssp. *japonica* cv. Nipponbare (OSj), *O. sativa* ssp. *indica* cv. 93-11 (OSi) and *O. glaberrima* cv. CG14 (OG). The unrooted tree was generated by the neighbor-joining method using ClustalX program. Numbers indicates bootstrap values with 1000 replicates. (EPS) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S1 ::: {.caption} ###### **List of identified genes in the 813 kb of the** ***O. glaberrima*** **cv. CG14** ***S~1~*** **regions.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S2 ::: {.caption} ###### **List of the different types of TE found in the** ***O. glaberrima S~1~*** **regions.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S3 ::: {.caption} ###### **List of identified pack-MULEs and enclosed genes in the** ***O. glaberrima*** **cv. CG14** ***S~1~*** **regions.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S4 ::: {.caption} ###### **Sequence comparison between orthologous coding sequences in the** ***O. glaberrima*** **cv. CG14 and** ***O. sativa*** **cv. Nipponbare** ***ADH*** **and** ***MOC1*** **regions.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S5 ::: {.caption} ###### **Sequence comparison between orthologous coding sequences in the** ***O. glaberrima*** **cv. CG14 and** ***O. sativa*** **cv. Nipponbare** ***S~1~*** **regions.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S6 ::: {.caption} ###### **New molecular markers designed in the** ***S~1~*** **regions.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Our thanks go to R.A. Wing and P.R. Marri (AGI, Arizona) for sharing data on the *O. glaberrima* whole genome sequence, and to P. Touzet et V. Poncet for their valuable comments on the manuscript. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was supported by the Generation Challenge Program. The Departement Soutien et Formation (DSF) from Institut de Recherche pour le Developpement (IRD), and Centro Internacional de Agricultura Tropical (CIAT) provided the PhD scholarship of A. Garavito. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: RG AG ML JT AG. Performed the experiments: RG AG FG SS. Analyzed the data: RG AG ML. Contributed reagents/materials/analysis tools: RG AG. Wrote the paper: RG AG ML. [^2]: ¤ Current address: UMR DIADE, Institut de Recherche pour le Développement (IRD), Montpellier, France

PubMed Central

2024-06-05T04:04:19.859381

2011-3-10

PMC3053391

Introduction {#s1} ============ Human oncogenic herpesviruses such as Epstein--Barr virus (EBV) and Kaposi\'s sarcoma-associated herpesvirus (KSHV) are closely linked to a variety of malignancies including nonkeratinizing nasopharyngeal carcinoma, gastric adenocarcinoma, Burkitt\'s lymphoma, Kaposi\'s sarcoma, primary effusion lymphoma, multicentric Castleman\'s disease, and various forms of lymphoproliferative disorders. Both EBV and KSHV are latent residents in B lymphocytes and show sporadic reactivation in lymphoepithelial tissues such as tonsils [@pone.0017809-Pauk1], [@pone.0017809-Sixbey1], [@pone.0017809-ThorleyLawson1]. Lytic reactivation of EBV or KSHV in epithelial cells of the nasopharynx is strongly influenced by the state of differentiation [@pone.0017809-Feederle1], [@pone.0017809-Johnson1], [@pone.0017809-Young1]. In addition, XBP-1s, a product of the master gene responsible for B cell differentiation, was recently suggested to be one of the physiological stimuli that trigger the lytic switch of EBV and KSHV in latently infected B cells [@pone.0017809-Sun1], [@pone.0017809-Wilson1]. For a cycling cell, growth arrest in the G1 phase implies one of the following fates to choose: quiescence (re-enters proliferation at a later time), apoptosis, differentiation or senescence [@pone.0017809-Blomen1], [@pone.0017809-Pfeuty1]. Among these four outcomes, differentiation and senescence share two features in common: dramatic chromosome remodeling [@pone.0017809-SekeriPataryas1] and lengthy development time (usually days). Cell senescence is a biochemical process exhibited by metabolically active cells whose cell cycles are frozen beyond the restriction point in G1 phase. First identified in *in vitro* cultured cells, cellular senescence occurs both in primary and cancer cell lines [@pone.0017809-Hwang1], [@pone.0017809-Serrano1]. In addition, the limit in proliferative capacity triggered by aberrant mitogenic signals of oncogenes, known as oncogene-induced senescence (OIS), is an alternative tumor suppressive mechanism that has been recently validated *in vivo* [@pone.0017809-Collado1], [@pone.0017809-Mooi1]. Senescence not only occurs in pre-malignant cells, but also appears in malignant tumors. In the latter case, senescence was usually produced by the removal of an essential oncogenic stimulus or the restoration of a tumor suppressor. For example, ablation of c-Myc in transgenic mouse models induced rapid tumor regression associated with senescence [@pone.0017809-Wu1]; systemic expression of a dominant-interfering Myc mutant in a preclinical mouse model with Ras-initiated lung adenocarcinoma triggered rapid tumor regression accompanied by senescence [@pone.0017809-Soucek1]. Lytic replication of herpesvirus occurs preferentially in the G1 phase of the cell cycle [@pone.0017809-Flemington1]. Accumulating evidence indicates that a number of viral immediate-early proteins actively exert a growth-arrest function by which the virus induces a less competitive environment for resources required for viral DNA replication. In this regard, when EBV undergoes lytic replication, host cells are protected from apoptosis and the DNA-synthetic machinery is blocked, although the activities of certain S-phase regulators increase [@pone.0017809-Inman1], [@pone.0017809-Kudoh1]. In addition, upon viral infection, ICP0 of herpes simplex virus induces cell cycle arrest in G1 by both p53-mediated and p53-independent pathways [@pone.0017809-Hobbs1]. Through the up-regulation of cyclins E and D, the IE2 protein of human cytomegalovirus potently arrests U373 cells and simultaneously blocks cellular DNA synthesis [@pone.0017809-Noris1]. Conceivably, G1 phase is not only a pivotal stage for cell fate determination [@pone.0017809-Blomen1], [@pone.0017809-Pfeuty1], but also is critical for virus fate, namely to maintain latency or to initiate a lytic replication episode. Among the identified immediate-early molecules, RTA ([r]{.underline}eplication and [t]{.underline}ranscription [a]{.underline}ctivator) is positionally and structurally conserved among the genomes of all gamma-herpesviruses. Ectopic expression of the EBV Rta in epithelial or B cells is capable of efficiently inducing the lytic cycle of EBV [@pone.0017809-Ragoczy1], [@pone.0017809-Zalani1]. Similarly, ectopic expression of KSHV RTA (K-RTA) in B or endothelial cells latently infected with KSHV leads to the successive expression of KSHV early and late genes [@pone.0017809-Sun2], [@pone.0017809-Lukac1]. We recently established doxycycline-inducible system of Rta in 293 cells, nasopharyngeal carcinoma cells (NPC-TW01), and laryngeal carcinoma cells (HEp-2). We found that, in the absence of BZLF1 and other EBV viral proteins, Rta alone can promote irreversible G1 arrest followed by cellular senescence in these epithelial cells ([@pone.0017809-Chen1] and unpublished). In the present study, we further demonstrate that in 293 cells, the doxycycline-inducible Rta not only reactivates EBV but also KSHV, to similar efficacy. While the precise mechanism of Rta-mediated KSHV reactivation is currently not resolved, results from comparative kinetics studies strongly indicate a casual role of Rta-induced G1 arrest in EBV and KSHV reactivations. Results {#s2} ======= EBV Rta alone is sufficient to initiate and complete lytic EBV replication in EREV8 cells {#s2a} ----------------------------------------------------------------------------------------- EBV Rta alone is known to disrupt EBV latency in both epithelial and B cells [@pone.0017809-Ragoczy1], [@pone.0017809-Zalani1]. 293TetER is a recently established 293 cell line that displays doxycycline (Dox) -controlled, conditional expression of EBV Rta [@pone.0017809-Chen1]. To confirm whether the expression of EBV Rta in 293TetER cells is sufficient to promote EBV lytic replication from the latent stage, rAkata-G418 EBV genome was transferred into 293TetER cells, yielding an EREV8 derivative line [@pone.0017809-Lee1]. To measure the induction rate of the EBV lytic cycle triggered by EBV Rta, Dox (50 ng/ml)-treated EREV8 cells were analyzed using an immunofluorescence assay and flow cytometry. As expected, the immunofluorescence assay showed a very high and homogenous expression of transgene Flag-EBV Rta ([Figure 1A](#pone-0017809-g001){ref-type="fig"}), and flow cytometry showed that ≈76% of the cells were positive. Similarly, considerable expression of immediate-early protein BZLF1 (≈82%) and late glycoprotein protein BALF4/gB (≈50%) were detected in the Dox-treated EREV8 cells ([Figure 1A](#pone-0017809-g001){ref-type="fig"}). Next, the expression kinetics of a panel of lytic proteins including BZLF1, BMRF1, BHRF1, and membrane protein gp350/220 in EREV8 cells were compared in parallel using western blot analysis. Although the degree of antibody affinity may vary, the overall kinetics of the different proteins was distinguishable ([Figure 1B](#pone-0017809-g001){ref-type="fig"}). The optimal expressions for each protein were in a hierarchical order: namely Flag-Rta (24--48 h), EBV immediate-early protein BZLF1 (48--72 h), early proteins BMRF1 and BHRF1 (72--144 h), and late membrane protein gp350/220 (120--144 h). Finally, the quantity of EBV genome equivalents encapsidated in the viral particles released from Dox-treated EREV8 cells at each time points were determined by comparative quantitative PCR (q-PCR). As shown in [Figure 1C](#pone-0017809-g001){ref-type="fig"}, the Dox-treated EREV8 cells continued to produce viral particles in an increasing manner until 144 h. Noticeably, at 96 h after induction, a subpopulation of cells started to round up and seemed to be full of granules. Anoikis-like detachment of cells from the petri dish was observed at later times (detailed below). These observations are reminiscent of proficient permissive replication of bacteriophage in *E. coli*. ::: {#pone-0017809-g001 .fig} 10.1371/journal.pone.0017809.g001 Figure 1 ::: {.caption} ###### Reactivation of EBV lytic replication by EBV Rta in EREV8 cells. \(A) Individual induction efficiency of Flag-EBV Rta (Flag), BZLF1, and late glycoprotein BALF4/gB were shown by an immunofluorescence assay. Untreated (--) or EREV8 cells treated with doxycycline (Dox) for 48 h were analyzed in parallel. Cells with immuno-positivity were quantified by flow cytometry and indicated in percentages for each detection. (B) Expression kinetics of EBV lytic proteins including BZLF1, BMRF1, BHRF1, and gp350/220 in Dox-induced EREV8 cells for the indicated times were analyzed by western blot analysis. β-actin served as a loading control. C6 and C72 indicate untreated cells at 6 h and 72 h, respectively. (C) Titration of viral particles released from 72--144 h Dox-treated EREV8 cells. Copy numbers of DNase I-resistant, encapsidated viral DNAs were determined by comparative quantitative PCR of EBV DNA polymerase gene (BALF5) using serial dilutions of Raji DNA as standards. Data are presented as means±SD from six independent PCR assays. :::  ::: To compare the "throughput" of lytic cycle replications initiated by the present system with that by conventional method (20 ng/ml 12-*O*-tetradecanoylphorbol-13-acetate (TPA) plus 3 mM butyrate), EREV8 cells treated with either method were performed simultaneously for a course of 96 h, a time when most of the chemical-treated cells were dead. Of note, in an immunofluorescence assay, while closed to 70% of Dox-treated cells expressed Flag-Rta at 24 h, only 43% of the chemical-treated EREV8 cells responded to TPA/butyrate (i.e. BZLF1 positive), indicating the presence of a considerable refractory subpopulation in the traditional induction system. For each time point, the expressions of EBV immediate-early protein Rta, BZLF1, and early protein BMRF1 were revealed by western blot analysis ([Figure 2A](#pone-0017809-g002){ref-type="fig"}) and the yields of encapsidated viral particles were determined by comparative q-PCR ([Figure 2B](#pone-0017809-g002){ref-type="fig"}). Interestingly, while Rta was detectable as early as 6 h and increasingly augmented until 48 h in the Dox-treated cells, Rta was only transiently upregulated at 24 h in the TPA/butyrate-treated cells. By contrast, the expressions of BZLF1 and BMRF1 were significantly augmented in the chemical-treated cells than in the Dox-treated group by 48 h. Furthermore, although TPA/butyrate-induced EREV8 cells appeared to shed EBV particles at earlier time (24 h), the overall viral yields in the Dox-treated cell were ≈1.5-higher than that in the chemical-treated cells at 96 h ([Figure 2B](#pone-0017809-g002){ref-type="fig"}). In addition, since the Dox-treated cells may continue to shed viral particles at an increasingly manner until 196 h ([Figure 1C](#pone-0017809-g001){ref-type="fig"}), it is estimated that from same number of EREV8 cells, the viral particles produced by Dox treatment will be at least 4-fold more than that by chemical induction. Taken together, these results indicated that although TPA/butyrate provided a faster and stronger stimulus for EBV reactivation in the EREV8 cells, yet the induction rate was poorer, the cells were sicker and the viral yield was lower than those triggered by Dox-inducible Rta. ::: {#pone-0017809-g002 .fig} 10.1371/journal.pone.0017809.g002 Figure 2 ::: {.caption} ###### Comparative studies of EBV reactivation induced by Dox (50 ng/ml) vs. conventional chemical method (20 ng/ml TPA plus 3 mM butyrate) in EREV8 cells. \(A) Expression kinetics of EBV Rta, BZLF1 at 6, 24, and 48 h in cells treated with indicated inducers. β-actin served as a loading control. (B)Viral particles released from Dox-treated or TPA/butyrate-treated EREV8 cells. Data are presented as means±SD from four independent PCR assays. :::  ::: EBV Rta alone is sufficient to initiate and complete lytic KSHV replication in ERKV cells {#s2b} ----------------------------------------------------------------------------------------- In the course of establishing EREV8, a control experiment was performed in which rKSHV.219 [@pone.0017809-Vieira1] was used to infect 293TetER cells and served to differentiate the specificity of EBV Rta for its cognate viral genome, as described previously [@pone.0017809-Sun2]. rKSHV.219 carries a genetic cassette that can be used to distinguish the stages of viral infection in the host cells: latent (green fluorescence) and lytic (red fluorescence) [@pone.0017809-Vieira1]. At first, we expected that Dox-induced EBV Rta would reactivate EBV but not KSHV genomes residing in the 293TetER cells. Somewhat surprisingly, upon the administration of Dox, a number of 293TetER cell clones harboring rKSHV.219 genomes exhibited strong red fluorescence, indicating lytic replication ([Figure 3A](#pone-0017809-g003){ref-type="fig"}). By contrast, only a low percentage of Dox-treated control 293Tet cells harboring rKSHV.219 genome exhibited red fluorescence ([Figure 3A](#pone-0017809-g003){ref-type="fig"}, 293Tet\_rKSHV\_C1), suggesting that EBV Rta was the determinant that triggered rKSHV.219 lytic cycle replication. To verify these observations further, five 293\_TetER\_rKSHV.219 cell clones were expanded, pooled, and collectively designated as ERKV. Stable latent infection of rKSHV.219 in ERKV cells was achieved by puromycin selection using previously described procedures [@pone.0017809-Vieira1]. Next, the expression kinetics of KSHV immediate-early protein K-RTA, early protein K-bZIP, and late protein K8.1 were studied by western blot analysis. Again, the overall expression pattern of these four molecules could be arranged in a cascade manner by their respective peak times: namely Flag-EBV Rta (24--48 h), immediate-early K-RTA and K-bZIP (48--72 h), and late glycoprotein K8.1 (72--96 h) ([Figure 3B](#pone-0017809-g003){ref-type="fig"}). Interestingly, expression of the three KSHV lytic proteins was extinguished at 168 h, suggesting no resources were available for virus multiplication. The titers of KSHV particles released into the culture medium at different time points were determined by comparative q-PCR of cell-free, encapsidated KSHV genome equivalents. The results showed that viral particles manufactured in ERKV cells were about 3-fold to that produced by EREV8 cells at 96 h (18 vs. 6 millions/ml), however, the production was plateaued afterwards ([Figure 3C](#pone-0017809-g003){ref-type="fig"}), reinforcing cellular resources for KSHV replication were exhausted after 96 h. ::: {#pone-0017809-g003 .fig} 10.1371/journal.pone.0017809.g003 Figure 3 ::: {.caption} ###### Reactivation of KSHV lytic replication by EBV Rta in ERKV cells. \(A) Latent and lytic infections of rKSHV.219, indicated by GFP and RFP, respectively, were inspected in control 293Tet (293Tet\_rKSHV\_C1) and 293TetER cells (293TetER\_rKSHV\_E2 and E10) treated with doxycycline (Dox) for 48 h. R/G represents the fraction of RFP-expressing cells in the population determined by using Image J (NIH). Five 293TetER\_rKSHV subclones with R/G \>70% were pooled and collectively referred to as ERKV. (B) The expression kinetics of KSHV lytic proteins (K-RTA, K-bZIP, K8.1) in control (C6 and C72) and Dox-treated (6--168 h) ERKV cells were examined by western blot analysis. (C) Titration of KSHV particles released from Dox-treated ERKV cells. Copy numbers of DNase I-resistant, encapsidated viral DNAs in each filtrated (0.45 µm) viral supernatants were determined by comparative quantitative PCR of KSHV DNA polymerase gene (ORF9) using serial dilutions of cosmid GB11 DNA as standards. Data are presented as means±SD from six independent PCR assays. (D) Titration of infectious KSHV particles from Dox-treated ERKV cells. Aliquots of filtrated supernatants were used to infect fresh 293 cells. Two days after infection, the numbers of GFP-positive cells, designated as "green 293 units", in each infection were counted under a fluorescence microscope. Error bars depict standard deviations of three independent counts. Two independent experiments were performed, one set of results is shown. (E) A luciferase reporter gene assay was used to screen the responsiveness of various viral and cellular promoters to EBV Rta (ER) and K-RTA (KR) in 293 cells. The error bars of each column indicate the standard deviation of each set of triplicate wells. The transfection efficiency of each sample was validated by Western blot analysis using M2 Flag monoclonal antibodies. :::  ::: To determine the infectivity of these viral particles, an aliquot of the filtrated supernatant was used to infect fresh 293 cells, and the green fluorescence-glowing cells, dubbed as "green 293 units" were determined by fluorescence microscopy ([Figure 3D](#pone-0017809-g003){ref-type="fig"}). The highest titer produced at Dox 96 h, 1.8×10^5^ units/ml, is ≈30-fold higher than that induced by the combination of sodium butyrate and K-RTA in the same 293 background described previously (Fig. 8D in [@pone.0017809-Vieira1]), indicating that this new system to induce lytic KSHV replication is very robust. Furthermore, since K-RTA is by far the only known immediate-early protein that is required and sufficient to complete a lytic cycle replication, to confirm whether K-RTA is the only gene activated by EBV Rta in ERKV cells, luciferase reporter gene assays were used to analyze the responsiveness to the cotransfected EBV Rta proteins of a panel of KSHV viral promoters. Two known responders of EBV Rta, namely promoters of EBV BGLF5 and cellular p21, were included as controls. As shown in [Figure 3E](#pone-0017809-g003){ref-type="fig"}, the EBV BGLF5 and cellular p21 promoter sequences were responsive to EBV Rta, whereas the three KSHV lytic promoters (PANp, ORF57p, and K-bZIPp) were preferentially responsive to K-RTA, as expected. These results established the respective specificity of EBV Rta and K-RTA for their cognate responsive elements in the present assay. Intriguingly, when the promoter of K-RTA was considered, even though the expression of EBV Rta was much less than that of K-RTA, the K-RTA promoter exhibited a significantly stronger responsiveness to EBV Rta than to K-RTA. Taken together, these results suggest that in Dox-treated ERKV cells, EBV Rta efficiently up-regulates the expression of K-RTA, followed by the activation of numerous KSHV lytic promoters and DNA replication elicited by K-RTA itself. In summary, unexpectedly, we found that EBV Rta alone is also sufficient for initiating and completing the lytic replication of KSHV in 293 cells. Similar to EREV8, ERKV cells consistently became aggregated and disrupted on the fourth day after Dox induction, suggestive of permissive viral replication in these cells. Long-term Dox-treated EREV8 and ERKV cells displayed growth arrest followed by cell death {#s2c} ----------------------------------------------------------------------------------------- We demonstrated previously that EBV Rta can initiate a sustained and irreversible G1 arrest, a hallmark of cellular senescence in both 293 and NPC cells [@pone.0017809-Chen1]. In the current study, we observed that EBV Rta alone is sufficient to induce and complete the lytic cycle of EBV and KSHV latent genomes in 293TetER cells. To characterize further the molecular phenotypes imposed by EBV Rta-mediated processes, the growth curves and metabolic activities of Dox-treated and -untreated 293TetER, EREV8, and ERKV were followed for eight days. As depicted in [Figure 4A](#pone-0017809-g004){ref-type="fig"}, the growth curves for 293TetER, EREV8, and ERKV were similar. A noticeable growth plateau was observed at 144 h, indicating the time when the cells were confluent and the nutrient was depleted from the culture media. In the Dox-treated group, the number of 293TetER, EREV8, and ERKV cells continued to increase until 48 h, after which 293TetER cell count remained stable through the end of the experiment, when the Dox-treated 293TetER cells became senescent, as described previously [@pone.0017809-Chen1]. By contrast, the number of live Dox-treated EREV8 and ERKV cells declined gradually from 72 to 192 h, suggesting that cell death occurred during this time. In parallel, the metabolic activity of each cell line under each treatment was measured using a WST-1 assay ([Figure 4B](#pone-0017809-g004){ref-type="fig"}). In general, the metabolic activities were consistent with the cell number counts. In the untreated group, all three lines exhibited a growth peak at 144 h, followed by an abrupt drop at 192 h, indicating nutrient deprivation and culture confluence. The metabolic activity was maintained in senescent 293TetER cells (a key feature of senescence ([@pone.0017809-Blagosklonny1], [@pone.0017809-Demidenko1]), but the metabolic activity in Dox-induced EREV8 and ERKV cells increased in the first 48 h and then declined progressively from 72 to 192 h ([Figure 4B](#pone-0017809-g004){ref-type="fig"}). Taken together, these results established that without Dox treatment, 293TetER, EREV8, and ERKV cells exhibit similar growth rates; yet, after Dox induction, EBV Rta led the three cell lines to different fates: 293TetER cells became senescent whereas EREV8 and ERKV cells died eventually. The different cell fates of Dox-treated 293TetER and 293TetER cells containing viral genomes (EREV8 and ERKV) were reflected in distinct cell morphologic changes. Specifically, disruptively rounded-up and anoikis-like cells started to be detectable in 96 h Dox-treated EREV8 and ERKV cells, and were especially prominent from 120 to 192 h ([Figure 4C](#pone-0017809-g004){ref-type="fig"}). By contrast, during the same time course, 293TetER cells remained flattened and enlarged without further changes in cell shape ([Figure 4C](#pone-0017809-g004){ref-type="fig"} and [@pone.0017809-Chen1]). These results suggest that the permissive lytic replications of EBV and KSHV in EREV8 and ERKV cells, respectively, may be the main cause of cell death. ::: {#pone-0017809-g004 .fig} 10.1371/journal.pone.0017809.g004 Figure 4 ::: {.caption} ###### Growth curves and metabolic rate of 293TetER, EREV8, and ERKV cells with or without Dox treatment. \(A) Cell growth rates of 293TetER, EREV8, and ERKV were determined by seeding triplicate wells (20,000 cells/well) in a 24-well tissue culture plate for each time point. In parallel, duplicate plates were prepared, and 50 ng/ml Dox was added to each sample 24 h after cell seeding. At the specified times, cells were counted using the trypan blue-exclusion method. Error bars denote standard deviations of triplicate wells. Four independent experiments were performed, one set of representative results is shown. ST, starting numbers. (B) Cell metabolic activity was measured using the WST-1 assay. Cell preparations and Dox treatment were identical to the procedure described in (A). Three independent experiments were performed, and one set of representative results is shown. (C) Disrupted morphologic changes were observed in 120 h Dox-treated EREV8 and ERKV cells, but not in 293TetER cells that exhibited cellular senescence. :::  ::: Rta modulated the expressions of G1 arrest genes in 293 and nasopharyngeal carcinoma (NPC-TW01) cells {#s2d} ----------------------------------------------------------------------------------------------------- In addition to 293TetER, we have recently established Tet-on Rta inducible system in NPC cell background, referred to as TW01TetER. As a step to dissect Rta\'s role in cell cycle, genome-wide transcriptome analysis was conducted in Dox-treated 293TetER and TW01TetER cells (GEO accession \# GSE24587). There were 120 genes commonly modulated by Rta in these two inducible cell lines (fold-change ≥1.9). Among these, more than 90% (109 out of 120) were modulated in the same direction in both cell lines. In addition, the inductions of FASN and MERTK were consistent with a previous microarray study (n = 5,000) in which human keratinocytes were infected with adenovirus vector expressing Rta [@pone.0017809-Li1], [@pone.0017809-Li2]. Thus, some of the transcriptional acts imposed by Rta may be conserved in different cell types. Next, gene ontology analysis [@pone.0017809-Huangda1], [@pone.0017809-Zhang1] was employed to classify these candidates into functionally-related gene sets. As such, the cell cycle-related genes were revealed with p-value≈0.01 in both analyses ([Table 1](#pone-0017809-t001){ref-type="table"}). Interestingly, although some of the genes, e.g. CDK6 in TW01TetER ([Table 2](#pone-0017809-t002){ref-type="table"}), whose expressions were not altered significantly at the mRNA level, we were able to confirm the expressions of five G1 arrest-related genes in both cell types by western blot analysis, including CCND2, CDK6, c-Myc, p21 and 14-3-3σ (detailed below). Therefore, Rta-induced G1 arrest seemed primarily originated from the transcriptional level. ::: {#pone-0017809-t001 .table-wrap} 10.1371/journal.pone.0017809.t001 Table 1 ::: {.caption} ###### Partial lists of gene ontology analyses of 120 genes commonly modulated by Rta in 293 cells and NPC-TW01 cells. ::: {#pone-0017809-t001-1} Common\_120\_DAVID [@pone.0017809-Huangda1] Gene no. p-Value --------------------------------------------------------------------- ---------- ---------- regulation of cell proliferation (GO:0042127) 15 1.43E-04 response to organic substance (GO:0010033) 14 2.23E-04 response to mechanical stimulus (GO:0009612) 5 2.99E-04 tissue morphogenesis (GO:0048729) 7 5.93E-04 regulation of apoptosis (GO:0042981) 14 6.35E-04 positive regulation of macromolecule metabolic process (GO:0010604) 14 1.15E-03 response to virus (GO:0009615) 5 3.60E-03 regulation of DNA metabolic process (GO:0051052) 5 4.23E-03 wound healing (GO:0042060) 6 4.89E-03 response to steroid hormone stimulus (GO:0048545) 6 5.00E-03 // **cell cycle** (GO:0007049) 11 1.34E-02 Common\_120\_WebGestalt [@pone.0017809-Zhang1] Gene no. p-Value -------------------------------------------------- ---------- ---------- response to stress (GO:0006950) 24 1.00E-03 organ morphogenesis (GO:0009887) 14 1.70E-03 regulation of cell proliferation (GO:0042127) 14 1.70E-03 response to organic substance (GO:0010033) 13 2.00E-03 response to chemical stimulus (GO:0042221) 18 3.60E-03 epithelium development (GO:0060429) 7 6.00E-03 anatomical structure morphogenesis (GO:0009653) 17 6.10E-03 regulation of DNA metabolic process (GO:0051052) 5 6.10E-03 response to steroid hormone stimulus(GO:0048545) 6 6.50E-03 response to virus (GO:0009615) 5 6.50E-03 // **cell cycle** (GO:0007049) 13 1.09E-02 Identified GO terms are sorted by p-Value. ::: ::: {#pone-0017809-t002 .table-wrap} 10.1371/journal.pone.0017809.t002 Table 2 ::: {.caption} ###### List of cell cycle related genes modulated by EBV Rta in 293 and NPC-TW01 cells. ::: {#pone-0017809-t002-2} Gene Symbol 293TetER TW01TetER Gene Description ------------------------------------------ --------------------------------------- ----------- ------------------------------------------------------- CCND1 −3.1[\#](#nt102){ref-type="table-fn"} −1.1 cyclin D1 CCND2 −3.1 −3.3 cyclin D2 CDK4 −1.3 −2.2 cyclin-dependent kinase 4 CDK6 −1.8 −1.1 cyclin-dependent kinase 6 CDKN1A 2.2 1.0 cyclin-dependent kinase inhibitor 1A (p21, Cip1) CHEK2 −1.2 −1.9 CHK2 checkpoint homolog (*S. pombe*) GNL3 −2.0 −2.1 guanine nucleotide binding protein-like 3 (nucleolar) H1F0 3.8 2.4 H1 histone family, member 0 HERC5 2.2 1.9 hect domain and RLD 5 HEXIM1 2.3 1.3 hexamethylene bis-acetamide inducible 1 HSPA2 2.8 2.3 heat shock 70kDa protein 2 IFITM1 5.0 1.8 interferon induced transmembrane protein 1 LMLN 1.9 2.2 leishmanolysin-like, metallopeptidase M8 family MYC −3.0 −4.5 v-myc myelocytomatosis viral oncogene homolog NEFH 6.6 1.3 neurofilament, heavy polypeptide NEFL −2.3 −3.1 neurofilament, light polypeptide NOLC1 −1.8 −2.4 nucleolar and coiled-body phosphoprotein 1 PTTG2 1.1 2.0 pituitary tumor-transforming 2 RRS1 −2.6 −2.5 RRS1 ribosome biogenesis regulator SFN 14.3 1.6 stratifin (14-3-3σ) TGFB2 −2.1 −1.3 transforming growth factor, beta 2 FASN[\#\#](#nt103){ref-type="table-fn"} 2.2 2.2 fatty acid synthase MERTK[\#\#](#nt103){ref-type="table-fn"} 8.3 21.5 c-mer proto-oncogene tyrosine kinase \# : Fold-change compared to the control groups. Details for the experimental design and data process procedures are referred to Gene Expression Omnibus (GEO) under accession number GSE24587. \#\# : Also observed in primary keratinocytes transduced by adenovirus vector expressing Rta [@pone.0017809-Li1], [@pone.0017809-Li2]. ::: Comparative analysis of Rta-mediated G1 arrest and viral reactivation {#s2e} --------------------------------------------------------------------- Previous results showed that Rta universally modulates the expressions of G1 signature proteins in 293 and NPC cells. These alterations not only support the idea that senescence is preceded by an irreversible G1 arrest [@pone.0017809-Blagosklonny1], [@pone.0017809-Campisi1], but also are reminiscent of a common function of herpesviral immediate-early genes, namely to halt cell cycle progression in G1 [@pone.0017809-Flemington1]. Therefore, we questioned whether G1 arrest was maintained in Dox-induced EREV8 and ERKV cells. To this end, short-term Dox-treated 293TetER, EREV8 and ERKV cells were subjected to flow cytometric analysis. As depicted in [Figure 5A](#pone-0017809-g005){ref-type="fig"}, in all three cell lines, G1-populations are increasingly proportional to the Dox induction time, indicating that regardless of viral genomes, Rta-mediated G1 arrest was sustained in both EREV8 and ERKV cells. In order to dissect the time sequence of Rta-mediated host G1 arrest and viral reactivations, we compared the expressions of cell cycle related genes to that of the viral lytic switches. As shown in [Figure 5B](#pone-0017809-g005){ref-type="fig"}, in all three cells lines, the decreased expressions of c-Myc, CDK6, CCND2 and increased expressions of p21, 14-3-3σ were temporally associated with the concentration of Dox-inducible Rta between 6 and 48 h. In addition, phosphorylated pRb (S807/S811) was accordingly diminished in all three cell lines, a strong indication of G1 arrest. In marked contrast, an evident induction of viral lytic switch genes, namely BZLF1 in EREV8 and K-RTA in ERKV cells, always lagged behind cellular gene alterations, suggesting that Rta-mediated host G1 arrest preceded the onset of viral reactivation. It is worth noting that the decrement of c-Myc before induction of K-RTA ([Figure 5B](#pone-0017809-g005){ref-type="fig"}, ERKV) agrees with a recent report in which elimination of c-Myc led to KSHV reactivation in primary effusion lymphoma cells [@pone.0017809-Li3]. Finally, by taking advantage of fluorescence markers residing in the rKSHV.219 genome, we observed that in the Dox 120 h-treated ERKV group, the remaining adherent cells with SA-β-Gal positivity were mostly void of virus lytic replication (not shown), indicating that exhibition of senescence marker and KSHV reactivation were mutually exclusive. Taken together, these results led us to hypothesize that in 293 cells, the Dox-inducible Rta efficiently induces a G1 arrest (6--48 h) that is an ideal environment for EBV or KSHV lytic cycle progression (48--168 h) and is a favorable preceding event for cellular senescence (120 h -- ∞). ::: {#pone-0017809-g005 .fig} 10.1371/journal.pone.0017809.g005 Figure 5 ::: {.caption} ###### Rta-mediated cell cycle arrest precedes the expressions of viral immediate-early genes. \(A) Dox-treated 293TetER, EREV8, and ERKV cells cultured for 24 and 48 h were subjected to flow cytometry analysis to quantify the cellular DNA content. The distributions of cells residing in the G2/M, S, G1, and subG1 stages at each time are shown. The results of three independent experiments were similar, and one representative dataset is shown. (B) Comparative expression kinetics (0--48 h) of cell cycle regulators or viral immediate-early proteins in Dox treated 293TetER, EREV8 and ERKV cells. Down-regulation of cell cycle activators (c-Myc, CDK6, CCND2, phosphorylated pRb) and up-regulation of cell cycle inhibitors (p21, 14-3-3σ) are temporally associated with the expression of Rta in all three cell lines. In comparison, EBV BZLF1 and KSHV K-RTA are not significantly augmented until 48 h, a time that alterations of cell cycle gene are nearly completed. α-tubulin served as a loading control. :::  ::: Discussion {#s3} ========== EBV Rta is a transcriptional activator with high plasticity in viral genome recognition [@pone.0017809-Chen2], [@pone.0017809-Gruffat1]. Results from microarray analysis in different cell backgrounds suggest that Rta also binds to and efficiently modulates the expression of host genome ([@pone.0017809-Li1], [@pone.0017809-Li2] and [Table 2](#pone-0017809-t002){ref-type="table"}). Here, we investigate the sequential events when Rta encounters host and viral genomes at the same time. First, it is confirmed that Rta efficiently modified the expressions of key cell cycle regulators of which three are related to cellular senescence (c-Myc [@pone.0017809-Wu1], p21 [@pone.0017809-Blagosklonny1], and 14-3-3σ [@pone.0017809-Schultz1]) ([Figure 5B](#pone-0017809-g005){ref-type="fig"}). Second, we observed that Rta-mediated cellular gene alterations preceded the induction of viral immediate-early genes BZLF1 and K-RTA. These phenomena strongly suggested that Rta may utilize a consensus theme to control cell cycles and viral reactivations. Our results support some previous findings [@pone.0017809-Kudoh1], [@pone.0017809-Cayrol1], [@pone.0017809-Izumiya1], [@pone.0017809-Rodriguez1], [@pone.0017809-Wu2], [@pone.0017809-Wu3] and disagree with some other ones [@pone.0017809-Guo1], [@pone.0017809-Swenson1], [@pone.0017809-Zacny1]. Rodriguez *et al.* demonstrated that a significant G1 bias was associated with early stages of chemically-induced EBV lytic cycle progression in NPC and B cells [@pone.0017809-Rodriguez1]. Kudoh *et al.* showed that induction of EBV lytic replication in Tet-On BZLF1 B95-8 cells completely arrested cell cycle progression at G1/S transition and blocked cellular DNA synthesis [@pone.0017809-Kudoh1]. When a single gene system is concerned, both EBV BZLF1 and KSHV K-bZIP elicited distinct pathways to arrest host cell cycle in G1 stage in various cellular backgrounds [@pone.0017809-Cayrol1], [@pone.0017809-Izumiya1], [@pone.0017809-Wu2], [@pone.0017809-Wu3]. Thus, our results suggest that the Rta-induced G1 arrest in EREV8 and ERKV cells indeed provided an adequate environment for virus reactivation. By contrast, Zacny *et al.* [@pone.0017809-Zacny1] and Swenson *et al.* [@pone.0017809-Swenson1] observed that Rta interacted with pRb that in turn released E2F1 and activated an S phase in Akata cells, U-2 OS cells and contact-inhibited fibroblasts. Guo *et al.* reported that over-expressions of BZLF1 or Rta in Raji cells resulted in degradation of pRb, accumulation of E2F1 and promotion of S phase entry [@pone.0017809-Guo1]. It bears to note that in our system the concentrations of E2F1 and pRb were not dramatically modulated by the Dox-inducible Rta ([@pone.0017809-Chen1] and unpublished). Thus, different cellular context may account for these discrepancies and more experiments are required to resolve this puzzle. Permissive EBV or KSHV replications have been previously demonstrated in differentiated cells *in vivo* and *in vitro* [@pone.0017809-Feederle1], [@pone.0017809-Johnson1], [@pone.0017809-Sun1], [@pone.0017809-Wilson1], [@pone.0017809-Greenspan1], [@pone.0017809-Hadinoto1]. In contrast, it is less clear for herpesviruses replicating in senescent cells. So far, papillomavirus E2 [@pone.0017809-Wells1], human cytomegalovirus IE2 [@pone.0017809-Noris1] and EBV Rta [@pone.0017809-Chen1] are the only known viral products involved in cellular senescence. E2 was previously shown to induce cellular senescence in HPV infected HeLa cells by restoring the functions of p53 and pRB [@pone.0017809-Psyrri1], [@pone.0017809-Wells2]. Whether IE2 or Rta-induced cellular senescence contributes to viral pathogenesis *in vivo* is worthy of further investigation. Of note, since both BZLF1 and Rta possess G1 arrest function, a synergistic effect of BZLF1 and Rta in cell cycle arrest is expected. Furthermore, Kalla *et al.* recently demonstrated that BZLF1 and Rta were expressed as immediate-early genes following primary EBV infection of B lymphocytes [@pone.0017809-Kalla1]. However, these early-expressed BZLF1 and Rta failed to initiate the EBV lytic cycle owing to the intruding viral genome was in an un-methylated status [@pone.0017809-Kalla1]. Therefore, we hypothesize that in such a transient-lytic phase where only the host genome is accessible, Rta (and BZLF1) may exert to trigger a cell senescence process. Among genes modulated by Rta depicted in [Figure 5B](#pone-0017809-g005){ref-type="fig"}, the sharply decreased expression of c-Myc by EBV Rta has two implications. First, one oncogenic role of c-Myc was suggested to be a repressor of cellular senescence [@pone.0017809-Wu1], [@pone.0017809-Zhuang1]. In our previous report, we demonstrated that EBV Rta efficiently induces cellular senescence in 293, NPC-TW01 and HONE-1 cells [@pone.0017809-Chen1]. Here we further confirm that in these senescent cells the decrement of c-Myc was one of the earliest events modulated by EBV Rta. Thus, decreased expression of c-Myc via Rta seems to participate in Rta-induced cellular senescence. Second, c-Myc is a negative regulator of KSHV lytic cycle replication [@pone.0017809-Li3], [@pone.0017809-Liu1]. RNAi-mediated knockdown of c-Myc resulted in disruption of KSHV latency and increment in mRNA and protein levels of K-RTA [@pone.0017809-Li3]. Consistent with these results, we observed that the reactivation of KSHV latent genome was preceded by a gradual decrement of c-Myc in the ERKV cells ([Figure 5B](#pone-0017809-g005){ref-type="fig"}). In addition, in a luciferase-reporting assay, the promoter sequences of K-RTA, but not those of K-bZIP, PAN, and ORF57 were preferentially activated by ectopic expression of Rta ([Figure 3E](#pone-0017809-g003){ref-type="fig"}). Thus, we hypothesize that either a direct act from Rta alone, or via down-regulated c-Myc, or both, are attributable to Rta-induced K-RTA synthesis in Dox-treated ERKV cells. EBV Rta is not the only variant that cross-reactivates KSHV; other viral factors including the HCMV *UL112-113* locus [@pone.0017809-Wells3] and HIV-1 *tat* protein [@pone.0017809-Harrington1] have also been ascribed to possess such functionality, suggesting that other viral infections may also participate in KSHV pathogenesis. Further, although permissive EBV or KSHV lytic replication were detectable *in vivo*, but a homogenous and thorough lysis of host cell by viral lytic replication is still lacking *in vitro*. Here, we have produced a model that provides a nearly permissive replication system for both EBV and KSHV that is controlled directly by EBV Rta. This system offers two advantages over the conventional approaches. First, the stimulus, 50 ng/ml Dox, is a very dilute, physiologically neutral compound. Compared with the conventional sodium butyrate or phorbol ester, Dox elicits far fewer, possibly no, undesirable effects on the treated cells. Second, the treatment produces homogenous results. Routinely, Flag-tagged EBV Rta and BZLF1 were detected in close to 80% of the 48 h Dox-treated EREV8 cells when assessed by an immunofluorescence assay. Similarly, more than 80% of the treated ERKV cells produced red fluorescence 48 h after induction. Our newly established EREV and ERKV cells thus provide a feasible system for elucidating host factors and viral determinants that contribute to regulate the EBV and KSHV reactivations. Materials and Methods {#s4} ===================== Cell culture {#s4a} ------------ 293TetER is a doxycycline inducible, EBV Rta conditional expression cell lines created by Virapower system™ (Invitrogen, Carlsbad, CA) [@pone.0017809-Chen1]. Same procedures were carried out to establish TW01TetER in which inducible Rta was expressed in a nasopharyngeal carcinoma cell line, NPC-TW01 [@pone.0017809-Lin1]. EREV8 is an EBV positive 293TetER derivative line generated by using cell-to-cell infection method [@pone.0017809-Lee1]. ERKV is a KSHV positive 293TetER derivative line that was stably infected with rKSHV.219 [@pone.0017809-Vieira1]. Specifically, 293TetER cells incubated with rKSHV.219 viral sup for 48 h were selected with 660 µg/ml puromycin for three weeks to obtain green fluorescent clones. Twelve such cell colonies were isolated, expanded, and determined for inducibility of KSHV lytic replication (red fluorescence) by 50 ng/ml doxycycline treatment. Five clones with high inducibility (70--90%) were pooled and used to compose the first generation of ERKV. 293TetER, EREV8 and ERKV cells were maintained in DMEM containing 10% Tet System Approved FBS (Clontech Laboratories, Mountain View, CA), 5 µg/ml blasticidin-S-HCl (Invitrogen) and 200 µg/ml zeocin (Invitrogen). To maintain the latently infected viral genomes, EREV8 and ERKV cultures were further supplemented with 400 µg/ml G418 and 660 µg/ml puromycin, respectively. Plasmids {#s4b} -------- pLenti4-Flag-CPO is a modified expression plasmid derived from pLenti4/TO/V5-DEST (Invitrogen). In brief, the original attR1 site to V5 epitope region (nt 2405--4203) in pLenti4/TO/V5-DEST was replaced with an in-frame DNA fragment encoding Kozak sequence, ATG, FLAG tag and a rare cutter CPO I site (5′CGGTCCG). Accordingly, the cDNAs of EBV Rta (M-ABA strain) and KSHV RTA (Genebank: U71367.1) were PCR-amplified with CPO I sites flanking at both ends, and subcloned into pLenti4-Flag-CPO. The resulting plasmids, namely pLenti4-Flag-ER and pLenti4-Flag-KR, were propagated in DH5α and used in further studies. The upstream sequences of p21 (2.4 kb), EBV BGLF5 (nt 108641 to 110053 of NC\_007605), K-RTA (nt 70240 to 71597 of U75698), PAN (nt 28159 to 28660), ORF57 (nt 81556 to 82005), and K-bZIP (74619 to 74849), were cloned in front of luciferase gene located in pGL3-Basic (Promega, Madison, WI), yielding pGL3-Basic-p21p, -BGLF5p, K-RTAp, -PANp, -ORF57p and -K-bZIPp, respectively. Transfection and luciferase reporter assay {#s4c} ------------------------------------------ Transfection was performed in 24-well plates. The next day when the cultured 293 cells were 90% confluent, appropriate amount of indicated plasmids were transfected into cells by using Lipofectamine™ 2000 (Invitrogen) according to the manufacturer\'s instructions. Twenty-four hr after transfection, cells were harvested for luciferase activity assay by using Dual-Glo lucifearse assay kit (Promega). In addition, an aliquot of cell lysates was subjected to western blot analysis for the normalization of each transfection efficiency. Titration of EBV and KSHV viral particles {#s4d} ----------------------------------------- Filtrated (0.45 µm) viral supernatant (160 µl) was incubated with 2 U DNase I (Invitrogen) at 37°C for 30 min followed by extraction of encapsidated EBV DNA using QIAamp MinElute virus spin kit (QIAGEN). Each comparative quantitative PCR reaction was composed of 4 µl diluted viral DNA, 5 µl Power SYBR Green Master Mix (Applied Biosystems, Foster City, CA), and 1 µl primer mix (2 µM). The primers used in the present study were as follows: detection of EBV genome, BALF5-forward (5′-CGGAGTTGTTATCAAAGAGGC-3′) and BALF5-reverse (5′-CGAGAAAGACGGAGATGGC-3′); detection of KSHV genome, ORF9-forward (5′-CCAACATCATCCAATGCCTC-3′) and ORF9-reverse (5′-GGGAAAAGTCACGGGAATG-3′). Known copy numbers of serially diluted EBV genome from Raji cellular DNA (50 copies/cell) were used as standards in titrating EBV viral particles. Known copy numbers of serially diluted cosmid GB11 DNA encompassing KSHV genome nt 1--35,022 (U75698) were used as standards in titrating KSHV viral particles. The reaction was conducted and detected by StepOnePlus™ Real-Time PCR system (Applied Biosystems). Infectivity assay of KSHV particles {#s4e} ----------------------------------- We followed the EBV infection procedure described by Hutt-Fletcher and colleagues with minor modification [@pone.0017809-Turk1]. Specifically, 293 cells were seeded onto 12-well plates at 1.2×10^5^ cells/well that produced ≅30% confluent monolayer 24 h later. Two hundred-µl undiluted, filtrated viral supernatant was gently applied onto the surface of cells. After 2 h of incubation on cells, 1.5 ml growth medium was added and the cells were reincubated for 48 h. To score the infectious units in each well, the culture supernatant was removed, cells were trypsinized and subjected to visual inspection for GFP expression under a fluorescence microscope (OLYMPUS BX51, Olympus UK Ltd, Essex SS2 5QH, UK). Western blot analysis {#s4f} --------------------- Cell lysates extracted by RIPA buffer were subjected to SDS-PAGE separation and transferred onto nitrocellulose membranes. The membranes were blocked for 1 h in 1× TBST containing 5% non-fat milk and then incubated with the indicated primary antibody overnight at 4°C. The blots were washed three times with 1× TBST for 5 min each. The blots were incubated with peroxidase-conjugated secondary antibody in blocking buffer for 1 h at room temperature. Blots were washed three times with 1× TBST for 5 min each and developed by SuperSignal West Pico chemiluminescent substrate kit (Pierce). Cell proliferation assay {#s4g} ------------------------ Cell proliferation was determined by using a WST-1 kit (Roche, Indianapolis, IN) or by trypan blue exclusion method as described previously [@pone.0017809-Chen1]. Immunofluorescence assay {#s4h} ------------------------ Cells were resuspended in PBS and dropped onto multiple-well diagnostic microscope slides and fixed in methanol/acetone (1∶1) at −20°C for 20 min. Cells were permeabilized with 0.1% Triton-X 100 at room temperature for 20 min. The slide was incubated with indicated primary antibody at room temperature for 1 h, washed three times in PBS for 5 min each, and incubated with FITC-conjugated secondary antibody at room temperature for 1 h. After PBS wash, the slide was incubated with Hoechst 33258 at room temperature for 20 min, washed with PBS, mounted in VECTASHIELD ™medium and inspected by fluorescence microscopy. To quantitate the percentage of positively immuno-reactive cells in the immunofluorescence assay, an aliquot of cells were analyzed in parallel by flow cytometric analysis. Flow cytometric analysis {#s4i} ------------------------ Cells were harvested by centrifugation, washed with phosphate-buffered saline (PBS), fixed in ice-cold 75% ethanol and stored at −20°C until all samples from different time points were collected. Of note, to quench the green and red fluorescence in ERKV cells, the fixation reagent was replaced with 95% methanol. Prior to flow cytometer analysis, the fixed cells were repelleted by centrifugation, permeabilized in PBS containing 0.1% Triton X-100 at room temperature for 30 min, and resuspended in PBS containing 50 µg/ml propidium iodide and 50 µg/ml RNaseA. After the cells were incubated in dark for 30 min, cell cycle profile analysis was carried out on 5,000 cells with a fluorescence activated cell sorter (FACSCalibur, Becton Dickinson, Franklin Lakes, NJ). The results were analyzed by using WinMDI v2.8 software. Antibodies {#s4j} ---------- Mouse monoclonal antibodies of EBV proteins were: Rta (467), BZLF1 (4F10), BMRF1 (88A9), BALF4/gB (L2), BHRF1 (3E8), and gp350/220 (72A1). Anti-KSHV RTA was provided by Dr. Keiji Ueda (Osaka University Medical School, Japan). Anti-KSHV K-bZIP was provided by Dr. Mengtao Li (University of Kentucky College of Dentistry, USA). All other antibodies were commercially available: KSHV K8.1 (ABI, Columbia, MD); CDK6, pRb/S807/S811 and p21 (Cell Signaling Technology, Danvers, MA); c-Myc (Santa Cruz Biotechnology, Santa Cruz, CA); 14-3-3σ (GeneTex, Irvine, CA); CCND2 (BD Pharmingen, Franklin Lakes, NJ); α-tubulin (Millipore, Billerica, MA); β-actin and M2-FLAG (Sigma-Aldrich, St. Louis, MO). We are indebted to Ching-Hwa Tsai, Mei-Ru Chen (National Taiwan University, Taiwan) and Shih-Tung Liu (Chang Gung University, Taiwan) for providing various antibodies of EBV proteins; Dr. Keiji Ueda (Osaka University Medical School, Japan) for anti-K-RTA. Dr. Mengtao Li (University of Kentucky College of Dentistry, USA) for anti-K-bZIP; Jeffrey Vieira (University of Washington, Seattle, USA) for rKSHV.219. We also acknowledge Mr. Shu-Wei Nien for technical support on microarray analysis, and the array services provided by Microarray Core Laboratory of National Health Research Institutes, Taiwan. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This study was supported by Taiwan NHRI CA-099-PP-17 and Department of Health DOH99-TD-C-111-004 to S.-F. Lin; NHRI CA-099-PP-13, National Science Council NSC98-3112-B-400-002, and NSC99-3112-B-400-009 to J.-Y. Chen. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: YJC WHT JYC SFL. Performed the experiments: YJC WHT YLC YCK SPC. Analyzed the data: YJC WHT YLC YCK JYC SFL. Wrote the paper: YJC SFL.

PubMed Central

2024-06-05T04:04:19.865394

2011-3-10

PMC3053392

Introduction {#s1} ============ Vascular smooth muscle cells (VSMCs) have enormous applications in regenerative medicine [@pone.0017771-Matsubayashi1], [@pone.0017771-Niklason1], [@pone.0017771-Oberpenning1]. Studies have demonstrated that smooth muscle-like cells (SMLCs) can be derived from bone marrow-[@pone.0017771-Kashiwakura1], [@pone.0017771-Ross1], adipose-[@pone.0017771-Rodrguez1], [@pone.0017771-Kim1] and umbilical cord blood-derived stem cells [@pone.0017771-LeRicousseRoussanne1]. Due to the easy expansion, human embryonic stem cells (hESCs) represent an alternative source of VSMCs particularly for old patients having stem cells with impaired function. Recent studies reported different strategies to differentiate hESCs into SMLCs by exposing a monolayer of undifferentiated hESCs to retinoic acid [@pone.0017771-Huang1] or a combination of cell culture medium and extracellular matrix environment [@pone.0017771-Xie1], [@pone.0017771-GerechtNir1], [@pone.0017771-Ferreira1] either in single-hESC- [@pone.0017771-Vo1], embryoid bodies (EBs)- [@pone.0017771-Ferreira1] or stromal cell- [@pone.0017771-Hill1] culture conditions. In one case, SMLCs transplanted subcutaneously in an animal model were able to contribute for the formation of functional blood microvessels [@pone.0017771-Ferreira1]. Despite these advances, several issues remain poorly understood: (i) what hESC-derived population has the most SMC potential, (ii) the bioactive molecules involved in the differentiation process, (iii) the modulatory effect of 3D environments in SMLCs, (iv) the functionality of the differentiated SMLCs, and (v) the level of organization of the contractile protein filaments. Here we evaluate the smooth muscle cell (SMC) differentiation of different cell populations isolated from human embryoid bodies grown in suspension for 10 days. The isolated cells were cultured in media supplemented with several inductive signals, including platelet-derived growth factor (PDGF~BB~), retinoic acid (RA), transforming growth factor beta 1 (TGF~β-1~) or a combination of PDGF~BB~ with TGF~β-1~. We show that CD34^+^ cells have higher SMC potential than CD34^−^ cells and PDGF~BB~ and RA are the most effective agents to drive the differentiation of hESCs into smooth muscle progenitor cells (SMPCs). We further demonstrate that these cells contract and relax in response to SMC agonists or inhibitors, respectively, and the effect is mediated by Rho A/Rho kinase- and Ca^2+^/CaM/MLCK-dependent pathways. In addition, cells encapsulated in 3D gel scaffolds further differentiate towards SMC lineage as confirmed by gene analysis. Finally, we show that Endothelin-1 induces the organization of the contractile protein filaments. Materials and Methods {#s2} ===================== An expanded [Materials and Methods](#s2){ref-type="sec"} section is provided in the online data supplement ([Materials and Methods S1](#pone.0017771.s012){ref-type="supplementary-material"}). hESC culture and embryoid body (EB) formation {#s2a} --------------------------------------------- Undifferentiated hESCs (passages 27--62; H9, WiCell, Wisconsin, <http://www.wicell.org/>) were grown on an inactivated mouse embryonic fibroblast (MEF) feeder layer, as previously described [@pone.0017771-Ferreira1]. To induce the formation of EBs, the undifferentiated hESCs were treated with 2 mg/mL type IV collagenase (Invitrogen, <http://www.invitrogen.com>) for 2 h and then transferred (2∶1) to low attachment plates (Corning, <http://www.corning.com>) containing 10 mL of differentiation medium \[80% KO-DMEM, 20% fetal bovine serum (FBS, Invitrogen), 0.5% L-glutamine, 0.2% β-mercaptoethanol, 1% nonessential amino acids and 50 U/ml∶50 µg/ml penicillin-streptomycin solution\]. EBs were cultured for 10 days at 37°C, 5% CO~2~ in a humidified atmosphere, with media changes every 3--4 days. Isolation and differentiation of CD34^+^, CD34^−^ and CD34^+^KDR^−^ cells {#s2b} ------------------------------------------------------------------------- CD34^+^ cells were isolated from EBs at day 10 according to a protocol previously reported by us [@pone.0017771-Ferreira1]. For some experiments, the CD34^+^ cells were further separated in CD34^+^KDR^−^ cells. In this case, cells were labeled with anti-VEGF R2/KDR-PE antibody (R&D, <http://www.rndsystems.com/>), then conjugated with anti-PE antibody coupled with magnetic beads, and finally the magnetically labeled cells were separated into CD34^+^KDR^+^ and CD34^+^KDR^−^ using a MS-MACS column (Miltenyi Biotec, <http://www.miltenyibiotec.com>). Isolated cells were grown on 24-well plates (1.5×10^4^ cells/cm^2^) coated with 0.1% gelatin and containing one of the following media: smooth muscle growth medium-2 (SMGM-2), endothelial growth medium-2 (EGM-2) or EGM-2 supplemented with PDGF~BB~ (50 ng/mL, Prepotech, <http://www.peprotech.com/>) or RA (1 µM, Sigma, <http://www.sigmaaldrich.com>) or TGF~β-1~ (10 ng/mL, Prepotech) or a mixture of PDGF~BB~ with TGF~β-1~ (50 ng/mL; 10 ng/mL). Human vascular smooth muscle cells (hVSMCs, isolated from the arteries of human umbilical cord, Lonza, <http://www.lonza.com>) were used as controls for the differentiation studies. Cells were cultured in SMGM-2 media (Lonza; until passage 6) being the medium changed every 2 days. Immunofluorescence analysis {#s2c} --------------------------- Cells were transferred to gelatin-coated slides containing differentiation medium, allowed to attach overnight, and then fixed with 4% paraformaldehyde (Electron Microscopy Sciences, <http://www.emsdiasum.com/microscopy>) for 15 min at room temperature. Cells were blocked with 1% (w/v) BSA and stained for 1 h with anti-human primary antibodies specific for smooth muscle α-actin (α-SMA, 1A4, Dako, <http://www.dako.com/>), smooth muscle myosin heavy chain (SM-MHC, SMMS-1, Dako) and calponin (CALP, Calponin1, Santa Cruz Biotec, <http://www.scbt.com/>). In each immunofluorescence experiment, an isotype-matched IgG control was used. Binding of primary antibodies to specific cells was detected with anti-mouse IgG Cy3 conjugate or anti-mouse IgG FITC (both form Sigma). Cell nuclei were stained with 4′, 6′-diamidino-2-phenylindole (DAPI) (Sigma) and the slides examined with either a Zeiss fluorescence microscope or Zeiss LSM 50 confocal microscope. Reverse transcription-polymerase chain reaction (RT-PCR) analysis {#s2d} ----------------------------------------------------------------- Total RNA from experimental groups was isolated using a protocol with TRIzol (Invitrogen) and RNeasy Minikit (Qiagen, Valencia, <http://www1.qiagen.com/>). cDNA was prepared from 1 µg total RNA using Taqman Reverse transcription reagents (Applied Biosystems, CA). Quantitative PCR (qPCR) was performed using Power SYBR Green PCR Master Mix and the detection using an ABI PRISM 7500 System (Applied Biosystems, <http://www3.appliedbiosystems.com>). Quantification of target genes was performed relative to the reference GAPDH gene: relative expression = 2^\[−(Ct^ ~sample~ ^−Ct^ ~GADPH~ ^)\]^. The mean minimal cycle threshold values (Ct) were calculated from quadruplicate reactions. Then, the relative gene expression in each experimental group was normalized to the relative gene expression found in hVSMCs. Primer sequences are published as supporting information (**[Table S1](#pone.0017771.s011){ref-type="supplementary-material"}**). For the RT^2^ Profiler™ PCR Array, cDNA was prepared from 1 µg total RNA using the RT^2^ PCR Array first strand kit (SABiosciences, <http://www.sabiosciences.com/>). RT-PCR assays were performed using the human extracellular matrix and adhesion molecules RT^2^ Profiler™ PCR Array (SABiosciences) on an ABI PRISM 7500 System. Data analysis was performed using analysis software provided by the kit manufacturer. Intracellular Ca^2+^ variation measurements {#s2e} ------------------------------------------- SMCs or hESC-derived cells were loaded with Fura-2 calcium fluorescent indicator [@pone.0017771-Grynkiewicz1] by incubation with 5 µM of the membrane permeable acetoxymethyl (AM) derivative FURA-2/AM (1 mM in DMSO, Molecular Probes, Invitrogen) and 0.06% (w/v) Pluronic F-127 (Sigma), using basal medium (M199, Sigma) as a vehicle (35 µl/well, not supplemented with serum nor antibiotics), for 1 h at 37°C in 5% CO~2~ and 90% humidity. Cells were then stimulated with 100 µM histamine (Sigma) [@pone.0017771-Bernardino1], 10^−7^ M bradykinin (Calbiochem, <http://www.merck-chemicals.com/>) [@pone.0017771-Drab1], 10^−5^ M angiotensin II (Calbiochem) [@pone.0017771-Drab1], 10^−5^ M carbachol (AlfaAesar) [@pone.0017771-Rodrguez1] or 50 mM KCl (Merck, <http://www.merck-chemicals.com>) [@pone.0017771-Bernardino1] by adding 1 µl of a stock solution. A detailed methodology for the fluorescence acquisition can be found on the online data supplement ([Materials and Methods S1](#pone.0017771.s012){ref-type="supplementary-material"}). Contractility assays {#s2f} -------------------- Agonist-induced contractile activity of the differentiated cells was evaluated as previously described [@pone.0017771-Rodrguez1], [@pone.0017771-Ferreira1]. hECS-derived cells cultured for 3 passages were washed with DMEM and contraction was induced by incubating these cells with 10^−5^ M carbachol in DMEM (Sigma) medium for 30 min. Contraction was calculated as the difference in cell area between time zero and 30 minutes. The same microscopic fields were imaged before and after treatment for contraction analysis. In a distinct experiment, cell relaxation was induced by incubation with 10^−4^ M atropine (AlfaAesar) in DMEM for 1 h followed by contraction with 10^−5^ M carbachol. Contraction was calculated as before. hVSMCs (3rd passage) were used as controls. Effect of RhoA/Rho kinase- and CaM/MLCK- agonists in cell contraction and maturation {#s2g} ------------------------------------------------------------------------------------ hVSMCs or hESC-derived SMPCs were treated with the inhibitor W-7 (12 µg/mL, Sigma) for 30 min and the agonist U46619 (1 µM, a CAM kinase-agonist) (Cayman Chemicals, <http://www.caymanchem.com>) in serum-free M199 medium for 3 days. At the end, cells were characterized for the expression of SMC markers by immunostaining. Finally, the contraction of U46619-treated cells was evaluated by embedding the cells in fibrin gels (3.5×10^4^ cells/50 µL) and measuring gel size at time 0 and 14 h by microscopy. This methodology was repeated to evaluate the effect of Rho kinase-agonist in cell contraction and maturation. In this case, the cells were exposed to the inhibitor Y27632 (13 µM) (Cayman Chemicals) and the agonist endothelin-1 (End-1, 10 nM, Sigma). Cytokine measurements {#s2h} --------------------- Cell culture supernatants were assayed for cytokines using a Bio-plex human 17-plex panel immunoassay kit (Bio-Rad, <http://www.bio-rad.com>) and cytokines concentrations were determined using Bio-Plex Manager 5, according to manufacturer\'s instructions. Samples and controls were run in triplicate, standards and blanks in duplicate. A detailed methodology can be found on the online data supplement (**[Materials and Methods S1](#pone.0017771.s012){ref-type="supplementary-material"}**). Co-culture of hESC-derived SMPCs with hVSMCs {#s2i} -------------------------------------------- To evaluate the effect of both soluble and insoluble factors, fluorescently-labeled hESC-derived SMPCs (stained with PKH67 dye, Sigma) were plated on top of mytomycin-inactivated hVSMC cultured on a 24 well plate for 2 days before use. This co-culture system was maintained for 5 days after which the overall cells were tripsinized, and fluorescent (PKH67^+^ cells) cells sorted by a FACS Aria (BD). The sorted cells were characterized by immunohistochemistry to evaluate the expression and organization of α-SMA, SM-MHC and calponin filaments. Culture of SMPCs in three-dimensional gels {#s2j} ------------------------------------------ Fibrin gels were obtained by the crosslinking of 20 mg/mL fibrinogen/TBS pH 7.4 in the presence of 50 U/mL thrombin/TBS pH 7.4 (both from Sigma-Aldrich). Fibrin gels (50 µL) were prepared by mixing the following components: 10 mg/mL fibrinogen, 2.5 mM CaCl~2~, 2 U/mL thrombin and 0.01 mg/mL aprotinin (Sigma-Aldrich). This solution was allowed to gel at 37°C in 100% relative humidity. hVSMC or hESC-derived SMPCs (3×10^5^) were encapsulated in fibrin gels (50 µl). Cells were centrifuged and resuspended in the fibrin gel precursor solution and included in 1 mL sterile syringes with cut tips. Polymerization was initiated at 37°C and allowed to proceed over 30 min. After polymerization, the cell constructs were removed from the syringe and placed in 24-well plates, containing specific medium, for up to 96 h. Statistical analysis {#s2k} -------------------- An unpaired *t* test or one-way ANOVA analysis of variance with Bonferroni post test was performed for statistical tests using SigmaStat. Results were considered significant when *P*\<0.05. Data are shown as mean ± SEM. Results {#s3} ======= Effect of initial cell population and inductive signals on cell proliferation {#s3a} ----------------------------------------------------------------------------- We evaluated the SMC differentiation potential of three cell populations isolated from EBs at day 10: CD34^+^ cells, CD34^−^ cells and CD34^+^KDR^−^ cells (mesenchymal origin [@pone.0017771-Vodyanik1]) ([**Figure 1A**](#pone-0017771-g001){ref-type="fig"}). The percentages of CD34^+^ and CD34^+^KDR^−^ cells in EBs were approximately 2% and 1.7%, respectively, after MACS isolation. As evaluated by flow cytometry (**[Materials and Methods S1](#pone.0017771.s012){ref-type="supplementary-material"}**), the purity of the cell populations CD34^+^, CD34^−^, and CD34^+^KDR^−^ was \>80%, \>98% and \>98%, respectively ([**Figure 1B**](#pone-0017771-g001){ref-type="fig"}). These cells were plated on gelatin coated dishes, at low density (1.5×10^4^/cm^2^), and cultured on basal media (EGM-2 or SMGM-2) supplemented or not with different inductive signals to guide their SMC differentiation: PDGF~BB~ (50 ng/mL) [@pone.0017771-Ross1], [@pone.0017771-Ferreira1], [@pone.0017771-Simper1], RA (1 µM) [@pone.0017771-Huang1], [@pone.0017771-Drab1], [@pone.0017771-Sinha1], TGF~β-1~ (10 ng/mL) [@pone.0017771-Ross1], [@pone.0017771-Sinha2], [@pone.0017771-Rensen1] or a combination of TGF~β-1~ (10 ng/mL) with PDGF~BB~ (50 ng/mL) ([**Figure 1A**](#pone-0017771-g001){ref-type="fig"}). These concentrations have been used previously in the differentiation of stem cells from different origins into vascular cells [@pone.0017771-Ross1], [@pone.0017771-Huang1], [@pone.0017771-Sinha2]. ::: {#pone-0017771-g001 .fig} 10.1371/journal.pone.0017771.g001 Figure 1 ::: {.caption} ###### The effect of initial cell population and inductive signals on cell proliferation. \(A) Protocols adopted to drive the differentiation of CD34^+^, CD34^+^KDR^−^ and CD34^−^ cells isolated from EBs at day 10 into the SMC lineage. (B) Flow cytometric analysis of hES-derived cells: CD34^−^ (B.1), CD34^+^ (B.2) and CD34^+^KDR^−^ (B.3) cells (in this last case isolated from the CD34^+^ cell population). The results show that CD34^−^ cells do not express CD34^+^ marker (B.1), CD34^+^ cells are formed by CD34^+^KDR^−^ and CD34^+^KDR^+^ cells (B.2), and CD34^+^KDR^−^ do not express the KDR marker (B.3). Percent of positive cells (dash plot) were calculated based in the isotype controls (grey plot). (C) Time-course proliferation of CD34^+^ (C.1), CD34^+^KDR^−^ (C.2) and CD34^−^ cells (C.3). :::  ::: CD34^+^ cells cultured on EGM-2 medium supplemented with PDGF~BB~ presented the highest proliferation (more than 8 population doublings over 20 days), followed by the ones cultured in EGM-2 medium supplemented with RA ([**Figure 1C.1**](#pone-0017771-g001){ref-type="fig"}). CD34^+^ cells grown in EGM-2 medium without PDGF~BB~ or RA proliferated poorly over 40 days ([**Figure 1C.1**](#pone-0017771-g001){ref-type="fig"}). Interestingly, CD34^+^ cells cultured in medium supplemented with TGF~β-1~ and PDGF~BB~ had low proliferation suggesting that TGF~β-1~ inhibited the effect of PDGF~BB~. CD34^+^KDR^−^ cells cultured on EGM-2 medium without any supplements proliferated extensively, showing more than 8 population doublings over 20 days ([**Figure 1C.2**](#pone-0017771-g001){ref-type="fig"}). Similar proliferation potential was observed for CD34^+^KDR^−^ cells grown in EGM-2 medium supplemented with RA, but not on medium supplemented with PDGF~BB~. The proliferation rate of CD34^−^ cells was also assessed in the media formulations tested for CD34^+^ and CD34^+^KDR^−^ cells. CD34^−^ cells cultured on EGM-2 medium supplemented with PDGF~BB~ showed the highest proliferation, having more than 8 population doublings over an 18 days period ([**Figure 1C.3**](#pone-0017771-g001){ref-type="fig"}). In contrast, cells grown in EGM-2 medium supplemented with RA did show a poor proliferation over more than 20 days. All taken together, the proliferation potential of the cells was dependent on the initial cell population and the supplements added to the basal media. Effect of initial cell population and inductive signals on SMC differentiation {#s3b} ------------------------------------------------------------------------------ Next we evaluated the expression of SMC-specific genes in the hESC-derived cells (passage 4) by quantitative RT-PCR ([**Figure 2**](#pone-0017771-g002){ref-type="fig"}). The genes analyzed included: α-smooth muscle actin (α-SMA), an early marker of SMC differentiation and highly specific marker for SMCs in adult animals [@pone.0017771-Frid1]; smooth muscle myosin heavy chain (SM-MHC), a later marker in SMC differentiation that seems to be highly restricted to SMCs [@pone.0017771-Owens1]; calponin and smooth muscle α-22, definitive SMC markers [@pone.0017771-Duband1]. The gene expression in the hESC-derived cells was normalized by the corresponding gene expression in hVSMCs. ::: {#pone-0017771-g002 .fig} 10.1371/journal.pone.0017771.g002 Figure 2 ::: {.caption} ###### Gene expression in hESC-derived cells evaluated by qRT-PCR. Gene expression in each experimental group was normalized by the corresponding gene expression observed in hVSMCs. Oct-4 was normalized by the expression in undifferentiated hESCs. (A) Gene expression of hESCs, CD34^+^ and CD34^−^ cells before differentiation. SMα-22 expression in CD34^−^ cells is very low (\<8.5×10^−6^) and not visible in the graph. CD34^+^ (B), CD34^+^KDR^−^ (C), and CD34^−^ (D) cells were cultured in SMGM-2 medium, EGM-2 medium, and EGM-2 medium supplemented with PDGF~BB~ or TGF~β-1~ or RA or TGF~β-1~ plus PDGF~BB~. Cells were characterized at passage 4 (≈20 days). Results are Mean ± SEM (*n* = 4); \* denotes statistical significance (*P*\<0.05). :::  ::: Undifferentiated CD34^+^ cells expressed low levels of SM-MHC (∼2.0%), SMα-22 (\<1%) and calponin (∼2.0%), in most cases comparable to the levels found in undifferentiated hESCs, and moderate levels of α-SMA (∼10%) ([**Figure 2**](#pone-0017771-g002){ref-type="fig"}). Culture of these cells in the presence of EGM-2 medium supplemented with PDGF~BB~ or RA contributed for the up-regulation of SMC genes, as confirmed by the significant increase in the expression of α-SMA (\>400-fold) and SM-MHC (\>1,600-fold) when compared with undifferentiated CD34^+^ cells (p\<0,05), and hVSMCS (α-SMA: \>39-fold; SM-MHC: \>27-fold) ([**Figure 2**](#pone-0017771-g002){ref-type="fig"}). Addition of TGF~β-1~, or TGF~β-1~ plus PDGF~BB~ increased the cellular expression of α-SMA, SM-MHC, SMα-22 and calponin when compared to undifferentiated CD34^+^ cells; however, the expression was one order of magnitude lower than the one in hVSMCS, suggesting that the CD34^+^-derived cells were not fully differentiated into SMCs. CD34^+^KDR^−^ cells cultured in SMGM-2 medium or EGM-2 medium in the presence or absence of PDGF~BB~ expressed higher levels of α-SMA (\>42-fold), SM-MHC (\>150-fold), SMα-22 (\>18-fold) and calponin (\>2-fold) than undifferentiated CD34^+^ cells ([**Figure 2**](#pone-0017771-g002){ref-type="fig"}). In addition, these cells showed similar levels of expression of α-SMA and SM-MHC to the one found in hVSMCs, suggesting that these cells shared features with SMCs. In contrast to the CD34^+^ and CD34^+^KDR^−^ cells, CD34^−^ cells cultured in several media formulations had lower expression of SMC markers, indicating low efficiency of differentiation into SMCs ([**Figure 2**](#pone-0017771-g002){ref-type="fig"}). Of note, all the hESC-derived cells previously described had low expression of Oct-4 confirming their loss of pluripotency. In addition, hESC-derived cells expressing high levels of SMC markers had no expression of PECAM-1 (endothelial cell marker) and α-actinin (a marker of cardiomyocytes) (data not shown) confirming again their SMC differentiation. Next, the expression and filament organization of contractile proteins was evaluated in cell populations having similar or higher α-SMA and SM-MHC gene expression than hVSMCs, i.e., CD34^+^RA, CD34^+^PDGF~BB~, CD34^+^KDR^−^PDGF~BB~ and CD34^+^KDR^−^EGM-2. All hESC-derived cells stained positive for α-SMA, SM-MHC and calponin (**[Figure S1](#pone.0017771.s001){ref-type="supplementary-material"} and [Figure S2](#pone.0017771.s002){ref-type="supplementary-material"}**). More than 70% of the cells expressed α-SMA as evaluated by FACS analyses (**[Figure S3](#pone.0017771.s003){ref-type="supplementary-material"}**). Individualized calponin filaments were observed in 16 to 60% of the overall cells; however, organized α-SMA protein filaments were only observed in CD34^+^PDGF~BB~ (6.0%) and CD34^+^KDR^−^EGM-2 (6.1%) cells, and no organized SM-MHC filaments were observed in the hESC-derived populations (**[Figure S3](#pone.0017771.s003){ref-type="supplementary-material"}**). In contrast, hVSMCs expressed high levels of organized α-SMA (84.6%), SM-MHC (94.6%) and calponin (80.1%) filaments. Collectively, gene and protein analyses indicate that several inductive signals (PDGF~BB~, or RA or EGM-2 basal medium) are able to drive the SMC differentiation of CD34^+^ or CD34^+^KDR^−^ cells, characterized by variable expression of SMC proteins and minimal assembly of the α-SMA protein into filaments. Some hESC-derived cells present a secretion profile similar to hVSMCs {#s3c} --------------------------------------------------------------------- Using the multiplex cytometric bead array method we investigated the secretion of 17 analytes by hVSMC, CD34^+^PDGF~BB~, CD34^+^RA, CD34^+^KDR^−^PDGF~BB~, CD34^+^KDR^−^EGM-2 and CD34^−^PDGF~BB~ cells. Out of the 17 analytes analyzed, hVSMCs secreted IL-6, IL-7, IL-8, G-CSF, IFN-γ, MCP-1 (MCAF) and TNF-α, above the detection limit (\>0.2 pg/mL) ([**Figure 3**](#pone-0017771-g003){ref-type="fig"}). Cytokines IL-6 and IL-8 were the most secreted cytokines. CD34^+^PDGF~BB~ and CD34^+^KDR^−^EGM-2 cells secreted the same analytes as hVSMCs. CD34^+^KDR^−^PDGF~BB~ and CD34^−^PDGF~BB~ cells did not secrete IL-7, while CD34^+^RA cells secreted all seven cytokines mentioned and 4 cytokines more (IL-1β, IL-2, IL-4 and GM-CSF) (**[Figure S4](#pone.0017771.s004){ref-type="supplementary-material"}**). ::: {#pone-0017771-g003 .fig} 10.1371/journal.pone.0017771.g003 Figure 3 ::: {.caption} ###### Secretomic profile of hVSMC and hECS-derived cells. Seventeen cytokines were measured simultaneously in medium collected from hVSMC, CD34^+^PDGF~BB~, CD34^+^RA, and CD34^−^PDGF~BB~ cells at passage 4. Results are Mean ± SEM (*n* = 3). :::  ::: Interestingly, the secretion profile of CD34^+^PDGF~BB~ cells is very similar to the one observed for hVSMCS, either in the type and concentration of the analytes secreted ([**Figure 3**](#pone-0017771-g003){ref-type="fig"}). In contrast, CD34^−^PDGF~BB~ cells secreted analytes at different concentration than hVSMCs, and the other cell populations seemed to be in an intermediate stage in terms of secretion profile ([**Figure 3**](#pone-0017771-g003){ref-type="fig"} **and [Figure S4](#pone.0017771.s004){ref-type="supplementary-material"}**). CD34^+^RA cells secreted higher levels of cytokines than all the remaining cell populations, including hVSMCs. Some hESC-derived cells like hVSMCs have the ability to contract when treated with vasoactive agonists and the process is mediated by Ca^2+^ {#s3d} -------------------------------------------------------------------------------------------------------------------------------------------- The ability of SMCs to contract in response to vasoactive agonists is mediated by an increase of intracellular Ca^2+^ levels which triggers the SMC contractile apparatus [@pone.0017771-Hathaway1]. To evaluate whether hESC-derived cells had contractility mediated by Ca^2+^, the cells were loaded with FURA-2, a Ca^2+^ sensitive dye, and their response to vasoactive agonists (bradykinin, angiotensin II, carbachol and histamine) and depolarization agents (KCl) was monitored by fluorescence ([**Figure 4A**](#pone-0017771-g004){ref-type="fig"}). The response profile was compared to the one observed for hVSMCs and human umbilical vein endothelial cells (HUVECs), as positive and negative controls, respectively. Intracellular free Ca^2+^ \[Ca^2+^\]~i~ levels increase when hVSMCs are exposed to bradykinin, angiotensin II and carbachol, while no measurable effect is observed in HUVECs. KCl induces a higher increase in the \[Ca^2+^\]~i~ levels in HUVECs than in hVSMCs, while histamine induces similar levels of \[Ca^2+^\]~i~ in both cell types but following different profiles. These response patterns are typical for HUVECs and hVSMCs [@pone.0017771-Dora1], [@pone.0017771-Takeda1], [@pone.0017771-Graier1], [@pone.0017771-Montiel1], [@pone.0017771-Kansui1]. ::: {#pone-0017771-g004 .fig} 10.1371/journal.pone.0017771.g004 Figure 4 ::: {.caption} ###### Contractility of hESC-derived cells. \(A) Concentration of intracellular Ca^2+^ in FURA-2-loaded cultured hESC-derived cells in response to several agonists (100 µM histamine, 10^−7^ M bradykinin, 10^−5^ M angiotensin II or 50 mM KCl). hVSMCs and HUVECs were used as positive and negative controls, respectively. Traces are representative of 4 independent experiments for each condition. (B) Contractility of hESC-derived cells when exposed to the effects of carbachol (10^−5^ M in DMEM medium, for 30 min) or atropine (10^−4^ M, 1 h) plus carbachol (10^−5^ M for 30 min). hVSMCs were used as controls. Results are Mean ± SEM (*n* = 3); ^\#^ denotes statistical significance (*P*\<0.05) in the same experimental group. :::  ::: CD34^+^RA and CD34^+^PDGF~BB~ cells exposed to histamine, bradykinin, angiotensin II, carbachol or KCl had similar response profiles as observed for hVSMCs ([**Figure 4A**](#pone-0017771-g004){ref-type="fig"} **and [Figure S5](#pone.0017771.s005){ref-type="supplementary-material"}**). For bradykinin and KCl the response intensity was lower than in hVSMCs, however similar intensity was observed for angiotensin II, carbachol and histamine. On the other hand, although CD34^+^KDR^−^RA, CD34^+^KDR^−^PDGF~BB~ and CD34^+^KDR^−^EGM-2 cells had similar response profiles as hVSMCs, in general the intensity of response was lower (**[Figure S6](#pone.0017771.s006){ref-type="supplementary-material"}**). In contrast to the previous hESC-derived populations, CD34^−^PDGF~BB~ cells showed a very low variation in the intracellular levels of Ca^2+^ when exposed to depolarization and vasoactive peptides. To examine whether hESC-derived cells were able to contract, they were subjected to the effects of carbachol and atropine ([**Figure 4B**](#pone-0017771-g004){ref-type="fig"}). With the exception of CD34^−^PDGF~BB~ cells, all cells were able to contract after exposure to carbachol (13 to 38% contraction after 30 min, depending on the cell population). In most cases, cell contraction was not significantly different (*P*\>0.05) from the one observed for hVSMCs. Moreover, with the exception of CD34^−^PDGF~BB~ cells, the muscarinic inhibitor atropin significantly blocked or reduced the carbachol-mediated effect (*P*\<0.05) ([**Figure 4B**](#pone-0017771-g004){ref-type="fig"}). Based on the expression of SMC proteins and genes, secretion of cytokines and cellular contraction, CD34^+^PDGF~BB~ and CD34^+^RA cells were selected for further characterization. These cells are likely at a smooth muscle progenitor cell stage (SMPCs), which can potentially be further induced into a more mature SMC phenotype. The contraction of both hESC-derived SMPCs and hVSMCs involves Rho A/Rho kinase- and Ca^2+^/CaM/MLCK-dependent pathways {#s3e} ----------------------------------------------------------------------------------------------------------------------- CaM/MLCK- and RhoA/Rho kinase-dependent pathways play a pivotal role in SMC contraction [@pone.0017771-Kim1], [@pone.0017771-Hathaway1]. The Ca^2+^/CaM pathway plays a key role in SMC contraction through the stimulation of MLCK-mediated phosphorylation of myosin light chain 20,000 Da (MLC~20~) [@pone.0017771-Hathaway1]. To assess whether Ca^2+^/CaM pathway was involved in the contraction of hESC-derived SMPCs, cells were exposed to the CaM-specific inhibitor W-7 [@pone.0017771-Frampton1], and then contraction was induced by exposing them to the CaM agonist U46619 [@pone.0017771-Kim1]. To facilitate the evaluation of cell contraction, cells were encapsulated in fibrin gels, and gel diameter was determined after 14 h. Pre-incubation of hESC-derived SMPCs with W-7 significantly inhibited U46619-induced contraction ([**Figure 5A**](#pone-0017771-g005){ref-type="fig"} **and** [**Figure 5B**](#pone-0017771-g005){ref-type="fig"}). Similar results were obtained for the control cells hVSMCs. Overall the results indicate the involvement of CaM/MLCK-kinase pathway in cell contraction. ::: {#pone-0017771-g005 .fig} 10.1371/journal.pone.0017771.g005 Figure 5 ::: {.caption} ###### Evaluation of the role of Rho A/Rho kinase- and Ca^2+^/CaM/MLCK-dependent pathways in the contraction of SMPCs and hVSMCs. CD34^+^PDGF or CD34^+^RA cells were initially treated for 3 days with serum-free medium containing the agonists endothelin-1 (End-1 (10 nM), A and B.1) or U46619 (1 µM) (A and B.2) in the presence or absence of the Rho-kinase inhibitor Y27632 (13 µM) or the calmodulin inhibitor W-7 (12 µg/mL). The cells were then encapsulated in fibrin gels and the diameter of the gel assessed at time 0 (1) and 14 h (2,3,4). Graph displays the variation in gel diameter (percentage) encapsulating hVSMCs (2), CD34^+^PDGF~BB~ cells (3) or CD34^+^RA cells (4). Data are shown as mean ± SEM (*n* = 6), \* *P*\<0.001 relatively to samples inhibited with Y27632 or W-7, by Student\'s *t*-test. :::  ::: To examine whether Rho kinase was involved in the hESC-derived SMPC contraction, the cells were pre-treated with the Rho kinase-specific inhibitor Y27632 [@pone.0017771-Kim1] and then contraction was induced by the agonist End-1 [@pone.0017771-Bouallegue1]. Cells treated with Y27632 and End-1 showed no significant contraction ([**Figure 5A**](#pone-0017771-g005){ref-type="fig"} **and** [**Figure 5B**](#pone-0017771-g005){ref-type="fig"}). In contrast, cells treated with End-1 but not Y27632 contracted significantly. Similar response profiles were obtained for hVSMCs and hESC-derived SMPCs. Collectively the results indicate the involvement of a Rho/Rho kinase-dependent pathway in SMC contraction. 3D culture of hESC-derived SMPCs modulates gene expression towards the expression observed in hVSMCs {#s3f} ---------------------------------------------------------------------------------------------------- We investigated the use of hESC-derived SMPCs for tissue engineering applications [@pone.0017771-Niklason1], [@pone.0017771-Oberpenning1]. Fibrin gels previously used for the encapsulation of neonatal SMCs [@pone.0017771-Ross2] were selected as scaffolds for the encapsulation of SMPCs. These gels allow cell attachment and can be remodeled by cellular metalloproteinases. SMPCs were encapsulated in fibrin gels for 3 days after which the cells were characterized at protein and gene levels. Gene expression of SMPCs (CD34^+^RA and CD34^+^PDGF) was compared to hVSMCs under the same culture conditions ([**Figure 6A**](#pone-0017771-g006){ref-type="fig"} **and** [**Figure 6B**](#pone-0017771-g006){ref-type="fig"}). The culture of SMPCs in 3D gels modulated the expression of SMC genes (α-SMA, SM-MHC or SMα-22) towards the one observed for hVSMCs cultured in 3D gels. We complemented these studies by evaluating the expression of extracellular matrix and adhesion molecules by a quantitative real-time PCR array. This array evaluated the expression of 84 genes involved in cell-cell and cell-matrix interactions. Again, the 3D culture of SMPCs modulated extracellular matrix and adhesion molecule genes towards the expression observed in hVSMCs ([**Figure 6B**](#pone-0017771-g006){ref-type="fig"}). The number of genes with similar expression increased from 23 to 58 or 9 to 53 when CD34^+^PDGF~BB~ cells or CD34^+^RA cells are cultured in 2D or 3D, respectively. Finally, gene expression associated with SMCs including collagen I and thrombospondin 1 [@pone.0017771-Majack1], integrins α2, α3, α5, αV and β1 [@pone.0017771-Deb1], the enzyme metalloproteinase 2 [@pone.0017771-Wang1], and the growth factor TGF~β-1~ was similar in SMPCs and hVSMCs ([**Figure 6C**](#pone-0017771-g006){ref-type="fig"} **and [Figure S7](#pone.0017771.s007){ref-type="supplementary-material"}**). ::: {#pone-0017771-g006 .fig} 10.1371/journal.pone.0017771.g006 Figure 6 ::: {.caption} ###### Characterization of hESC-derived SMPCs encapsulated in 3D fibrin gel scaffolds. SMPCs were characterized after being encapsulated in fibrin gels for 3 days. (A) SMC gene expression on CD34^+^RA (A.1) and CD34^+^PDGF (A.2) cells, as assessed by qRT-PCR. Gene expression of hESC-derived SMPCs was normalized by gene expression of hVSMCs under the same culture conditions. Results are Mean ± SEM (*n* = 4). (B) Comparison of extracellular matrix and adhesion molecules-related gene expression in CD34^+^PDGF~BB~ (B.1) or CD34^+^RA cells (B.2) with hVSMCs cultured in 2D (tissue culture polystyrene) or cultured in 3D fibrin gels. Gene expression was evaluated using a RT^2^ Profiler™ PCR array. (C) Normalized extracellular matrix and soluble factor gene expression of SMPCs relatively to hVSMCs, both cultured in 3D or 2D systems. :::  ::: Co-culture of hESC-derived SMPCs with hVSMCs induces the assembly of α-SMA and calponin into filaments {#s3g} ------------------------------------------------------------------------------------------------------ Next we sought to investigate whether hESC-derived SMPCs could be further matured into a SMC phenotype with an organized contractile filament network. To accomplish this goal, hESC-derived SMPCs were initially labeled with PKH67 fluorescent dye and co-cultured on top of hVSMCs for 5 days. After this time, the cells were sorted and characterized by immunocytochemistry. Remarkably, hESC-derived SMPCs showed significant improvement in the organization of the fibers after contact with hVSMCs: 36.3% and 41.2% of CD34^+^PDGF~BB~ and CD34^+^RA cells, respectively, had organized α-SMA fibers, while 64.8% and 73.8% had organized calponin fibers ([**Figure 7A**](#pone-0017771-g007){ref-type="fig"} **and** [**Figure 7B**](#pone-0017771-g007){ref-type="fig"}). These results show that hESC-derived SMPCs are plastic cells and can be induced to differentiate into a more mature SMC phenotype displaying an organized contractile network. ::: {#pone-0017771-g007 .fig} 10.1371/journal.pone.0017771.g007 Figure 7 ::: {.caption} ###### Maturation of hESC-derived SMPCs. (A and B) Expression and organization of SMC proteins on hESC-derived SMPCs cultured on top of inactivated-hVSMCs for 5 days. hVSMCs were used as controls. Bar corresponds to 50 µm. Results are Mean ± SEM (*n* = 10). (C and D) Expression and organization of SMC proteins on hESC-derived SMPCs treated with vasoactive agents for 3 days. Bar corresponds to 50 µm. Results are Mean ± SEM (*n* = 8); \*, \*\*, \*\*\* denote statistical significance (*P*\<0.05, *P*\<0.01, *P*\<0.001, respectively). :::  ::: Vasoactive agents U46619 and End-1 improve the organization of contractile protein fibers {#s3h} ----------------------------------------------------------------------------------------- Next, we identified molecules able to maturate the hESC-derived SMPCs into SMCs having an organized contractile network. We sought to evaluate the effect of the agonists U46619 and End-1 involved in the CaM/MLCK- and RhoA/Rho kinase-contraction pathways, respectively. In CD34^+^RA cells, each agonist independently improved dramatically the organization of α-SMA and calponin fibers, being End-1 the most effective. The addition of both agonists did not significantly improve the effect of End-1 ([**Figure 7C**](#pone-0017771-g007){ref-type="fig"} **and** [**Figure 7D**](#pone-0017771-g007){ref-type="fig"}). Importantly, the induction effect of the vasoactive agents was not observed on CD34^+^PDGF~BB~ ([**Figure 7D**](#pone-0017771-g007){ref-type="fig"}) or CD34^−^PDGF~BB~ cells (**[Figure S8](#pone.0017771.s008){ref-type="supplementary-material"}**) showing that the inductive effect is dependent on the differentiation history of the SMPCs. The mature SMCs (derived from CD34^+^RA) express SMC markers including α-SMA, SM-MHC and calponin but not the cardiac marker α-actinin (**[Figure S9](#pone.0017771.s009){ref-type="supplementary-material"}**). Furthermore, the cells show no expression of troponin T, a protein found in skeletal and cardiac muscle but not in smooth muscle and vascular endothelial-cadherin, a protein typically expressed in endothelial cells (**[Figure S10](#pone.0017771.s010){ref-type="supplementary-material"}**). Discussion {#s4} ========== In this study, we demonstrate that CD34^+^ cells have higher SMC potential than CD34^−^ cells. We further show that RA or PDGF~BB~ drive the differentiation of CD34^+^ cells into SMPCs We have characterized the differentiated cells at gene and at protein level, their secretome, the ability to contract when incubated with several pharmacological agents, and contraction mechanism mediated by Ca^2+^. Envisioning the use of these cells for vascular engineering, they were encapsulated on 3D fibrin scaffolds and characterized at gene and functional levels. Finally, we identified End-1 as a key molecule to induce the polymerization of contractile proteins in SMPCs (CD34^+^RA cells). We examined the capacity of three hESC populations (CD34^+^, CD34^+^KDR^−^, CD34^−^) isolated from EBs at day 10, and cultured as single cells on media supplemented with inductive signals including PDGF~BB,~ RA, TGF~β-1~ and TGF~β-1~ plus PDGF~BB~ to differentiate into SMCs. Notably, TGF~β-1~ [@pone.0017771-Ross1], [@pone.0017771-Sinha2], [@pone.0017771-Dickson1], PDGF~BB~ [@pone.0017771-Ross1], [@pone.0017771-Ferreira1], [@pone.0017771-Simper1], and RA [@pone.0017771-Huang1], [@pone.0017771-Drab1] have been reported to be very important inductive signals for SMC differentiation from different initial stem cell populations. We report that from all inductive signals tested in this study, RA and PDGF~BB~ are the most effective in guiding the differentiation of hESC-derived CD34^+^ cells into SMPCs, based on gene and protein analysis, response to the depolarization agents and vasoactive peptides, contraction profile, secretion of cytokines, and behavior in a 3D scaffold. Although SMPCs express most of the SMC markers, they exhibit disorganized contractile proteins. We further show that CD34^+^ cells have higher propensity to yield contractile SMPCs than CD34^−^ cells, when exposed to the same inductive signals. Interestingly, CD34^+^KDR^−^ cells which have been reported to be of mesenchymal origin [@pone.0017771-Vodyanik1], can give rise to SMPCs but they respond less efficiently to vasoactive peptides and depolarization agents, have lower contractile properties than CD34^+^PDGF cells, and have a different cytokine secretion profile as compared to hVSMCs. In line with previous results [@pone.0017771-Yamashita1], our results suggest that cells expressing KDR receptor are needed for an efficient SMC differentiation. The development of mature contractile SMCs from stem cells occurs in multiple steps comprising (i) the commitment to the SMC lineage, (ii) the differentiation into early immature and (iii) the maturation into the mature contractile phenotype [@pone.0017771-Owens2]. Previously, we have reported that CD34^+^ cells could give rise to SMLCs when cultured in medium supplemented with PDGF~BB~; however, the differentiation of SMLCs was not complete since the assembly of α-SMA or SM-MHC proteins into filaments was not observed [@pone.0017771-Ferreira1]. Curiously, similar results have been obtained in this work for the cell populations tested and exposed to different inductive signals including RA and TGF~β-1~. Our results indicate that the co-culture of SMPCs with fully differentiated hVSMCs induces the assembly of α-SMA and calponin proteins into individualized filaments. This indicates that the cells are able to maturate into a fully contractile phenotype. It is known that both assembly and disassembly of actin filaments is regulated by RhoA [@pone.0017771-Hellstrand1]. Consistent with this, our results show that End-1, an agonist of RhoA pathway, dramatically increases actin polymerization in CD34^+^RA cells and the cells exhibit individualized α-SMA and calponin filaments. However, such inductive effect was not observed in CD34^+^PDGF~BB~ cells, and this could be due to the inhibition of mature SMC marker expression by PDGF~BB~ [@pone.0017771-Dandr1]. Strikingly, these cells are contractile ([**Figure 5**](#pone-0017771-g005){ref-type="fig"}) despite presenting a small percentage of polymerized actin fibers (6%; **[Figure S3](#pone.0017771.s003){ref-type="supplementary-material"}**). Nevertheless, it is known that the total amount of actin that undergoes polymerization after induction of contraction is relatively small [@pone.0017771-Gunst1]. Further research is needed to study the molecular processes that regulate the assembly of actin filaments in smooth muscle tissues and the nature of the actin filaments network that are formed during contractile activation [@pone.0017771-Gunst1]. Contractile and synthetic SMCs, which represent the two ends of a spectrum of SMCs with intermediate phenotypes, clearly show different morphologies [@pone.0017771-Rensen1]. The SMPCs derived in this work seem to have the contractile phenotype, as they are spindle-shaped [@pone.0017771-Hao1], express proteins involved in SMC contraction including α-SMA, SM-MHC, calponin and SMα-22 [@pone.0017771-Rensen1], [@pone.0017771-Owens1], and contract when exposed to carbachol and vasoactive peptides being this effect reversed by the presence of the respective antagonists. We show that SMPCs are able to contract to a plethora of vasoactive agents including carbachol, angiotensin II, histamine, thromboxane-mimetic U46619, endothelin-1, and bradykinin, as hVSMCs. Most of these agonists act through different receptors coupled to G-proteins activating membrane-bound phospholipase C, which leads to the formation of inositol-1,4,5-triphosphate (IP~3~) and diacylglycerol (DAG) [@pone.0017771-Hathaway1]. Through different mechanisms, both molecules induce the accumulation of Ca^2+^ in the cell cytoplasm [@pone.0017771-Hathaway1]. The increase of intracellular Ca^2+^ activates the cell\'s contraction machinery. Our results agree with previous ones showing that hESC-derived SMCs respond to bradykinin, histamine and carbachol by increasing at different levels the intracellular concentration of Ca^2+^ [@pone.0017771-Hill1]. The accumulation of intracellular Ca^2+^ in hESC-derived SMCs exposed to carbachol is minimal, indicating that the contraction of the cell may involve the activation of other intracellular players (e.g. protein kinase C) than Ca^2+^, as described in other studies [@pone.0017771-Harnett1]. Our results show that hESC-derived SMPC contraction is mediated by the activation of CaM/MLCK- and Rho/Rho kinase-dependent pathways [@pone.0017771-Kim1], [@pone.0017771-Hathaway1]. A similar mechanism has been reported recently for SMLCs obtained from human adipose tissue-derived mesenchymal stem cells [@pone.0017771-Kim1]. The results also indicate that the absence of fully organized protein filaments did not prevent the contraction of SMLCs. This is in line with results reported previously by us [@pone.0017771-Ferreira1] and by others [@pone.0017771-Vo1] despite the absence of maturity and organization in the contractile machinery of the cell. SMCs have key biological functions in terms of contraction and secretion of soluble signaling molecules [@pone.0017771-Gerthoffer1]. In the present study we report for the first time the secretion profile of hES-derived SMPCs. CD34^+^ cells that differentiate in medium supplemented with PDGF~BB~ have similar secretome profile as hVSMCs. They express high levels (\>100 pg/mL) of IL-6 and IL-8, moderate levels (between 100 and 10 pg/mL) of IFN-γ, and small levels of (between 10 and 1 pg/mL) of IL-7, G-CSF, MCP-1 and TNF-α. IL-6 is a cytokine with potent inflammatory properties and metabolic effects in SMCs. It has been shown that IL-6 secretion is inversely correlated to glucose consumption [@pone.0017771-Mayr1]. IL-8 is a cytokine that induces the proliferation and chemotaxis of smooth muscle cells and increases the activity of mitogen-activated protein kinase (MAPK) [@pone.0017771-Yue1]. IFN-γ acts on vascular smooth muscle cells to induce cellular proliferation [@pone.0017771-Wang2]. Although several studies have reported the differentiation of different stem cells into SMCs, very few evaluated the impact of the 3D environment at the geno- and phenotype of the differentiated cells [@pone.0017771-Ross1]. Gene studies have demonstrated that human aortic SMCs encapsulated in a 3D collagen matrix have significantly less focal adhesions, lower proliferation, and lower α-SMA expression than cells in 2D [@pone.0017771-Li1]. However, to our knowledge, no study has compared the gene expression of SMLCs from different origins with fully differentiated SMCs when cultured in 2D or 3D systems. Our results show that hESC-derived SMPCs encapsulated for 3 days in fibrin gels express similar gene levels of α-SMA, SM-MHC, and SMα22 as encapsulated hVSMCs. In addition, considering an array of 84 genes related to cell-cell and cell-matrix interactions, 3D-cultured-SMPCs had a more hVSMC-similar gene expression profile than 2D-cultured SMPCs. This shows that 3D scaffolds may induce further the differentiation of SMPCs into SMCs. Several factors might account for the differences found between 2D and 3D culture systems including (i) ECM stiffness and (ii) ECM 3D environment. It has been demonstrated that the stiffness of the ECM has a high impact in the cytoskeletal and focal adhesion assembly of SMCs [@pone.0017771-Peyton1]. Furthermore, SMCs cultured within 3D polyethylene glycol-fibrinogen [@pone.0017771-Peyton2] or collagen [@pone.0017771-Li1] gels had less proliferation, stress fibers and focal adhesion than on 2D culture systems. Future studies should evaluate the effect of both factors in the modulation of geno- and phenotype of the differentiated cells over the time and study the underlining mechanism. In this study we assessed the effect of the 3D matrix after 3 days of cell encapsulation. During this time, previous studies have shown that SMCs are able to migrate and remodel the ECM [@pone.0017771-Naito1]; however, it would be interesting to extend these studies in time. Although very recent studies have identified and characterized hESC-derived populations with SMC potential [@pone.0017771-Vo1], [@pone.0017771-Hill1], identified bioactive agents involved in their SMC differentiation, and evaluated their functionality, our study advance these results in several ways. First, it provides a detailed analysis of the phenotype, secretome and function (unraveling the mechanism of cell contraction) of hESC-derived SMCs at levels not documented before. In contrast to the previous studies that have used monolayer-based differentiation protocols, our methodology uses an EB-differentiation step to isolate progenitor cells, which might have advantages for the scale-up of the process through the use of bioreactors. Second, our study identifies a methodology to induce the organization of the contractile protein filaments. Third, it demonstrates the importance of a 3D environment to modulate the activity of hESC-derived SMLCs. Recent studies reported the derivation of pluripotent stem cells from human somatic cells by retroviral transduction [@pone.0017771-Takahashi1]. These reprogrammed cells share many features with hESCs in terms of morphology, gene expression and differentiation potential, and open a new avenue for the potential derivation of autologous cells for regenerative medicine and drug screening. Indeed, the differentiation of iPSCs into SMCs has been recently reported [@pone.0017771-Lee1]. It remains to be determined whether the methodology described in this work can be applied on iPSCs and this is an issue that will be evaluated in future work. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### **SMC proteins are expressed on hESC-derived SMPCs.** CD34^+^, CD34^+^KDR^−^ and CD34^−^ cells differentiated under different media conditions express α-SMA, calponin and SM-MHC, as evaluated by immunofluorescence. hVSMCs were used as a positive control and HUVECs as a negative control for the SMC markers. In all figures, bar corresponds to 50 µm. (TIFF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### **Isotype controls for SMPC immunostainning.** Stained the isotype controls IgG~2A~ and IgG~1~ for the SMC markers: α-SMA, SM-MHC and Calponin. Cell nuclei were stained with 4′, 6′-diamidino-2-phenylindole (DAPI). Cells were labeled with mouse anti--human IgG~2A~ and IgG~1~ antibodies. Bar corresponds to 50 µm. (TIFF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S3 ::: {.caption} ###### **Organization and expression of contractile proteins.** (**A**) Quantification of organized α-SMA, SM-MHC and calponin filaments. (B) Expression of α-SMA in differentiated CD34^+^, CD34^+^KDR^−^ and CD34^−^ cells. Cells were differentiated for 3 passages (approximately 18 days after cell seeding). hVSMCs and HUVECs were used as positive and negative controls, respectively. In all graphs, the percentages of positive cells were calculated based in the isotype controls (gray plot) and are shown in each histogram plot. (TIFF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S4 ::: {.caption} ###### **Secretomic profile of hESC-derived cells.** 17 cytokines were measured simultaneously in the medium collected from CD34^+^KDR^−^PDGF~BB~ and CD34^+^KDR^−^EGM-2 cells. A standard range of 0.2 to 3,200 pg/mL was used. Samples and controls were run in triplicate, standards and blanks in duplicate. (TIFF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S5 ::: {.caption} ###### **Contractility of hESC-derived cells.** Cells were loaded with FURA-2/AM and their response to carbachol (10^−5^ M) was monitored by fluorescence. The response profile was compared to the one observed for hVSMCs and HUVECs, as positive and negative controls, respectively. (TIFF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S6 ::: {.caption} ###### **Contractility of hESC-derived cells.** Cells were loaded with FURA-2/AM and their response to vasoactive agonists (bradykinin (10^−7^ M), angiotensin II (10^−5^ M) and histamine (100 µM) and depolarization agents (KCl; 50 mM) was monitored by fluorescence. The response profile was compared to the one observed for hVSMCs and HUVECs, as positive and negative controls, respectively. (TIFF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S7 ::: {.caption} ###### **Integrin gene expression.** Gene expression on CD34^+^PDGF~BB~ and CD34^+^RA cells was normalized by gene expression on hVSMCs, both cultured in 3D or 2D systems. Gene expression was obtained from the RT^2^ Profiler™ PCR array. (TIFF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S8 ::: {.caption} ###### **Expression and organization of SMC proteins on CD34^−^PDGF~BB~ cells treated with vasoactive agents for 3 days.** A) Quantification by immunocytochemistry analysis. Results are Mean ± SEM (*n* = 8). B) Expression of a-SMA and calponin in CD34^−^PDGF~BB~ cells treated with End-1 (10 nM) for 3 days. Bar corresponds to 50 µm. (TIFF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S9 ::: {.caption} ###### **Gene expression in CD34^+^PDGF~BB~ and CD34^+^RA cells after treatment with vasoactive agents for 3 days.** Gene expression in CD34^+^PDGF~BB~ and CD34^+^RA cells was normalized by gene expression in hVSMCs. Results are Mean ± SEM (*n* = 4). (TIFF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S10 ::: {.caption} ###### **Expression and organization of troponin T (TrpnT) and vascular endothelial-cadherin (VECad) on CD34^+^RA (A) and CD34^+^PDGF~BB~ (B) cells treated with End-1 for 3 days.** Human cardiac tissue (for TrpnT) and HUVECs (for VeCad) were used as positive controls. Bar corresponds to 50 µm. (TIFF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S1 ::: {.caption} ###### **Primers used for Real Time PCR.** PCR conditions: initial denaturation step at 94°C for 5 min; 40 cycles of denaturation at 94°C for 30 sec, annealing at 60°C for 33 sec and extension at 72°C for 30 sec. At the end was performed a final 7 minutes extension at 72°C. After amplification, the melting curve profile or agarose gel electrophoresis was used to determine the specificity of PCR products. (TIFF) ::: ::: {.caption} ###### Click here for additional data file. ::: Materials and Methods S1 ::: {.caption} ###### (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: The authors would like to thank Renata Gomes for help in the preparation of this manuscript. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was supported by a Marie Curie-Reintegration Grant, MIT-Portugal program, Crioestaminal, Associação Viver a Ciência and Fundação para a Ciência e a Tecnologia (PTDC/SA-BEB/098468/2008 and PTDC/CTM/099659/2008 to L.F.; SFRH/BD/40077/2007 to H.V.). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: HV LF. Performed the experiments: HV RN. Analyzed the data: HV RN MG LF. Wrote the paper: HV LF.

PubMed Central

2024-06-05T04:04:19.870009

2011-3-10

PMC3053393

Introduction {#s1} ============ The integration of information from several sensory modalities offers an enriched perception of the world and provides a more robust method for representing and recognizing objects. Multisensory integration increases information content and disambiguates information that might otherwise have multiple interpretations [@pone.0017777-Ernst1]. Furthermore, integrating multisensory information enhances the reliability of sensory estimates [@pone.0017777-Ernst2], [@pone.0017777-Alais1] and increases the speed of perceptual learning [@pone.0017777-Seitz1], [@pone.0017777-Kim1]. How is multisensory information stored in the brain? The neuronal basis of multisensory integration has been investigated in several behaviors [@pone.0017777-Stein1], [@pone.0017777-Ghazanfar1], [@pone.0017777-Shams1]. However, relatively few studies have directly assessed the effect of experience on the neuronal representation of multisensory information: Familiar and unfamiliar audiovisual stimuli evoke differential activation of the posterior superior temporal sulcus [@pone.0017777-Hein1], left inferior frontal cortex, intraparietal sulcus [@pone.0017777-Naumer1], occipitotemporal junction and parahippocampal gyrus [@pone.0017777-Tanabe1]. However, at the level of the single neuron, comparable studies of stimulus familiarity are lacking and therefore the neuronal basis of multisensory memory formation remains unclear. Additionally, one would like to know whether the mechanisms underlying the storage of unisensory and multisensory information resemble one another. Despite the prevalence of multisensory stimuli in the natural world, many studies of object recognition have investigated the representation of unisensory information [@pone.0017777-Xiang1], [@pone.0017777-Horn1], [@pone.0017777-Nakamori1]. It therefore remains to be tested whether the findings of these studies extend to multisensory information and whether one can explain multisensory information storage in terms of the storage of information about its unisensory components. To address these questions, we studied an animal model of object recognition; filial imprinting, in which young birds learn to recognize an audiovisual stimulus [@pone.0017777-Sluckin1]. We recorded neurons from a critical forebrain region [@pone.0017777-McCabe1], [@pone.0017777-Horn2]; the intermediate and medial mesopallium (IMM) of imprinted and naïve domestic chicks during presentation of an audiovisual imprinting stimulus and novel object, and of their auditory and visual components. We presented a fully balanced stimulus set ([Table 1](#pone-0017777-t001){ref-type="table"}) that included incongruent audiovisual combinations, in which the auditory and visual components of an imprinting stimulus and novel object were mismatched. This experimental design allowed us to compare the effect of imprinting on unisensory and multisensory neuronal responses and to investigate the nature of any multisensory representation formed through experience. We find that imprinting enhanced the mean magnitude of neuronal response to unisensory components of the imprinting stimuli but not to the multisensory imprinting stimulus itself. Rather imprinting most strongly enhanced the response of neurons to a mismatched audiovisual stimulus combining the visual component of the imprinting stimulus and auditory component of a novel object. ::: {#pone-0017777-t001 .table-wrap} 10.1371/journal.pone.0017777.t001 Table 1 ::: {.caption} ###### Stimulus set presented to each animal with abbreviations. ::: {#pone-0017777-t001-1} Visual Component --------------------- ------------------ ---------------- ---------------- None (Unisensory) Not Applicable **V~IS~** **V~NO~** Imprinting stimulus **A~IS~** **A~IS~V~IS~** **A~IS~V~NO~** Novel Object **A~NO~** **A~NO~V~IS~** **A~NO~V~NO~** Each stimulus was presented for 4 seconds 15--20 times with a minimum inter-stimulus interval of 4 s. ::: Results {#s2} ======= We recorded activity from 157 neurons in the IMM of three imprinted and three naive chicks (see [Methods](#s4){ref-type="sec"}) during presentation of an audiovisual imprinting stimulus (IS) and novel object (NO) and their auditory and visual components. We characterized the response of each neuron to each stimulus using response magnitude - the firing rate during presentation expressed as a percentage of the baseline firing rate measured before presentation. Visual Stimuli (V~IS~ and V~NO~) {#s2a} -------------------------------- We found that neurons recorded from imprinted chicks responded more strongly to the visual component of the imprinting stimulus (V~IS~) than the visual component of the novel object (V~NO~) whereas neurons recorded from naïve chicks did not. [**Figure 1a**](#pone-0017777-g001){ref-type="fig"} shows the firing rate of two neurons recorded from an imprinted chick and a naïve chick before and during presentation of visual stimuli. Across the neuronal population ([**Fig. 1b**](#pone-0017777-g001){ref-type="fig"}), we found that the mean response magnitude was significantly greater towards V~IS~ than V~NO~ in the neurons recorded from imprinted by not naïve chicks (ANOVA, interaction between effects of group and stimulus: *F* ~1,\ 155~ = 6.01, *P* = 0.015). ::: {#pone-0017777-g001 .fig} 10.1371/journal.pone.0017777.g001 Figure 1 ::: {.caption} ###### Single cell and population responses to visual stimuli. (**A**) Raster plot and peri-stimulus time histograms illustrating the firing rate of two neurons recorded from a naïve and an imprinted chick before and during presentation of the visual components of the imprinting stimulus (V~IS~) and novel object (V~NO~). Percentage values indicate the response magnitude calculated as the firing rate during presentation (0 to 4 s) expressed as a percentage of pre-stimulus baseline firing rate (−4 to 0 s). (**B**) Mean (± s.e.m.) response magnitude of neuronal populations recorded in naïve (white: n = 85) and imprinted (black: n = 72) chicks. (\*) indicates a significant interaction between group and stimulus (*P* = 0.015). :::  ::: Auditory Stimuli (A~IS~ and A~NO~) {#s2b} ---------------------------------- Similarly, the neuron presented in [**figure 1a**](#pone-0017777-g001){ref-type="fig"} recorded from an imprinted chick responded more strongly to the auditory component of the imprinting stimulus (A~IS~) than that of the novel object (A~NO~) whereas the neuron recorded from a naïve chick showed little difference in response between stimuli ([**Fig. 2a**](#pone-0017777-g002){ref-type="fig"}). Across the recorded populations, the mean response of neurons from imprinted chicks to A~IS~ was greater than to A~NO~ whereas the mean response of neurons from naïve chicks to A~IS~ was weaker than to A~NO~ ([**Fig. 2b**](#pone-0017777-g002){ref-type="fig"}) (interaction between group and stimulus: *F* ~1,\ 155~ = 5.86, *P* = 0.017). ::: {#pone-0017777-g002 .fig} 10.1371/journal.pone.0017777.g002 Figure 2 ::: {.caption} ###### Single cell and population responses to auditory stimuli. (**A**) Raster plot and peri-stimulus time histograms illustrating the firing rate of the same neurons shown in [figure 1](#pone-0017777-g001){ref-type="fig"} before and during presentation of the auditory components of the imprinting stimulus (A~IS~) and novel object (A~NO~). Percentage values indicate the response magnitude calculated as the firing rate during presentation (0 to 4 s) expressed as a percentage of pre-stimulus baseline firing rate (−4 to 0 s). (**B**) Mean (± s.e.m.) response magnitude of neuronal populations recorded in naïve (white) and imprinted (black) chicks. (\*) indicates a significant interaction between group and stimulus (*P* = 0.017). :::  ::: Congruent Audiovisual Stimuli (A~IS~V~IS~ and A~NO~V~NO~) {#s2c} --------------------------------------------------------- In contrast, single neurons recorded from the imprinted chick such as that presented in [**Fig. 1**](#pone-0017777-g001){ref-type="fig"} and [**Fig. 2**](#pone-0017777-g002){ref-type="fig"} responded similarly to the audiovisual imprinting stimulus (A~IS~V~IS~) and novel object (A~NO~V~NO~) and there was little difference in response to the audiovisual imprinting stimulus between neurons recorded from imprinted and naïve chicks ([**Fig. 3a**](#pone-0017777-g003){ref-type="fig"}). Comparison between neuronal populations recorded from imprinted and naïve chicks revealed no main effect of group or stimulus and no interaction between these factors ([**Fig. 3b**](#pone-0017777-g003){ref-type="fig"}). ::: {#pone-0017777-g003 .fig} 10.1371/journal.pone.0017777.g003 Figure 3 ::: {.caption} ###### Single cell and population responses to congruent audiovisual stimuli. (**A**) Raster plot and peri-stimulus time histograms illustrating the firing rate of the same neurons from [figures 1](#pone-0017777-g001){ref-type="fig"} and [2](#pone-0017777-g002){ref-type="fig"} before and during presentation of the audiovisual imprinting stimulus (A~IS~ V~IS~) and novel object (A~NO~ V~NO~). Percentage values indicate the response magnitude calculated as the firing rate during presentation (0 to 4 s) expressed as a percentage of pre-stimulus baseline firing rate (−4 to 0 s). (**B**) Mean (± s.e.m.) response magnitude of neuronal populations recorded in naïve (white) and imprinted (black) chicks. (ns) indicates the absence of interaction between group and stimulus. :::  ::: The effect of imprinting on mean response magnitude to the imprinting stimulus varied with modality ([**Fig. 4**](#pone-0017777-g004){ref-type="fig"}): Comparison between visual and audiovisual modalities demonstrated that enhancement of mean response magnitude to V~IS~ was absent for A~IS~V~IS~ (interaction between group and modality; *F* ~1,\ 155~ = 3.93, *P* = 0.049). Similarly, comparison between auditory (A~IS~) and audiovisual (A~IS~V~IS~) modalities demonstrated that the enhancement of mean response magnitude was limited to auditory stimuli (*F* ~1,\ 154~ = 7.91, *P* = 0.006). Thus imprinting leads to the modification of neuronal responses that are limited to the unisensory components of the imprinting stimulus. Comparison between the effects of imprinting on responses to the audiovisual novel object and its auditory or visual components revealed no interactions between modality and group. ::: {#pone-0017777-g004 .fig} 10.1371/journal.pone.0017777.g004 Figure 4 ::: {.caption} ###### Comparison between the effects of imprinting on different modalities. Mean (± s.e.m.) response magnitude to the audiovisual imprinting stimulus (A~IS~V~IS~) and its auditory (A~IS~) and visual components (V~IS~). (\*\*) indicates interaction between the effects of imprinting on auditory and audiovisual stimuli (*P* = 0.006). (\*) indicates interaction between the effects of imprinting on auditory and audiovisual stimuli (*P* = 0.049). :::  ::: Incongruent Audiovisual Stimuli (A~IS~V~NO~ and A~NO~V~IS~) {#s2d} ----------------------------------------------------------- By mismatching the auditory and visual components of the imprinting stimulus and novel object, it was possible to create two incongruent audiovisual stimuli (A~IS~V~NO~ and A~NO~V~IS~). We found that single neurons recorded from imprinted chicks responded more strongly to the combination A~NO~V~IS~ than neurons recorded from naïve chicks ([**Fig. 5a**](#pone-0017777-g005){ref-type="fig"}). Comparing neuronal populations revealed a significant effect of group (*F* ~1,\ 155~  = 26.23, *P*\<0.001) ([**Fig. 5b**](#pone-0017777-g005){ref-type="fig"}). However, there was no effect of imprinting on the mean response to the alternative incongruent combination A~IS~V~NO~ (*P*\>0.1). ::: {#pone-0017777-g005 .fig} 10.1371/journal.pone.0017777.g005 Figure 5 ::: {.caption} ###### Single cell and population responses to incongruent audiovisual stimuli. (**A**) Raster plot and peri-stimulus time histograms illustrating the firing rate of the same neurons from [figures 1](#pone-0017777-g001){ref-type="fig"}--[](#pone-0017777-g002){ref-type="fig"} [3](#pone-0017777-g003){ref-type="fig"} before and during presentation of incongruent audiovisual stimulus (A~NO~ V~IS~ and A~IS~ V~NO~). Percentage values indicate the response magnitude calculated as the firing rate during presentation (0 to 4 s) expressed as a percentage of pre-stimulus baseline firing rate (−4 to 0 s). (**B**) Mean (± s.e.m.) response magnitude of neuronal populations recorded in naïve (white) and imprinted (black) chicks. (\*) indicates a significant effect of group on mean response magnitude to A~NO~ V~IS~ (*P*\<0.001). :::  ::: Multisensory Integration {#s2e} ------------------------ We also investigated whether multisensory integration was affected by imprinting by calculating an additivity index for each neuron for each audiovisual stimulus. In order to calculate additivity, we measured the change in firing rate during presentation of an audiovisual stimulus (AV) and its auditory (A) and visual (V) components ([**Fig. 6a**](#pone-0017777-g006){ref-type="fig"}). We then calculated the difference between the change in firing rate during presentation of AV and the sum of changes in firing rate during presentation of A and V, and divided this by the total sum of changes in firing rate (see [**Fig. 6a**](#pone-0017777-g006){ref-type="fig"} and [Methods](#s4){ref-type="sec"}). The resulting variable therefore ranges from −1 to 1 with values greater than and less than zero indicate subadditivity and superadditivity respectively, whereas zero indicates that the change in firing rate during presentation of the audiovisual stimulus is equal to the sum of changes in firing rated during presentation of its auditory and visual components. ::: {#pone-0017777-g006 .fig} 10.1371/journal.pone.0017777.g006 Figure 6 ::: {.caption} ###### Experience-dependent audiovisual integration. (**A**) The response of a neuron recorded from an imprinted chick to the audiovisual imprinting stimulus (A~IS~V~IS~) and its auditory (A~IS~) and visual components (V~IS~) (see also [Figs. 1](#pone-0017777-g001){ref-type="fig"}--[3](#pone-0017777-g003){ref-type="fig"}). The change in firing rate was calculated for each stimulus and used to calculate the additivity index. (**B**) Mean (± s.e.m.) additvity index of neuronal populations recorded from naive and imprinted chicks to the audiovisual imprinting stimulus (A~IS~V~IS~) and novel object (A~NO~V~NO~) and to incongruent audiovisual stimuli (A~NO~V~IS~ and A~IS~V~NO~). (\*\*\*) indicates significant effect of group on additivity index of A~NO~V~IS~ (*P*\<0.001). :::  ::: [**Figure 6b**](#pone-0017777-g006){ref-type="fig"} shows the mean additivity index for neurons recorded from naïve and imprinted chicks for each audiovisual stimulus. In accordance with our earlier findings, imprinting led to the increase in additivity in the case of incongruent responses to A~NO~V~IS~ (*T* ~153~ = 4.68, *P*\<0.001). This result can be explained by the strong increase in audiovisual response induced by imprinting, coupled with the weaker increase in familiar visual response and decrease in unfamiliar auditory response. The difference in mean additivity between groups was not significant for any of the other audiovisual stimuli. Discussion {#s3} ========== Unisensory and multisensory neuroplasticity {#s3a} ------------------------------------------- We report an imprinting-related enhancement in responses of neurons within the IMM for the auditory and visual components of an imprinting stimulus but not the audiovisual imprinting stimulus itself. This leads us to conclude that imprinting-related enhancement of the response magnitude of IMM neurons is limited to the unisensory components of an imprinting stimulus and does not extend to the audiovisual compound. Our findings are consistent with earlier reports of the selective enhancement of neuronal responsiveness to auditory and visual components of an imprinting stimulus [@pone.0017777-Horn1], [@pone.0017777-Brown1], [@pone.0017777-Nicol1]. Our findings also support preliminary results of more recent work in which the proportion of IMM neurons found to be responsive to the visual component of an imprinting stimulus increased following imprinting whereas the proportion responsive to the audiovisual imprinting stimulus did not (Nicol & Horn, *Proceedings of the Physiological Society* 2009 Cardiff, UK. Available at [www.physoc.org/Proceedings](http://www.physoc.org/Proceedings): Last Accessed Jan 2011). However, our findings conflict with an earlier study by Brown and Horn [@pone.0017777-Brown1] in which the proportion of sites within the IMM responsive to an audiovisual imprinting stimulus increased with imprinting (albeit, this increase was smaller than that reported for the visual component of the imprinting stimulus). This finding led to the assumption that imprinting similarly affects responsiveness to the audiovisual imprinting stimulus and its visual component [@pone.0017777-Jackson1]. Our findings challenge this assumption and suggest that unisensory and multisensory stimuli cannot be considered equivalent in the study of the neurophysiological basis of filial imprinting. The disparity between the present findings and those of Brown and Horn cannot be attributed to the difference in measurement used to characterize neurons (proportion of responsive sites vs. response magnitude) because reanalysis of our data according to the same method confirmed our finding: imprinting enhanced the proportion of sites responsive to the visual component of the imprinting stimulus but not the audiovisual stimulus itself ([Table 2](#pone-0017777-t002){ref-type="table"}). Furthermore, recent work by other investigators has also found a dissociation in the effects of imprinting on proportion of responsive neurons to the audiovisual imprinting stimulus and its visual component (Nicol & Horn, *Proceedings of the Physiological Society* 2009 Cardiff, UK. Available at [www.physoc.org/Proceedings](http://www.physoc.org/Proceedings): Last Accessed Jan 2011). Our results may differ from those of Brown and Horn because of differences in recording method: In the current study, we used tetrodes to identify the responses of single neurons whereas the earlier results were obtained using multi-unit recordings of the activity of clusters of neurons. It is possible that multi-unit recordings are limited in their sensitivity as a particularly responsive neuron can cause an entire cluster to be identified as responsive when the majority of units are unresponsive. Tetrodes allow the separation of neurons within such a cluster and therefore may avoid such biases, providing a more sensitive index of neuronal activity that could explain the difference between past studies and the present findings. ::: {#pone-0017777-t002 .table-wrap} 10.1371/journal.pone.0017777.t002 Table 2 ::: {.caption} ###### Proportion of neurons responsive to the audiovisual imprinting stimulus and its visual component. ::: {#pone-0017777-t002-2} [Modality]{.underline} [Group]{.underline} [Proportion responsive]{.underline} ------------------------ --------------------- ------------------------------------- ------- Visual Naïve 17/85 20.0% Imprinted 24/72 33.3% Audiovisual Naïve 39/85 45.9% Imprinted 33/72 45.8% Responsive neurons were defined as those whose firing rate during stimulus presentation significantly differed from the baseline firing rate before stimulus presentation (T-test: see ref. [@pone.0017777-Brown1] for details). ::: The dissociation between changes in mean response magnitude to the multisensory imprinting stimulus and its unisensory components may be explained by the principle of inverse effectiveness. This principle describes the phenomenon occurring in both mammals and birds in which the effect of adding an additional modality to a stimulus on response magnitude is inversely proportional to the original salience of the stimulus when presented alone [@pone.0017777-Meredith1], [@pone.0017777-Zahar1], [@pone.0017777-Kayser1]. In the current study, the enhancement of neuronal responses to the auditory or visual components of the imprinting stimulus may lead to a reduction in the effectiveness of adding a second modality when the audiovisual imprinting stimulus is presented. This would lead to a relatively constant mean response magnitude to the audiovisual imprinting stimulus despite an increase in response magnitude to auditory and visual components, as we report. This interpretation is supported by the relatively weak correlation between response magnitude of neurons to the visual and audiovisual imprinting stimulus ([**Fig. 7a**](#pone-0017777-g007){ref-type="fig"}) and not novel object ([**Fig. 7c**](#pone-0017777-g007){ref-type="fig"}). However, there are strong correlations between the response of neurons to audiovisual stimuli and their auditory components, both for the imprinting stimulus ([**Fig. 7b**](#pone-0017777-g007){ref-type="fig"}) and the novel object ([**Fig. 7d**](#pone-0017777-g007){ref-type="fig"}). These correlations would not be predicted by the principle of inverse effectiveness; however it is possible that correlations are present because auditory stimuli are sufficiently salient that little enhancement through multisensory integration occurs anyway ([**Fig. 4**](#pone-0017777-g004){ref-type="fig"}). Under such circumstances, correlations between auditory and audiovisual responses might be expected as the audiovisual stimulus is no more salient than its auditory component. ::: {#pone-0017777-g007 .fig} 10.1371/journal.pone.0017777.g007 Figure 7 ::: {.caption} ###### Principle of inverse effectiveness. Axes indicate the response magnitude (% of baseline activity) of individual neurons recorded from naïve (grey) and imprinted (black) chicks to unisensory (x-axis) and multisensory stimuli (y-axis). Equations and r-values indicate regression and correlation coefficients respectively. (**A**) Visual component (V~IS~) vs. audiovisual imprinting stimulus (A~IS~V~IS~). (**B**) Auditory component (A~IS~) vs. audiovisual imprinting stimulus (A~IS~V~IS~). (**C**) Visual component (V~NO~) vs. audiovisual novel object (A~NO~V~NO~). (**D**) Auditory component (A~NO~) vs. audiovisual novel object (A~NO~V~NO~). :::  ::: The relationship between the behavior of chicks and the responses of IMM neurons to the audiovisual imprinting stimulus and novel object remains unclear. Imprinted but not naïve chicks were able to discriminate between the audiovisual imprinting stimulus and novel object ([**Fig. 8**](#pone-0017777-g008){ref-type="fig"}) yet the mean response magnitude of neurons to both stimuli in both groups were similar ([**Fig. 4**](#pone-0017777-g004){ref-type="fig"}). It therefore seems unlikely that response magnitudes of IMM neurons directly contribute to the discrimination between imprinting stimulus and novel object; rather our results suggest the neurons within the IMM may serve to identify unexpected auditory properties of the imprinting stimulus (see below). ::: {#pone-0017777-g008 .fig} 10.1371/journal.pone.0017777.g008 Figure 8 ::: {.caption} ###### Behaviour during presentation of audiovisual stimuli. Preference scores for audiovisual imprinting stimulus. Box plots indicate median preference scores (center bar), upper and lower quartiles (box) and whiskers represent the range. (\*\*) Comparison of individual medians revealed that the preferences for the audiovisual imprinting stimulus of imprinted (n = 8) but not naïve chicks (n = 9) were greater than chance (50%, sign test: naïve chicks, *P*\>0.5; imprinted, *P* = 0.008). :::  ::: Audiovisual Incongruence {#s3b} ------------------------ We report that the mean response magnitude to the incongruent audiovisual stimulus A~NO~V~IS~ (the visual component of the imprinting stimulus combined with the auditory component of the novel object) was greater in imprinted than naïve chicks. This result should be interpreted with caution as the difference in response magnitude derived mainly from the unusually weak responses to A~NO~V~IS~ recorded from neurons in naïve chicks. There is no prior reason to expect this audiovisual stimulus to differ so notably in its salience to naïve chicks from other audiovisual stimuli, raising the possibility that the finding is anomalous. However, there is also no reason to believe that the recording of neuronal responses to A~NO~V~IS~ was any less accurate than all other audiovisual stimuli: Presentation order was randomized across neuronal tests, making it unlikely that a consistent time of presentation biased the results. Furthermore, the analysis of neuronal activity and calculation of response magnitude following single unit isolation was automated for all stimuli, and therefore any inaccuracy in measurement of response magnitude for A~NO~V~IS~ should also be manifest in the measurement of response magnitude for all other stimuli. Moreover, it is also notable that of all audiovisual stimuli, A~NO~V~IS~ evoked the strongest mean response magnitude in the population recorded from imprinted chicks. We therefore believe that it is unlikely that the imprinting-related enhancement of mean response magnitude to A~NO~V~IS~ was anomalous but rather interpret the effect tentatively as a result of imprinting that may reveal important details regarding the function of IMM neurons; namely the detection of incongruous auditory accompaniments to the visual imprinting stimulus. The suggestion that imprinting enhances neuronal responses to incongruous auditory accompaniments of the visual imprinting stimulus is consistent with the finding that the mean response magnitude to the visual component of the imprinting stimulus (V~IS~) was stronger in imprinted than naïve birds, as presentation of V~IS~ in the absence of any call could be considered an incongruent auditory condition given the original imprinting exposure was to the audiovisual compound A~IS~V~IS~. The suggestion is also consistent with the more general proposal that neurons in the IMM respond to unexpected variations from the original imprinting experience as neurons in the IMM also respond more strongly to A~IS~V~IS~ when presented in an unfamiliar than familiar visual context following imprinting (Town & McCabe, Unpublished). It remains to be seen whether this hypothesis accurately predicts the effects of imprinting on neuronal responses to more ethologically relevant stimuli such as live hens and naturalistic situations. Multisensory integration in the IMM {#s3c} ----------------------------------- IMM neurons recorded from imprinted chicks responded strongly to auditory stimuli and to the visual component of the imprinting stimulus demonstrating that, at least following imprinting, information from multiple sensory modalities is integrated in the IMM. The ability of IMM neurons to respond to multiple modalities of sensory information is consistent with previous findings [@pone.0017777-Brown1], [@pone.0017777-Nicol1] and the projection of afferents from visual (optic tectum, arcopallium intermedium, nidopallium and the Wulst) and auditory (Field L) regions of the brain to the IMM [@pone.0017777-Bradley1]. The afferents sent to the IMM from the nidopallium may also convey somatosensory information [@pone.0017777-Bradley1] and chicks show the ability to imprint on tactile information [@pone.0017777-Taylor1]. It would therefore be interesting to test whether IMM neurons also respond to somatosensory stimuli and whether these responses are dependent upon imprinting. When calculating additivity in the IMM, we found that mean values of neurons recorded from imprinted chicks were near zero for the incongruent audiovisual stimulus A~NO~V~IS~ (mean ± s.e.m  =  −1.78±4.5; comparison vs. 0: *P*\>0.5). This suggests that on average, the sum of changes in neuronal activity during presentation of the audiovisual stimulus was similar to the sum of changes in neuronal activity during separate presentation of its unisensory components. Mean additivity was also near zero for the incongruent stimulus A~IS~V~NO~ and audiovisual imprinting stimulus; however neither is likely to reflect audiovisual integration at the population level because the visual component (V~NO~) of the stimulus did not evoke strong responses from neurons and therefore a mean additivity value near zero may reflect similar responses to audiovisual and auditory stimuli. Thus, at least during presentation of A~NO~V~IS,~ neurons may integrate visual and auditory information. Remaining Questions {#s3d} ------------------- At present it remains unclear how incongruence detection is performed, or what its function might be. Additionally, it is unclear how auditory and visual information may be integrated, at least during the response of neurons recorded from imprinted chicks to A~NO~V~IS~. Much of this obscurity stems from the lack of structural and biochemical knowledge about neurophysiologically characterized neurons. In terms of incongruence detection and multisensory integration, the underlying mechanisms will depend upon the form in which information reaches the IMM. It is not clear whether auditory and visual information are provided through separate (auditory and visual) or mixed (auditory, visual and audiovisual) channels. Recent evidence has demonstrated that the optic tectum, a structure thought to provide visual input to the IMM, is capable of multisensory integration in the Barn Owl [@pone.0017777-Zahar1]. Therefore, it may be likely that IMM neurons receive unisensory and multisensory information at synapses from structures that were originally described as unimodal. This speculation remains to be confirmed and will require paired recordings of neurons from the IMM and their presynaptic inputs from other regions of the brain. Furthermore, understanding the computations performed during synaptic integration within the dendritic tree will require the application of techniques such as *in-vivo* calcium-imaging to measure post-synaptic potentials at multiple synapses in deep tissue of behaving animals. With regard to the function of incongruence detection, future studies will need to elucidate the regions of the brain to which specific neurons, responding most strongly to A~NO~V~IS~ and V~IS~, project. Neurons may project axons locally within the IMM or to relatively distant cognitive and motor regions such as the hyperpallium apicale, arcopallium (homologous to the mammalian amygdala) and striatum [@pone.0017777-Bradley1] and therefore imprinting-related responses of an IMM neuron to A~NO~V~IS~ and V~IS~ may serve one or more of several functions (e.g. social recognition, emotional behavior or generation of motor output) depending on its innervations pattern. It is notable therefore, that imprinted chicks did not differ greatly in the extent to which they approached A~NO~V~IS~ or A~IS~V~IS~ (i.e. congruent and incongruent stimuli) during stimulus presentations suggesting that the mean response magnitude of neurons within the IMM and approach behavior in the training wheel are not directly linked (median approach distance: A~IS~V~IS~  =  1.25 m, A~NO~V~IS~  =  1.23 m). The nature of synapses in innervated regions will also be of crucial importance in understanding the function of IMM neurons; within the IMM there is an imprinting-related enhancement of potassium stimulated GABA (γ-aminobutyric acid) release suggesting that imprinting may alter the balance of inhibition within the IMM [@pone.0017777-McCabe2]; however understanding how changes in inhibitory synapses affect the circuits in which IMM neurons take part will require knowledge of the neurophysiological properties of pre- and post-synaptic neurons. Conclusions {#s3e} ----------- In summary, we report a dissociation between the effects of imprinting on the responses of IMM neurons to an audiovisual imprinting stimulus and its auditory and visual components, challenging the existing assumption that the effects of imprinting on unisensory and multisensory responsiveness are equivalent. We report an enhancement in mean response magnitude to an incongruent audiovisual stimulus, suggesting that neurons within the IMM may signal incongruous auditory accompaniments to the visual component of the imprinting stimulus. In future, it will be important to simultaneously characterize the neurophysiological, structural and biochemical properties of neurons in order to better understand the function of the IMM during imprinting. Materials and Methods {#s4} ===================== Subjects {#s4a} -------- Subjects were domestic chicks (*Gallus gallus domesticus*, Ros 308 Strain) incubated and reared in darkness within incubators maintained at 32--35°C. All procedures were performed in accordance with the UK Animals (Scientific Procedures) Act 1986, under the UK Home Office Project License No. 80/2276 and were approved by the University Biomedical Support Services (UBSS) ethical review committee, University of Cambridge. Imprinting {#s4b} ---------- Approximately 24 hours after hatching, 30 chicks were imprinted using methods similar to those described elsewhere [@pone.0017777-McCabe1]. Briefly; chicks were placed in running wheels within a darkened training box maintained at 30°C and exposed to an imprinting stimulus - a rotating, illuminated red box presented in conjunction with a maternal hen call (Call A, see below for further details about stimuli) - for two sessions of 60 minutes separated by an hour interval in which chicks were returned to incubators. Imprinted chicks were then identified by their ability to discriminate between the imprinting stimulus and novel object (a rotating, illuminated blue cylinder) in a sequential preference test in which chicks were returned to training wheels and presented with the visual component of the imprinting stimulus and novel object for two periods of four minutes in an ABBA design in which stimulus order was counterbalanced; neither stimulus was accompanied by explicit auditory stimulation. For each presentation, the distance the chick ran was measured and a preference score calculated as the distance run towards the imprinting stimulus (IS) as a percentage of the total distance run during the test (IS +NO): Preference  =  100 × \[IS / (IS + NO)\] For the imprinted group, only chicks with strong preferences for the imprinting stimulus (\>70%: n = 8) were selected for surgical implantation of microelectrodes. Chicks in the naïve group (n = 9) remained in a holding incubator and received no exposure to the imprinting stimulus prior to implantation. Microelectrode Design and Implantation {#s4c} -------------------------------------- Neuronal activity was recorded using four platinum/iridium wires (Neuralynx, Bozeman, MT, USA) wound together and bonded (tetrodes)[@pone.0017777-Gray1]. In order to penetrate the brain, tetrodes were mounted onto thin (dia. 125 µm) tungsten wire (Advent Research Materials, Oxford, UK) with cyanoacrylate superglue. The resultant structure was then fixed into a guide cannula and the protruding end coated in 1,1′-dioctadecyl−3,3,3′,3′-tetramethylindocarbocyanine perchlorate (DiI) (Sigma); a neuronal tracer allowing electrode localization [@pone.0017777-DiCarlo1]. The tetrode tips were then gold-plated to an impedance of 0.2--0.4 MΩ prior to surgery. Chicks were anaesthetized (0.12 ml Equithesin, intraperitoneal)[@pone.0017777-Davies1] and positioned in a sterotaxic frame. A craniotomy was performed 0.8 mm lateral to the midline and 2.5 mm anterior to the frontoparietal suture and the dura mater removed. A microdrive assembly [@pone.0017777-CipollaNeto1] was then glued to the dorsal surface of the skull, allowing one tetrode to be positioned over the left or right IMM. A reference electrode was also placed under the skull permitting differential recording and the assembly was stabilized in dental cement. At the end of surgery, each tetrode was then advanced approximately 1.25 mm over a period of 2.5 hrs. Neuronal recording {#s4d} ------------------ Following recovery from surgery overnight, neuronal activity was detected in the awake animal, placed in a modified running wheel in which a tether connected the microdrive to recording equipment: Signals were amplified 10,000 times, band-pass filtered between 300 and 3,000 Hz (CyberAmp; Axon Instruments, Union City, CA, USA) and sampled at 14 kHz for offline analysis (Power1401 laboratory interface and Spike2; Cambridge Electronic Design, Cambridge, UK). Tetrodes were advanced until spontaneous neuronal activity was detected and chicks were then presented with familiar and unfamiliar visual, auditory and audiovisual stimuli. Detailed accounts of visual and auditory stimuli can be found elsewhere [@pone.0017777-McCabe1], [@pone.0017777-VanKampen1]. Briefly; the visual stimulus was either a red and black box (9×17.5×18 cm; l×w×h), or a blue and white cylinder (diameter, 15.5 cm; height, 19 cm). Both were illuminated from within by 24 W bulbs, rotated at 30 revolutions per minute and placed 65 cm from the running wheel. During stimulus presentation, current was provided to the stimuli to cause illumination and rotation. Between presentations, the stimuli were dim and static and elicited little interest from the animals. Auditory stimuli were maternal calls (Calls A and B) recorded from two hens and presented at approximately 75 dB using a cassette player controlled by a TTL pulse from the Power1401 laboratory interface and a pair of loud speakers placed out of view of the animal. Audiovisual stimuli consisted of all possible combinations of visual and auditory components of the imprinting stimulus and novel object ([Table 1](#pone-0017777-t001){ref-type="table"}). All stimulus presentations lasted four seconds and throughout presentation of visual and audiovisual stimuli, chicks were required to look towards the visual stimulus with both eyes during presentation. This was ensured by monitoring head position via video camera and excluding presentations in which either or both eyes were turned away from the stimulus. Stimuli were presented in a consecutive sequence an average of 15 times and the stimulus order was randomized between chicks. The approach behavior of the chick was also recorded during stimulus presentation and this data was used to confirm the ability of subjects to discriminate between audiovisual as well as visual stimuli: For each four second presentation, the distance run during presentation was measured and preference scores calculated using the mean distance run towards the audiovisual imprinting stimulus and novel object. Following a testing session, tetrodes were advanced at least 200 µm to avoid repeated sampling of the same neurons. Chicks were then returned to holding incubators for at least 45 minutes between tests. In six birds, spontaneous activity was not satisfactorily detected at any depth and therefore only behavioral data from these birds were analyzed. Electrode Localization {#s4e} ---------------------- At the end of the experiment, chicks were euthanized (0.1 ml Euthatal, intraperitoneal) and perfused transcardially with 0.9% saline and 4% paraformaldehyde in PBS (Sigma). The brain was removed and stored in 4% paraformaldehyde in PBS until 24 hours before sectioning, at which point the brain was transferred to 20% sucrose (Sigma). Frozen sections were then cut at 180 µm thickness and tetrode location confirmed by the presence of DiI stained neurons. Data from five subjects were excluded because tetrodes were positioned outside the IMM. Data Analysis {#s4f} ------------- Single units were isolated from recorded data using standard cluster cutting techniques [@pone.0017777-Gray1], [@pone.0017777-Lewicki1]. Briefly; events with amplitudes between two and ten times the background noise on at least one channel of the tetrode were selected by threshold detection (Spike2). Events were then sorted by waveform parameters and principal components using k-means and manual clustering. Events that did not resemble action potentials on at least one channel were discarded. Single unit isolation was assessed using spike interval histograms and visual inspection of waveform shape; the minimum interval between spikes was greater than 2 ms for all neurons. Following isolation, the times of stimulus presentation and spikes were saved and subsequently analyzed in Matlab (Mathworks, Natick, MA, USA). Single-unit responses were then assessed using the normalized response magnitude, calculated as: RM  =  100 × (P/B). Where P is the mean firing rate of a neuron during the 4 s stimulus presentation, and B is the firing rate in the 4 s baseline period before presentation. Mean response magnitudes to the audiovisual imprinting stimulus and novel object, their auditory and visual components and incongruent audiovisual stimuli were compared between imprinted and naive birds in a 2×2 (stimulus × group) analysis of variance (ANOVA; Genstat, VSN International, Hemel Hempstead, UK). Modality replaced stimulus as a factor for comparisons between visual or auditory and audiovisual imprinting stimuli. By comparing naïve and imprinted birds it was possible to control for stimulus salience. By comparing neuronal responses to the imprinting stimulus and novel object or their unisensory components, it was possible to determine whether the effect of imprinting was generalized or specific to the imprinting stimuli previously experienced. Therefore by observing the interaction of group and stimulus, we could exclude the influences of stimulus salience or generalization from our interpretation. Regression and correlation coefficients used to describe the relationship between the magnitudes of responses to audiovisual stimuli and their auditory and visual components were calculated in Matlab. For each neuron we also characterized the integration of auditory and visual information using the additivity index of multisensory integration \[modified from 33\]. In our index, additivity was calculated in two stages, firstly we calculated the corrected the firing rate of a neuron in response to an audiovisual stimulus (AV) by deducting the baseline firing rate before presentation (B~AV~) from the firing rate during presentation (P~AV~). The same corrections were also applied to responses to auditory (A) and visual (V) components of the audiovisual stimulus: AV  =  P~AV~ − B~AV~ A  =  P~A~ − B~A~ V  =  P~V~ − B~V~ Corrected firing rates were then used in the following equation to calculate additivity: Additivity  =  (AV − A − V) / (\|AV\| + \|A\| + \|V\|) Denominator values were made absolute because the combination of positive and negative values (i.e. responses at a rate lower than baseline firing rate) could lead to cancellation that made the total sum of neuronal activity inaccurately low. For each audiovisual stimulus, the mean additivity index was compared between imprinted and naïve birds by t-test. We thank Zoltan Cseresnyes for assistance with confocal microscopy and Gabriel Horn for valuable discussion. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This research was supported by a BBSRC studentship and a research grant from the Balfour Trust fund. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: SMT. Performed the experiments: SMT. Analyzed the data: SMT. Contributed reagents/materials/analysis tools: SMT BJM. Wrote the paper: SMT BJM.

PubMed Central

2024-06-05T04:04:19.874064

2011-3-10

PMC3053394

Introduction {#s1} ============ Paracoccidioidomycosis (PCM) is a systemic granulomatous disease caused by *Paracoccidioides brasiliensis*, a termally dimorphic fungus widespread in Latin America mainly affecting rural workers [@pone.0017885-Visbal1]. PCM is the most prevalent systemic endemic mycosis in South America notably in Brazil, Colombia, Venezuela and Argentina [@pone.0017885-Odds1]. The fatal acute PCM affects the reticule endothelial system, whereas the chronic PCM affects mainly the lung, which shows a granulomatous inflammation with an inefficient cellular immune response [@pone.0017885-Visbal1], [@pone.0017885-SanBlas1]. Antifungal chemotherapy is required to control the disease. The conventional treatment of PCM is based on sulfonamides, amphotericin B and azole derivates. Extended periods of therapy are usually required to warrant a good clinical response and avoid relapses [@pone.0017885-Visbal1]. But, the prolonged time of drugs administration causes frequent self-exclusion of the patient from treatment [@pone.0017885-Visbal1]. The introduction of azoles marked an advance in the treatment of fungal diseases, PCM among them. Azoles act on ergosterol biosynthesis at the C-14-demethylation stage, and the resulting ergosterol depletion and accumulation of 14-methylated sterols interferes with the functions of ergosterol as a membrane component, altering the normal permeability and fluidity of the fungal membrane [@pone.0017885-Odds1]. Imidazoles (ketoconazole) and triazoles (fluconazole, saperconazole, and itraconazole) have been extensively used for the treatment of PCM and have proven effective for clinical purposes, showing fewer side effects than amphotericin B [@pone.0017885-Visbal1]. The above mentioned antifungal antibiotics have drawbacks such as long time of medication (sulfa derivates and azoles), severe renal side effects (amphotericin B), unresponsiveness of some patients to the treatments (all antifungals) [@pone.0017885-Visbal1], [@pone.0017885-Hahn1], [@pone.0017885-Pereira1], [@pone.0017885-Pereira2], [@pone.0017885-deAlmeida1], [@pone.0017885-Paniago1] and high cost (azoles, lipid formulations of amphotericin B). Therefore, the search for new and more effective strategies to conventional chemotherapy for *P. brasiliensis* and other fungal pathogenic species, with fewer or no side effects, continues [@pone.0017885-SanBlas1]. Because of the search for these new alternatives for PCM treatment, several candidate antigen molecules and its mechanisms of protection against *P. brasiliensis* are being studied. Among these molecules, the recombinant protein rPb27 represents an attractive candidate due to its great potential to control this disease as it was demonstrated in previous work [@pone.0017885-Reis1], in which this protein showed a significant degree of protection in the lungs, livers and spleens of mice immunized with it and posteriorly challenged with a virulent strain of *P. brasiliensis*. In this same work it was shown that this protein is component of F0 fraction of this fungus. This fraction had already demonstrated protective ability in experimental PCM [@pone.0017885-Diniz1]. The association of imunotherapeutics and antifungal agents to treat PCM has also been investigated. The administration of the peptide P10, a 15-amino acid peptide identified in the glycoprotein Gp43, that have already shown the capacity to elicit the secretion of Th1 type cytokines [@pone.0017885-Marques1], as an adjuvant to the chemicals used in the PCM therapy, showed an improvement of the therapeutic effectiveness of some antifungal agents [@pone.0017885-Marques1], [@pone.0017885-Marques2]. In this work we evaluated the immunotherapeutic potential of rPb27 immunization with or without fluconazole chemotherapy to treat PCM as well as the cytokines profile and IgG isotypes production induced by this combined treatment in experimental PCM using BALB/c mice. Results {#s2} ======= Cloning, expression and purification of recombinant rPb27 {#s2a} --------------------------------------------------------- Cloning and sequencing of rPb27 DNA resulted in 773 bp comprising the entire rPb27 gene open reading frame and some nucleotides that were added by the plasmid. Nucleotide sequence homology was 100% in relation to the coding region of the 27 kDa *P. brasiliensis* hypothetic protein - access number AA49615 (Data not shown). Nucleotide sequence of this protein was cloned into a pET-DEST42 expression vector where a recombinant protein of approximately 27 kDa was expressed with a his-tag and then purified by affinity cromatography. After this purification it was possible to obtain a single protein with a good yield, as demonstrated by SDS-PAGE and western blotting assays ([Fig. 1](#pone-0017885-g001){ref-type="fig"}). ::: {#pone-0017885-g001 .fig} 10.1371/journal.pone.0017885.g001 Figure 1 ::: {.caption} ###### Purified rPb27 profile. **A**. SDS-PAGE analysis of rPb27 after purification on HiTrap™ Chelating HP (Amersham Biosciences, Uppsala Sweden). Aliquot of 20 µl of purified rPb27 was separated on 10% polyacrylamide gels, under reducing conditions, followed by Comassie-blue staining. **B**. Western blotting analysis of purified rPb27 using mouse IgG anti-his-tag (Amersham Biosciences, Uppsala Sweden). Purified rPb27 was subjected to 10% SDS-PAGE, under reducing conditions, followed by eletrophoretic transfer to nitrocellulose paper. The membrane was incubated with mouse IgG anti-his-tag (1∶100) and revealed with goat anti-mouse IgG conjugated with peroxidase (1: 10000). 1, molecular marker. 2, purified rPb27. The molecular weight (kDa) of molecular marker proteins is showed on the left. :::  ::: Organ CFU from intratracheally infected BALB/c mice immunized with rPb27 and/or treated with fluconazole {#s2b} -------------------------------------------------------------------------------------------------------- To explore the combined effect of rPb27 immunization and fluconazole treatment in BALB/c mice, animals immunization and chemotherapy started 30 days after infection. Analysis of organ CFU was done after 40 and 90 days of treatment. A significant reduction of fungi recovered from lung, spleen and liver of animals (CFU) was obtained in mice immunized with rPb27 and treated with fluconazole at the first time point. In the lung of these animals it was determined, 40 days post treatment, a 60% reduction in the CFUs in relation to infected-only group. Besides, in the liver and spleen of these animals there wasn\'t recovered any fungi colonie at the two time points analyzed. The rPb27 immunization alone failed to reduce fungi recovered from these organs. And the treatment with the antifungal drug alone also failed to reduce the number of CFU in the lung and liver at two time points, and in the spleen after 90 days of treatment despite a reduction after 40 days ([Fig. 2](#pone-0017885-g002){ref-type="fig"} A, B). ::: {#pone-0017885-g002 .fig} 10.1371/journal.pone.0017885.g002 Figure 2 ::: {.caption} ###### Fungal recovery in lung, spleen and liver of infected mice. The CFUs were estimated 40 (A) and 90 days (B) post treatment in organs from mice infected intratracheally with 3×10^5^ *P. brasiliensis* yeast cells and subjected to fluconazole treatment combined or not with rPb27 immunization. Control mice were only infected with *P. brasiliensis* (Infected), adjuvant mice were inoculated with *C. parvum*-Al(OH)~3~ with fluconazole treatment (Adjuvant/T) or not (Adjuvant), and rPb27 mice were immunized with recombinant protein combined to fluconazole treatment (rPb27/T) or not (rPb27). All groups of mice were infected with the same number of yeast cells. Bars represent the Log~10~(UFC/g) means and standard deviations from organs of 3 to 5 animals in each group. \* significant (p\<0,05) difference in relation to the group of mice only infected. :::  ::: In the lungs despite a significant reduction of fungi recovered at the first time point analyzed in mice immunized with rPb27 and treated with fluconazole ([Fig. 2A](#pone-0017885-g002){ref-type="fig"}), after 90 days of treatment this number increased and matched the levels of infected-only group ([Fig. 2B](#pone-0017885-g002){ref-type="fig"}). Lung, spleen and liver histopathology from BALB/c mice vaccinated with rPb27 and/or treated with fluconazole {#s2c} ------------------------------------------------------------------------------------------------------------ Histopathology of lung, liver and spleen sections showed differences in granuloma lesions. Lungs of animals only infected with *P. brasiliensis*, infected and submitted to fluconazole treatment, or infected and immunized with rPb27 ([Fig. 3](#pone-0017885-g003){ref-type="fig"}, [4](#pone-0017885-g004){ref-type="fig"}) showed, at the two time points investigated, giant confluent granulomes with innumerable viable yeast cells of *P. brasiliensis*, and extensive tissue destruction. While in the liver and spleen ([Fig. 3](#pone-0017885-g003){ref-type="fig"}, [4](#pone-0017885-g004){ref-type="fig"}) of these groups it was observed multiple foci of granulomatous inflammation, containing fungal cells, however the group infected and treated with fluconazole and the group immunized with rPb27 presented a decrease on lesions size in the liver after 90 days of treatment ([Fig. 4](#pone-0017885-g004){ref-type="fig"}) in relation to infected-only group. And at the first time point analyzed ([Fig. 3](#pone-0017885-g003){ref-type="fig"}) the group infected and treated with fluconazole didn\'t present any considerable lesion in the liver and spleen. The groups of mice that were innoculated with adjuvant with or without chemotherapy presented similar pattern of lesions than infected-only mice in the three organs analyzed (data not shown). ::: {#pone-0017885-g003 .fig} 10.1371/journal.pone.0017885.g003 Figure 3 ::: {.caption} ###### Representative histopathology of lungs, livers and spleens from infected mice, after 40 days of treatment. BALB/c mice were euthanized 40 days after treatment. The lungs, spleens, and livers were excised, fixed in 10% buffered formalin, and embedded in paraffin for sectioning. The sections were stained with hematoxylin--eosin and examined microscopically. Infected, group only infected with *P. brasiliensis*. Infected/T, same as Infected, but treated with fluconazole. rPb27, group infected and posteriorly immunized with rPb27. rPb27/T, same as rPb27, but treated with fluconazole. In each lung photos, the scale bar represents 427.3 µm, while in each liver and spleen photos, the scale bar represents 56.9 µm. :::  ::: ::: {#pone-0017885-g004 .fig} 10.1371/journal.pone.0017885.g004 Figure 4 ::: {.caption} ###### Representative histopathology of lungs, livers and spleens from infected mice, after 90 days of treatment. BALB/c mice were euthanized 90 days after treatment. The lungs, spleens, and livers were excised, fixed in 10% buffered formalin, and embedded in paraffin for sectioning. The sections were stained with hematoxylin--eosin and examined microscopically. Infected, group only infected with *P. brasiliensis*. Infected/T, same as Infected, but treated with fluconazole. rPb27, group infected and posteriorly immunized with rPb27. rPb27/T, same as rPb27, but treated with fluconazole. In each lung photos, the scale bar represents 416.6 µm, while in each liver and spleen photos, the scale bar represents 55.6 µm. :::  ::: Mice immunized with rPb27 and also treated with fluconazole showed after 40 days of treatment a considerable reduction in the size of lesions at lungs, presenting few pulmonary compact granulomes with no confluence ([Fig. 3](#pone-0017885-g003){ref-type="fig"}). These granulomes were more organized and with a reduced number of yeast cells that all other groups analyzed ([Fig. 5](#pone-0017885-g005){ref-type="fig"}). The lungs of other infected groups (Infected, Infected/T, Adjuvant, Adjuvant/T and rPb27) presented a desorganized granulome with a great number of viable yeast cells of *P. brasiliensis* ([Fig. 5](#pone-0017885-g005){ref-type="fig"}). After 90 days of treatment the group immunized with rPb27 and also treated with fluconazole presented in lungs a pattern of granulome similar to infected group ([Fig. 4](#pone-0017885-g004){ref-type="fig"}, [5](#pone-0017885-g005){ref-type="fig"}), and in the spleen and liver this group didn\'t present any considerable injury at the two time points analyzed ([Fig. 3](#pone-0017885-g003){ref-type="fig"} and [4](#pone-0017885-g004){ref-type="fig"}). ::: {#pone-0017885-g005 .fig} 10.1371/journal.pone.0017885.g005 Figure 5 ::: {.caption} ###### Representative granulome histopathology of lungs from infected mice after 40 and 90 days of infection. BALB/c mice were euthanized 40 and 90 days after treatment. The lungs, spleens, and livers were excised, fixed in 10% buffered formalin, and embedded in paraffin for sectioning. The sections were stained with hematoxylin--eosin and examined microscopically. Infected, group only infected with *P. brasiliensis*. Infected/T, same as Infected, but treated with fluconazole. rPb27/T, group infected and posteriorly immunized with rPb27 and treated with fluconazole. In each photo, the scale bar represents 25.9 µm. :::  ::: Recombinant rPb27 specific immunoglobulin responses in infected mice {#s2d} -------------------------------------------------------------------- The specific antibody response to rPb27 in infected mice were evaluated by ELISA. After three rounds of vaccination, significant levels of anti-rPb27 IgG were detected in the sera of mice immunized with rPb27, treated or not with fluconazole. The other groups infected with *P. brasiliensis* didn\'t present significant antibodie levels for this protein, as well as the group without any intervention (data not shown). Significant levels of anti-rPb27 specific IgG1, IgG2a and IgG2b isotypes were detected in the sera of mice immunized with rPb27 treated or not with fluconazole compared to sera from noninfected mice after 40 ([Fig. 6A](#pone-0017885-g006){ref-type="fig"}) and 90 days of treatment ([Fig. 6B](#pone-0017885-g006){ref-type="fig"}), while significant levels of IgG3 were found only in the group rPb27 after 40 ([Fig. 6A](#pone-0017885-g006){ref-type="fig"}) and 90 days of treatment ([Fig. 6B](#pone-0017885-g006){ref-type="fig"}). The other groups infected with *P. brasiliensis* (Infected, Infected/T, Adjuvant, Adjuvant/T) didn\'t present significant levels of these isotypes (data not shown). ::: {#pone-0017885-g006 .fig} 10.1371/journal.pone.0017885.g006 Figure 6 ::: {.caption} ###### IgG isotypes production against rPb27 by infected and immunized mice. Antibody response against rPb27 in mice infected and immunized with this recombinant protein associated or not with fluconazole chemotherapy was determined by ELISA assay after 40 (A) and 90 (B) days of treatment. Control, mice without any intervention. rPb27, group infected and posteriorly immunized with rPb27. rPb27/T, same as rPb27, but treated with fluconazole. Bars represent the means and standard deviations of optical density (O.D.) at 1∶400 serum dilution in each experimental group (n = 3). \* significant (p\<0,05) difference in relation to the control group. \# significant (p\<0,05) difference in relation to the rPb27 group. :::  ::: Discussion {#s3} ========== Considering the increase in the worldwide incidence of fungal infections, the development of new therapies for these diseases has become of great importance to public health [@pone.0017885-Chakrabarti1]. The current therapy to treat mycosis is based on polienes and azoles, depending on the severity of the infection [@pone.0017885-ShikanaiYasuda1]. To treat severe cases of PCM amphotericin B followed by itraconazole and sulfametoxazole are indicated. The main disadvantage of using amphotericin B is its occasional toxicity. In the case of itraconazole or sulfametoxazole, the long period of treatment required may cause patients to quit medication, possibly leading to recurrence of disease [@pone.0017885-ShikanaiYasuda1]. Given these dificulties, new approaches to the treatment of systemic fungal infections need to be developed. Alternative strategies to conventional chemotherapy for fungal diseases have been explored. Combined therapy of immunotherapeutics and antifungal agents in the treatment of PCM has been investigated, and have demonstrated a great potential to enhance antifungal effect and to prevent relapses [@pone.0017885-Marques1], [@pone.0017885-Marques2]. Here we reported the efficacy of rPb27 immunization combined with fluconazole chemotherapy in the treatment of PCM. rPb27 is a recombinant protein that was firstly described by McEwen and coworkers [@pone.0017885-McEwen1], and have demonstrated a great potential in immunodiagnosis of PCM [@pone.0017885-Diez1], [@pone.0017885-Ortiz1], [@pone.0017885-Ortiz2] and as a vaccine candidate against this mycosis, as demonstrated in previous work, in which the immunization with this protein showed a significant degree of protection in the lungs, livers and spleens of mice posteriorly challenged with a virulent strain of *P. brasiliensis* [@pone.0017885-Reis1]. In this work the expression and purification of this protein showed a single protein with approximatelly 27 kDa of molecular mass. After 40 and 90 days of treatment the fungal load was examined in lungs, livers and spleens of infected mice. The pattern of CFU in these organs showed an additive effect of rPb27 immunization and drug treatment. In lungs at the first time point analyzed this combined treatment reduced in 60% the CFU number in relation to control groups (untreated infected animals and those that received adjuvant), however, this CFU number increased after 90 days of treatment reaching the Infected group levels. This relapsing effect in lungs can be assigned to the short time of treatment. In the liver and spleen the combined treatment reduced in 100% the CFU number in comparison to control groups, at the two time points analyzed. Showing that, in these organs the combined administration of rPb27 and fluconazole controlled the relapsing effects. Histopathological analysis of infected mice confirmed the therapeutic effect of rPb27 immunization combined with fluconazole chemotherapy, by leading to fungal elimination, reaching sterility in spleen and liver despite the short time of treatment. This sterility wasn\'t reached by other authors using similar strategies of therapeutic vaccines [@pone.0017885-Marques1], [@pone.0017885-Marques2], [@pone.0017885-Ribeiro1]. In the mice lungs rPb27 immunization with fluconazole chemotherapy reduced CFU in this organ, at the first time point analyzed. In addition, histopathological analyses of the group that received this combined treatment showed less lung tissue compromised 40 days after treatment, presenting large preserved areas with compact, well organized granulomes containing few *P. brasiliensis* yeast cells. But after 90 days of treatment the lungs of this group showed an increase in lung tissue compromised and its granulomes reached the extension and organization of those from control groups, which validates the CFU assay results. Preliminary cytokine screening showed that animals that received rPb27 immunization with fluconazole chemotherapy presented high levels of TNF-α when compared with only-infected group after 40 days of treatment (data not shown). This result conforms with the literature that TNF-α has a protective role in *P. brasiliensis* infection and may have contributed to the reduced size of lung tissue compromised in addition to a decrease on CFU numbers and to a better organization of granulomes when compared to only-infected group. TNF-α is a Th1 cytokine that according to previous work has a great importance to control dissemination and growth of the fungi, and inflammatory response in PCM [@pone.0017885-Souto1]. These authors used mice with homologous disruption of the TNF-α receptor p55 infected with *P. brasiliensis* and demonstrated that these animals, but not wild-type mice, were unable to control the growth of yeast cells and the mice succumbed to infection by day 90 after infection. Besides, inflammatory granulomas weren\'t found in these knockout animals. These findings showed that TNF-α mediate resistance to *P. brasiliensis* infection [@pone.0017885-Souto1]. In fact, the granulomatous inflammatory reaction, a specialized and efficient tissue response against certain parasites [@pone.0017885-Adams1], *P. brasiliensis* among them, requires TNF-α which is supposed to be responsible for attracting and activating effector cells as well as for macrophage accumulation and differentiation [@pone.0017885-Tracey1]. Besides this, a recent work showed that spleen cells from infected female mice produced more TNF-α than those from males in response to paracoccin stimulus, which contributes to the greater resistance to PCM presented by female in relation to male mice [@pone.0017885-Pinzan1]. Summarizing, the rPb27 immunization combined to fluconazole chemotherapy was efficient to avoid fungal dissemination to spleen and liver of mice and to control yeast cells growth in the lungs after 40 days of treatment, but this control in lungs was lost after the suspension of chemotherapy, 90 days post treatment, this can be due to the short time of treatment, only thirty days. In the literature it was mentioned that resistance to *P. brasiliensis* infection is determined mainly by the ability of the host to restrict fungal dissemination rather than control fungal growth at the primary site of infection [@pone.0017885-Cano1]. This containment of the fungus in the lungs can be the first step to PCM control, and, possibly, with a larger time of chemotherapy this disease could be controlled. The specific antibody levels was also evaluated. We demonstrated that immunization of BALB/c mice with rPb27 induced significant levels of all isotypes evaluated after 40 and 90 days of treatment. When this immunization was combined with fluconazole chemotherapy mice developed a pulmonary-restricted PCM associated with low mortality rates and production of significant levels of IgG1, IgG2a and IgG2b isotypes after 40 and 90 days of treatment, with a small production of IgG3 only 40 days after treatment. The presented data showed that the immunization with rPb27 promoted enhanced antifungal protection by fluconazole chemotherapy, decreasing fungal load in lungs after 40 days of treatment and avoiding fungal dissemination to other sites of infection (liver and spleen). Thus, rPb27 can be explored as an adjuvant for PCM therapy. Materials and Methods {#s4} ===================== Ethics Statement {#s4a} ---------------- All animals were handled in strict accordance with good animal practice as defined by the relevant national and local animal welfare bodies, and all animal work was approved by the Ethics Committee on Animal Experimentation (CETEA) from the Universidade Federal de Minas Gerais (UFMG). The protocol number is 24/2006. Animals used in this work {#s4b} ------------------------- Adult male BALB/c mice (6--8 weeks old) were purchased from Centro de Bioterismo, ICB-UFMG (Belo Horizonte, MG, Brazil), and maintained under standard laboratory care as previously described [@pone.0017885-Diniz1]. *P. brasiliensis* strain {#s4c} ------------------------ Virulent *P. brasiliensis* human isolate was obtained from a patient with active PCM, whose case was reported by Araujo and coworkers [@pone.0017885-AraujoSde1], maintained in YPD agar medium \[0.5% yeast extract, 0.5% peptone, 1.5% D-glucose, 1.5% agar, pH 7.0\] (Sigma, St. Louis, MO, USA) at 36°C and collected on the seventh day of culture. The viability of fungal suspensions determined by staining with Janus Green B vital dye method [@pone.0017885-Dias1] (Merck. Darmstadt, Germany) was always higher than 90%. The virulence of the human isolate was checked in each experiment by infecting intratracheally BALB/c mice and recovering the yeast cells from their organs. Subcloning and sequencing of rPb27 DNA {#s4d} -------------------------------------- The sequence of the recombinant rPb27 was already cloned by our group in previous work into the expression vector pGEX 4T-2 (GIBCO BRL), which produces a recombinant protein fused to glutathione S-transferase (GST) [@pone.0017885-Reis1]. In order to facilitate the purification procedure and avoid the contamination of purified protein with thrombin, the rPb27 sequence was transferred to the expression vector pET-DEST 42 (Invitrogen, Carlsbad, USA) which express the recombinant protein with a C-terminal His-tag. The cloning procedure was based on Gateway® technology (Invitrogen, Carlsbad, USA), that uses an strategy based on the recombinational properties of bacteriophage lambda [@pone.0017885-Landy1]. The primer set for amplification of the rPb27 sequence included the forward primer: 5′-GGGG ACA AGT TTG TAC AAA AAA GCA GGC TTC GAA GGA GAT AGA ATG GCA CGA GCG CTC AGT TC -3′ and reverse primer: 5′- GGG GAC CAC TTT GTA CAA GAA AGC TGG GTC GTT GTG GAA GAC AGC GCT GCA -3′. The PCR annealing temperature was 56°C. The PCR products were separated in 1% agarose gel. The blunt-end PCR products were then cloned into pDONR 221 vector according to the manufacturer\'s protocol (Invitrogen, USA). The reaction mixture was incubated overnight at 25°C. The reaction was then stopped with 10 minutes Proteinase K incubation. After this ste, competent *E. coli* (TOP10) was transformed by electroporation with the pDONR 221/rPb27 construct according to the manufacturer\'s protocol (Invitrogen, USA). Positive clones were selected on (LB) medium containing 50 µg/mL kanamycin. Plasmid DNA was isolated using the high pure plasmid extraction kit, Mini-prep QIAprep Spin miniprep 150 (Qiagen, Hilden, Germany). The presence of the insert was confirmed by PCR and, finally, to confirm the fidelity of the sequence, DNA sequencing was performed. This construct is called the entry clone. The LR recombination reaction was then carried out between the entry clone and destination vector, pET-DEST42, according to the manufacturer\'s instructions (Invitrogen, USA). Competent *E. coli* (BL21) were transformed by electroporation with the products of LR recombination according to the manufacturer\'s protocol. Positive clones were analysed by culturing them on LB medium containing 100 µg/mL ampicillin. The recombinant plasmid pET-DEST42/rPb27 was checked for accurate insertion by restriction enzyme analysis, and the inserted fragment was sequenced on a MegaBACE DNA Analysis System (Amersham Biosciences, Buckinghamshire, England) [@pone.0017885-Sanger1]. The sequence homology with rPb27 was analyzed using the algorithm ClustalW available on the Internet: <http://www.ebi.ac.uk/clustalw/>. Expression in *E. coli* and purification of rPb27 {#s4e} ------------------------------------------------- The recombinant protein, rPb27, was expressed in *E. coli* using the expression vector pET-DEST42 (Invitrogen, USA), which produces a recombinant protein with a C-terminal his-tag. Purification of the recombinant protein was therefore undertaken by HiTrap^TM^ Chelating HP according to the manufacturer\'s instructions (Amersham Biosciences, Uppsala Sweden). Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and Western blotting {#s4f} ----------------------------------------------------------------------------------------- Purified rPb27 was subjected to continuous electrophoresis using SDS 10% polyacrylamide gels under reducing conditions [@pone.0017885-Laemmli1]. The separated proteins were stained with Coomassie blue or transferred to a nitrocellulose membrane [@pone.0017885-Towbin1], blocked with 1.6% of casein in phosphatebuffered saline (PBS), pH 7.4, for 1 h at room temperature and then incubated with mouse IgG anti-his-tag (Amersham Biosciences, Uppsala Sweden) diluted 1∶100 in 0.25% casein-PBS for 1 h at room temperature. After this step the membranes were incubated with specific secondary antibody peroxidase-conjugated diluted 1∶10000 in 0.25% casein-PBS for 1 h at room temperature and developed with diaminobenzidine (0.6 mg/ml) in PBS, pH 7.4, plus H~2~O~2~ (1 µl/ml) and NiCl~2~ (10 µl/ml). The reaction was stopped with deionized water. Intratracheal infection of BALB/c mice {#s4g} -------------------------------------- Male BALB/c mice were inoculated intratracheally with 3×10^5^ viable yeast cells of virulent *P. brasiliensis*, grown on YPD-agar and suspended in sterile PBS. Briefly, mice were anesthetized i.m. with 40 µl of a solution containing 57% of ketamine (Dopalen, Vetbrands, Brazil) and 43% of xylazine (Dopaser, Laboratório Calier do Brasil LTDA, Brazil); after approximately 10 min, their necks were hyper-extended, and the trachea was exposed at the level of the thyroid and injected with 3×10^5^ yeast cells in 50 µl of PBS using a 30-gauge needle. The incisions were sutured with 4-0 silk. Immunization of mice with rPb27 {#s4h} ------------------------------- After 30 days of infection, groups of male BALB/c mice were immunized by subcutaneous injection of 50 µg rPb27 in the presence of 100 µg *Corynebacterium parvum* and 1 mg aluminum hydroxide, Al(OH)~3~, as an adjuvant. Animals were boosted three times, at two week intervals, with the same amount of antigen. Control mice were inoculated with adjuvant. Chemotherapy of infected mice {#s4i} ----------------------------- After 30 days of infection animals were treated for 30 days during which groups of mice received doses every 24 h of fluconazole 10 mg/Kg (Mantena laboratories limited). All drug administration were intraperitoneal. Groups of mice studied {#s4j} ---------------------- Male BALB/c mice were divided in seven groups of 12 animals each: Immunized with rPb27 treated with fluconazole (rPb27/T) or not (rPb27); injected with *C. parvum*-Al(OH)~3~ as an adjuvant fluconazole treated (Adjuvant/T) or not (Adjuvant); infected only (Infected); infected and treated with fluconazole (Infected/T); and a control group without any intervention (Control). Fungal recovery in organs of infected mice {#s4k} ------------------------------------------ Organ colony-forming units (CFUs) were evaluated after 40 and 90 days of treatment in the lungs, spleens and livers, which were removed, weighed and homogenized in PBS. The final suspension was placed on brain heart infusion (BHI, Difco) agar supplemented with 4% fetal calf serum and 5% spent culture medium of *P. brasiliensis* as a growth factor. Gentamycin was added at 40 mg/l. The plates were incubated at 36°C and read after 20 days. The results were expressed as the number of log~10~ of viable *P. brasiliensis* CFUs per gram of tissue per mouse. Enzyme-linked immunosorbent assay (ELISA) {#s4l} ----------------------------------------- ELISA plates were coated overnight at 4°C with 1 µg/100 µl of purified rPb27 recombinant protein in 0.05 M carbonate/bicarbonate buffer, pH 9.6. Wells were blocked for 1 h at 37°C with 1.6% casein in PBS solution. Plates were then incubated with 100 µl of anti-rPb27 mouse serum or non-immunized mouse serum samples diluted 1∶400 in 0.25% casein in PBS for 1 h at 37°C. Washes were performed with PBS-T20 (0.15 M PBS, pH 7.4; 0.1% Tween-20) between incubations. For each well, 100 µl of peroxidase-conjugated goat anti-mouse IgG, IgG1, IgG2a, IgG2b or IgG3 antibodies (Southern Biotechnology Associates, Inc., Birmingham, AL, USA) diluted 1∶5000 were added and incubated for 1 h at 37°C. After additional washes, peroxidase activity was assayed with 100 µl of TMB ELISA substrates solution (Thermo Scientific Pierce). Color development was stopped with 50 µl of 2N H~2~SO~4~. The optical density at 450 nm was measured with an automated ELISA reader (ELX 800 BIO-TEK Instruments Inc.). Histopathology of lung, spleen, and liver of experimental groups {#s4m} ---------------------------------------------------------------- BALB/c mice were euthanized 40 and 90 days after treatment. The lungs, spleens, and livers were excised, fixed in 10% buffered formalin, and embedded in paraffin for sectioning. The sections were stained with hematoxylin--eosin and examined microscopically [@pone.0017885-Reis1]. Statistical analysis {#s4n} -------------------- Data were analyzed statistically by one-way ANOVA followed by the Bonferroni test or Student t-test associated when necessary to the non-parametric Mann--Whitney test, with the level of significance set at p\<0.05. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was supported by Fundação de Amparo à Pesquisa de Minas Gerais (FAPEMIG), Conselho Nacional de Desenvolvimento Cientifico e Tecnologico (CNPq) and Pró-reitoria de Pesquisa da Universidade Federal de Minas Gerais. Viviane C. Fernandes is a PhD student with a scholarship from CNPq. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: VCF AMG. Performed the experiments: VCF EMNM JNB JBC. Analyzed the data: VCF AMG. Contributed reagents/materials/analysis tools: VCF AMG RS. Wrote the paper: VCF AMG.

PubMed Central

2024-06-05T04:04:19.877313

2011-3-10

PMC3053395

Introduction {#s1} ============ The grasses, the approximately 10,000 species in the family Poaceae, are one of the most ecologically and economically significant taxa on the planet. Comparative mapping of diverse grass species led to the conclusion that they are all similar in gene content and order [@pone.0017855-Gale1], [@pone.0017855-Moore1] to the point that it was argued grasses could be treated as a single genetic system, sharing map data, markers, and leveraging inter-specific hybrids to dissect the genes responsible for morphological variation between different grass lineages [@pone.0017855-Bennetzen1]. In other words, knowledge gained from the study of any one grass species could be quickly and directly applied to all other species in the family. Among the grasses, maize is without question the species with the longest and most comprehensively documented history of genetic investigation. The rich genetic resources found in maize are the result of over a century of genetic investigation beginning with R. A. Emerson\'s small but distinguished group in the early 20^th^ century; see B. McClintock\'s unpublished note on this group [@pone.0017855-McClintock1]. The resulting set of characterized genes has the potential to be of great value in the genomics era and sets maize apart from many model systems of more recent origin. Until now the applications of this information in a genomic context have been severely limited by the lack of reliable connections between the data produced by geneticists studying individual genes and the datasets produced by genomicists who generally work at the level of whole genomes. We curated a dataset of 464 "classical" maize genes supported by citations from at least three publications, mutant phenotype data, or direct requests from the maize community using data presented in MaizeGDB: The Maize Genetics and Genomics Database (<http://www.maizegdb.org>) [@pone.0017855-Lawrence1], [@pone.0017855-Lawrence2]. Using manual annotation we connected these well characterized maize loci to gene models created by maizesequence.org, the group that recently published a sequence of the maize genome. To increase the utility of this dataset we also identified orthologous genes at syntenic locations in the genomes of three other grass species with published genomes: rice [@pone.0017855-Goff1], sorghum [@pone.0017855-Paterson1], and brachypodium [@pone.0017855-The1]. The evolutionary relationships of these grass species and a number of other notable grasses are shown in [Figure 1](#pone-0017855-g001){ref-type="fig"}. This initial classical gene list was distributed to the maize community with links to software that graphically presented our pan-grass synteny data and links to the MaizeGDB locus pages where all data regarding individual maize genes is archived. ::: {#pone-0017855-g001 .fig} 10.1371/journal.pone.0017855.g001 Figure 1 ::: {.caption} ###### Phylogenetic relationships of notable and sequenced grass species. Branch lengths not to scale. \*The genome sequencing of foxtail millet by the joint genome institute is complete, but has not yet been published. Therefore it is not included in our analyses (SI 1). \*\*Projects to sequence the genomes of barley and wheat are announced or in progress. :::  ::: The maize lineage, a branch that included both *Zea* and *Tripsacum*, experienced a whole genome duplication an estimated 5--12 million years ago [@pone.0017855-Bomblies1]--[@pone.0017855-Schnable1]. This duplication created two homeologs (syn. homoeologs, ohnologs, syntenic paralogs) co-orthologous to single copy genes in other, unduplicated, grass species. The nearest unduplicated outgroup species with a sequenced genome is *Sorghum bicolor*. For many genes, the two duplicated copies were functionally redundant and one copy or the other has been lost from the genome of modern maize by an intrachromasomal recombination deletion mechanism [@pone.0017855-Woodhouse1]. Pairs of chromosomes orthologous to each of the ten chromosomes of sorghum can be reconstructed within the maize genome [@pone.0017855-Wei1]. In all ten cases, one chromosome copy in maize has lost a significantly greater proportion of genes conserved syntenically in rice and sorghum across its entire length, and these chromosome copies are grouped together into the maize2 subgenome, while the chromosome copies that experienced lower rates of post-tetraploidy gene loss are grouped together into the maize1 subgenome [@pone.0017855-Schnable2]. Here we show that the genes of interest to maize geneticists are much more likely to be syntenically conserved across all grasses than the average gene supported by full length cDNA evidence. We also found that maize genes identified by a mutant phenotype are disproportionately found on maize1. The bias is true both for genes with a retained duplicate from the whole genome duplication, and singletons whose duplicate copies have been deleted. This finding was predicted by our previously published hypothesis that deletions of duplicate gene copies from the maize1 subgenome are more likely to impact fitness than deletions of copies of the same genes from maize2, as maize1 genes tend to be expressed at higher levels than their duplicates on maize2 [@pone.0017855-Schnable2]. We provide all our data on gene locus to gene model mapping, and identification of orthologous genes in other grasses and the homeologous gene in maize, if present, locations in the hopes that these data will be of use to others in the research and teaching community ([Supplemental Information S1](#pone.0017855.s003){ref-type="supplementary-material"}). Results {#s2} ======= Comparing gene models of individually cloned genes to gene models released by the maize genome sequencing consortium {#s2a} -------------------------------------------------------------------------------------------------------------------- Manual mapping of experimentally validated genes to the maize genome provided a chance to error-check the version\_2 gene models released by maizesequence.org. Overall most gene models agreed with previously cloned gene model data ([Supplemental Information S1](#pone.0017855.s003){ref-type="supplementary-material"}). Aside from missed UTR exons and the genes which were classified as supported only by *ab initio* prediction despite being supported by sequences in GenBank, the most frequent error we identified were genes that had been split into multiple unlinked gene models by maizesequence.org. This generally resulted from apparent mistakes in the ordering of contigs within BACs. The overall error rate was substantially reduced in the B73\_refgen2 release, which increased the percent of contigs with order and orientation information from 30 to 80% [@pone.0017855-Wei2]. However this form of error remains present in version 2. For example the coding sequence of the gene *aspartate kinase-homoserine dehydrogenase1* is split into three separate gene models ([Figure 2A](#pone-0017855-g002){ref-type="fig"}). ::: {#pone-0017855-g002 .fig} 10.1371/journal.pone.0017855.g002 Figure 2 ::: {.caption} ###### Examples of manually identified errors in maize gene annotations. Graphics from GEvo comparative sequence alignment tool. Annotated cDNAs from GenBank are compared to regions of the maize B73\_refgen2 genome. Features on the forward strand are displayed above the dotted line, and features on the reserve strand are displayed below the line. Grey lines mark the extent of gene models with CDS sequences in green and UTR sequences in blue. Orange bars mark the gaps between assembled contigs of the maize genome (stretches of N\'s). Red boxes connected by lines show sequences identified as homologous by blastn. A. A comparison of the coding sequence of *aspartate kinase-homoserine dehydrogenase1* to the region of maize chromosome 4 that contains the three gene models --from left to right, GRMZM2G365423, GRMZM2G389303, and GRMZM2G437977 \-- among which the exons of this gene have been divided. An interactive version of this graphic can be regenerated in GEvo using the following link: <http://genomevolution.org/r/25xh> B. A comparison of *cytokinin oxidase1* to GRMZM2G146644, a gene model which includes the 5\' and 3\' ends of *cko1* but has also incorporated unrelated exons from another maize genome contig. Regenerate analysis: <http://genomevolution.org/r/25s5> C. The coding sequence of ferredoxin homeolog2 which maps to a region of the maize genome annotated as the 3\' UTR of GRMZM2G147266. Regenerate analysis: <http://genomevolution.org/r/25s7>. :::  ::: The most dramatic example of an erroneous gene model is provided by *cytokinin oxidase1*, where the 5\' and 3\' regions of the coding sequence mapped to the same gene model -- GRMZM2G146644 -- but the gene model included apparently unrelated exons from a contig inserted between the two ends of *cytokinin oxidase1* ([Fig. 2B](#pone-0017855-g002){ref-type="fig"}). In an additional two cases -- *male sterile45* and *ferritin homolog2* \-- the entire CDS of a gene mapped to regions annotated as UTR ([Figure 2C](#pone-0017855-g002){ref-type="fig"}). We provide proofing links in our master classical maize gene list so that a researcher can immediately visualize obvious annotation problems using the GEvo comparative genomics tool (a CoGe application) used to generate [Figure 2](#pone-0017855-g002){ref-type="fig"} ([Supplemental Information S1](#pone.0017855.s003){ref-type="supplementary-material"}) [@pone.0017855-Lyons1]. Comparing human to computational identification of maize genes using known sequences {#s2b} ------------------------------------------------------------------------------------ Subsequent to the February, 2010 release of our initial version of classical maize gene list to the maize genetics community, maizesequence.org released a list of gene models mapped to named loci in the MaizeGDB database using the Xref computational pipeline (<http://www.maizesequence.org/info/docs/namedgenes.html>). Comparing their machine-annotated dataset to our version 2 list, we identified 152 cases of overlapping assignment of classical maize genes and named maize genes ([Supplemental Information S1](#pone.0017855.s003){ref-type="supplementary-material"}). The remaining 316 classical maize genes identified by manual annotation were not caught by the computational pipeline. In 140 of the overlapping cases, both lists assigned loci to the same gene model. The remaining 12 cases were further investigated using multiple independent GenBank records, as well as genetic location data recorded on MaizeGDB locus pages. In two cases the Xref assignment was clearly correct and the appropriate corrections were made to our list. In nine cases sequence and genetic location data supported the manual assignment over that of Xref. No conclusion could be reached in the final case. Identification of orthologs of classical maize genes in other grasses {#s2c} --------------------------------------------------------------------- The current release of the maize genome -- B73\_refgen2 -- contains over 110,000 annotated genes, many of which have already been identified as gene fragments or genes encoding transposon related proteins. To develop a subset of genes comparable to our classical gene list we adopted an approach used previously [@pone.0017855-Eveland1] restricting ourselves to the subset of annotated maize genes supported by sequenced full length cDNA evidence (see [Methods](#s4){ref-type="sec"}) [@pone.0017855-Alexandrov1], [@pone.0017855-Soderlund1]. In total we identified 34,579 genes supported by full length cDNAs including 81.9% of the unique genes on our classical maize gene list and 75% of the unique genes which were originally identified by a visible mutant phenotype. Using the online syntenic analysis tool SynMap [@pone.0017855-Lyons2], we found that, compared to the average maize gene supported by full length cDNA evidence, classical maize genes, including those with known mutant phenotypes, are much more likely to possess conserved homologs at orthologous syntenic locations -- true orthologs \-- in *Japonica* rice, sorghum, and brachypodium ([Figure 3](#pone-0017855-g003){ref-type="fig"}). ::: {#pone-0017855-g003 .fig} 10.1371/journal.pone.0017855.g003 Figure 3 ::: {.caption} ###### Syntenic conservation of the classical maize genes in other grasses. Comparison of the proportion of genes identified by a mutant phenotype prior to cloning (N = 111), all classical maize genes (N = 464), and all maize genes supported by full length cDNA evidence (N = 34579) for which syntenic orthologs could be identified in the other three grass species with sequenced genomes: sorghum, rice, and brachypodium. :::  ::: Distribution of classical maize genes and mutant phenotype genes between subgenomes {#s2d} ----------------------------------------------------------------------------------- The maize genome is comprised of two subgenomes maize1 and maize2 [@pone.0017855-Schnable2]. Each subgenome is orthologous to the entire genomes of sorghum, rice, and brachypodium. These other grass genomes have remained unduplicated since the radiation of the grasses. The two subgenomes are distinguished by expression of retained duplicate genes and gene loss rates. Maize1 genes tend to be expressed at higher levels than their retained homeologs on maize2, and maize2 has lost copies of more genes syntenically retained in other grass species than maize1 [@pone.0017855-Schnable2]. The distribution of syntenically retained classical maize genes between the two subgenomes of maize roughly mirrors that of all syntenically retained genes supported by full length cDNA evidence. [Figure 4](#pone-0017855-g004){ref-type="fig"} plots these data for all 34,579 genes supported by full length cDNA evidence, the 468 genes of the classical gene list, and the subset of 102 genes on the classical gene list identified by mutant phenotype prior to cloning. Given the bias towards greater expression of maize1 homeologs, the slight bias towards higher numbers of maize1 genes with retained homeologs among genes supported by full length cDNA evidence was expected, but this finding is not of significant interest. However, among syntenically retained genes which were first identified by a visible mutant phenotype, the bias towards the maize1 subgenome is significantly greater than for the classical maize gene list as a whole (p = .028, Fisher Exact Test), and members of homeologous gene pairs located on maize1 were twice as likely as the duplicate copies on maize2 to be originally identified by mutant phenotype \-- 29 maize1 genes with homeologs vs. 14 maize2 genes with homeologs (significantly different from a 50/50 split p = .0222, Chi-square test). ::: {#pone-0017855-g004 .fig} 10.1371/journal.pone.0017855.g004 Figure 4 ::: {.caption} ###### Distribution of classical maize genes between the two maize subgenomes. Comparison of the distribution of genes retained syntenically in at least one other grass species between the two subgenomes of maize as well as whether genes possess retained homeologs from the maize whole genome duplication. For syntenically retained maize genes with full length cDNA support N = 17956. For the subset of the classical maize gene list that are syntenically retained N = 429. For the subset of genes that were first identified by mutant phenotype and are syntenically retained N = 102. :::  ::: Discussion {#s3} ========== The benefits of manual gene annotation {#s3a} -------------------------------------- Our manual proofing of the classical maize gene list shows that, as tempting as it may be to rely primarily on inexpensive *in silico* annotation techniques, manual structural annotation provided a significant amount of important information to B73\_refgen2. Tools are available that allow interested researchers to proof and improve the structural annotations of their favorite genes [@pone.0017855-Wilkerson1]. Having those improvements incorporated into official genome annotations would benefit the entire community. Syntenic conservation of classical maize genes {#s3b} ---------------------------------------------- The idea that genetic collinearity among the grasses could be used to accelerate the research across the whole family is a venerable one [@pone.0017855-Gale1], [@pone.0017855-Moore1], [@pone.0017855-Bennetzen2]. Enthusiasm for this concept of treating the grasses as a single genetic system waned as the sequencing of multiple grass genomes demonstrated that a significant fraction of transcribed genes are not syntenically retained across species, limiting the benefits of cross-species mapping and trait dissection. Our finding that 37% of maize genes supported by full-length cDNA are not retained at a syntenic position in other grass species, and almost 50% of cDNA supported genes apparently inserted into their present locations prior to divergence of the BEP clade, represented by both rice and brachypodium, is in agreement with previous studies. Research in arabidopsis, using papaya as an outgroup, estimated that half of all annotated genes in that species belonged to a "gray" genome of genes which had transposed into nonsyntenic positions within the last 70 million years [@pone.0017855-Freeling1]. A recent study in Drosophila found that knockouts of recently inserted -- within the last 35 million years -- and ancient syntenically conserved genes produced lethal phenotypes at statistically similar rates [@pone.0017855-Chen1]. Genes belonging to the gray genome of maize are essentially unexplored. The genes of greatest interest historically seem to be precisely those that are retained in the same syntenic position in the genomes of all grass species. It may be that, in plants, genes essential for day to day function, such as those involved in key biochemical and developmental pathways, are by definition less likely to transpose or, when they transpose, are less likely to rise to fixation within a species. A small but significant number of mutant genes in maize were identified using map-based cloning approaches relying on rice synteny, prior to the publication of the maize genome. While map-based cloning and comparison of maize to rice certainly did occur, we think it unlikely that this explanation accounts for the magnitude of our results. The techniques used in this paper allowed us to identify with high confidence, lost or transposed genes by first identifying a predicted orthologous syntenic location in the target grass genome. Even the genes which are not retained in all species can be a starting point for hypothesis driven research, a use we support via Gevo links to enable quick visual comparisons of orthologs or predicted locations in multiple grass species ([Supplemental Information S1](#pone.0017855.s003){ref-type="supplementary-material"}). For example, c1 and pl1 are two homeologous maize genes that regulate the biosynthesis of anthocyanin. Both genes have been studied extensively by the maize genetics community. A syntenic co-ortholog of the two genes is retained in the genomes of both sorghum and rice. However the gene is absent from orthologous region of the brachypodium genome ([Figure S1](#pone.0017855.s001){ref-type="supplementary-material"}) which prompted us to investigate further and find the gene was not present anywhere in the brachypodium genome ([Figure S2](#pone.0017855.s002){ref-type="supplementary-material"}). We conclude from this brief research foray that this portion of the anthocyanin biosynthetic regulatory pathway may be significantly different or completely absent in brachypodium, opening avenues for further research. Increased bias towards the maize1 subgenome of mutant phenotype genes {#s3c} --------------------------------------------------------------------- A bias towards maize1 for the classical maize genes was expected given the greater total number of retained genes present in that subgenome. However, when we examined the subset of the classical maize gene list identified by a mutant phenotype prior to cloning, the bias of this dataset towards the dominant subgenome -- maize1 -- was significantly greater than could be explained by the difference in total gene numbers between the two subgenomes. Interestingly this bias is also statistically significant for genes with a retained homeolog on the opposite, homologous subgenome, maize2. Since there is one gene copy present in each subgenome for this class of gene, *a priori* evidence of gene function, the expectation was that mutations of either copy would be about equally likely to produce a mutant phenotype. This was not the case. Rather, our finding that maize1 is the preferred location of genes with mutant phenotypes even when a homeologous duplicate is present suggests that the loss of maize1 copies may be more likely to result in visible impacts of the sort which might catch the eye of researchers, or farmers, in the field. As impacts on plant morphology visible to researchers are likely to have a pronounced impact on plant fitness, this finding is certainly consistent with our previously published hypothesis that the deletion of a gene from maize1 is more likely to be selected against than the deletion of the same gene from maize2 [@pone.0017855-Schnable2]. The corollary is even more interesting: knockout phenotypes do not appear to be behaving as if gene function was buffered by a duplicate copy of the same gene expressed in the same cells. For the moment, our working hypothesis is that maize1 gene copies have predominantly retained the ancestral function of the gene in the pre-duplication ancestor of maize, leaving maize2 copies free to potentially adopt new, or less essential functions. This prediction is fully testable on a gene-by-gene basis through investigation of the function of orthologous genes we identify in the closely related and unduplicated species sorghum. Conclusion {#s3d} ---------- This pilot study demonstrates the usefulness of traditional genetics data in the genomics era, and the importance of model species like maize with long histories of genetic investigation. A large number of morphological mutants in maize remain uncloned. The ability to identify high confidence orthologs in all grass species with sequenced genomes combined with the unrivaled economic and ecological significance of the Poaceae means investigation of a gene or gene family in any one of these species can quickly benefit researchers working around the world to answer a wide range of questions in different grass species. We hope that the tools, datasets, and links provided here ([Supplemental Information S1](#pone.0017855.s003){ref-type="supplementary-material"}), as well as our preliminary findings, will support continued insights based on pan-grass comparative genetics. Materials and Methods {#s4} ===================== Classical maize genes were identified from the list of maize loci maintained by MaizeGDB [@pone.0017855-Lawrence1], [@pone.0017855-Lawrence2] and include genes with associated GenBank sequence records with greater than three referencing papers in the database, additional cloned genes with known mutant phenotypes, as well as genes added after soliciting community input. Genes were initially mapped to the sequenced maize genome using LASTZ, and then visually proofed and corrected using GEvo part of the CoGe comparative genomics platform (<http://genomevolution.org/CoGe/>) [@pone.0017855-Lyons1]. These GEvo links are provided to aid continued research and permit proofing and verification of our results. The full length cDNA-supported gene set was constructed using the \'semi-strict assembly\' collection of full length cDNAs provided by the maize cDNA project (<http://www.maizecdna.org>) [@pone.0017855-Alexandrov1]. Full-length cDNAs were aligned to B73\_refgen2 gene models using LASTZ, and those models supported by a full length cDNA with \>95% identity and \>90% coverage were included in the set. Homeologous genes in maizes and orthologous genes in other grasses were identified using SynMap [@pone.0017855-Lyons2] with the optional Quota Align filters; SynMap is a web based tool available at <http://www.genomevolution.org/CoGe/SynMap.pl>. When no syntenic gene was identified, a predicted location was generated based on syntenically conserved flanker genes. Predicted orthologous locations longer than 1 MB were excluded as were predicted homeologous locations in maize longer than 2 MB. Our classical maize gene list provides a GEvo link that permits quick visual comparisons among grass orthologs and the predicted locations of deleted grass genes. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### **Absence of a gene homologous to c1/pl1 in the predicted orthologous location of brachypodium.** GEvo Graphic (see legend of [Figure 2](#pone-0017855-g002){ref-type="fig"}) showing the conservation of similar genes in the same positions up and downstream of the homeologous maize genes *colored alurone1* and *purple plant1.* The same flanking genes are found in the same positions relative to the single orthologous genes in the sorghum and rice genomes. The location of these same genes has been used to predict the location where an orthologous genes in brachypodium should be located, however no sequence -- annotated as a gene or otherwise -- homologous to c1/pl1 is present at the predicted location. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### **The a maximum likelihood tree showing the phylogenetic relationships of** ***colored alurone1/purple plant1*** **-like genes in maize, sorghum, rice, and brachypodium.** Based on syntenic location, these genes are predicted to fall into three clades of orthologous genes marked in yellow, green, and purple. The two genes most similar to c1/pl1 in brachypodium both fall into separate gene clades based on both tree topology and syntenic location. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Supplemental Information S1 ::: {.caption} ###### (XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: We thank Lisa Harper and Eric Lyons for involvement in the original discussion of the concept of a classical maize gene list and MaizeGDB for maintaining the information on maize genetic loci which made this research possible. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**Funded by NSF DBI-0701871 to MF and Chang-Lin Tien Graduate Fellowship to JCS. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: JCS. Performed the experiments: JCS. Analyzed the data: JCS. Wrote the paper: JCS MF.

PubMed Central

2024-06-05T04:04:19.879823

2011-3-10

PMC3053396

Introduction {#s1} ============ The recent identification of the Dermokine (*Dmkn*) gene came from different studies carried out to identify new genes specifically expressed during the late stage of epidermis differentiation [@pone.0017816-Matsui1], [@pone.0017816-Moffatt1], [@pone.0017816-Toulza1], [@pone.0017816-Bazzi1], [@pone.0017816-Naso1]. Mapped to human chromosome 19q13.1, *Dmkn* spans 25 exons. Its expression leads to four groups of transcripts according to three different transcriptional start sites, two transcriptional termination sites, and several alternative coding exons [@pone.0017816-Toulza1]. The corresponding isoforms were named α, β, γ and δ. The δ transcripts, spanning from exon 6 to exon 25, are radically different from the α, β and γ transcripts [@pone.0017816-Toulza1]. First, they show a very broad pattern of expression, including numerous tissues and organs [@pone.0017816-Toulza1], [@pone.0017816-Brandenberger1], [@pone.0017816-Kimura1], [@pone.0017816-Wakamatsu1], whereas α, β and γ mRNAs expression is mainly restricted to epidermis. Second, unlike α-, β- and γ-groups, δ mRNAs do not encode a putative signal peptide and are predicted to produce cytosolic proteins. This was confirmed by the expression of recombinant Dmknδ in transfected 293/EBNA cells [@pone.0017816-Toulza1]. Finally, the δ family of transcripts is represented by a surprisingly broad number of members. We cloned up to 9 different cDNAs from human epidermis, potentially encoding 6 different Dmknδ proteins [@pone.0017816-Toulza1]. Rab proteins make up the largest subfamily of small GTPases that play central roles in intracellular membrane trafficking. So far, in humans, the Rab family has been shown to have more than 60 proteins scattered around distinct intracellular compartments, where they regulate vesicle budding, transport and fusion [@pone.0017816-Zerial1], [@pone.0017816-Stenmark1]. Rab proteins cycle between an active (GTP-bound) and an inactive (GDP-bound) state. The nucleotide switch leads to a Rab conformational change which determines the interaction with specific regulators and effectors that are located both on membranes and in the cytosol [@pone.0017816-Pfeffer1]. For example, the GDP/GTP exchange factors (GEFs) catalyze the conversion from the GDP- to GTP-bound state, whereas GTPase-activating proteins (GAPs) catalyze GTP hydrolysis [@pone.0017816-Schwartz1]. Among the Rab family of proteins, Rab5 is a key player in the early endocytic pathway. It regulates clathrin-coated vesicle-mediated transport from the plasma membrane to the early endosomes as well as homotypic early endosome fusion. Moreover, it has also been implicated in endosome motility along microtubules [@pone.0017816-Nielsen1] and actin filaments [@pone.0017816-Lanzetti1] and also in growth factor signalling [@pone.0017816-Chen1]. The three Rab5 paralogues Rab5a, b and c [@pone.0017816-Bucci1], encode isoforms showing distinct tissue distributions [@pone.0017816-Gurkan1]. At least 20 cytosolic proteins specifically interact with active Rab5, highlighting the complexity of the downstream regulation by this GTPase [@pone.0017816-Christoforidis1]. The Dmknδ share no sequence similarity with any known protein. In order to elucidate its role we thus performed yeast two-hybrid screening and identified the Rab5 proteins as partners. By GST pull-down experiments and confocal microscopy analysis of transiently transfected HeLa cells, we further characterized the involvement of Dmknδ in the early endosomal trafficking. Materials and Methods {#s2} ===================== Yeast two-hybrid screening {#s2a} -------------------------- The yeast reporter strain AH109 was sequentially transformed with pGBKT7-Dmknδ5 and a cDNA library (Matchmaker human keratinocyte library in pGAD10, Clontech) following the instructions of the Matchmaker Gal4 two-hybrid system (Clontech). The double transformants were plated on selective medium lacking tryptophan, leucine and histidine and grown at 30°C for 5 days. Positive colonies were then picked, plated on selective medium lacking tryptophan, leucine, histidine and adenine, and tested for β-galactosidase activity using a replica plate assay. About 2.5 million library clones were screened. Library plasmids from positive colonies were isolated using Fast Prep (Thermo Scientific), rescued into E. coli DH5α and sequenced. Antibodies {#s2b} ---------- Primary antibodies were: polyclonal anti-Rab5b and anti-Rab7 (Santa Cruz Biotechnology), monoclonal anti-Rab11, anti-LAMP1 and anti-EEA1 (BD Biosciences), monoclonal anti-clathrin (Abcam), monoclonal anti-GST (Cell Signaling Technology) and anti-GFP (Novus Biologicals). Alexa-Fluor- 555 secondary antibodies were obtained from Invitrogen. Cell culture and transfection {#s2c} ----------------------------- HeLa cells were cultured in Dulbecco\'s Modified Eagle\'s Medium (DMEM) plus GlutaMAX^TM^ supplemented with 10% heat-inactivated foetal bovine serum, 50 U/ml penicillin and 50 µg/ml streptomycin (Invitrogen) at 37°C in 5% CO~2~. HeLa cells were transfected with plasmid constructs using JetPEI reagent (Polyplus Transfection), according to the manufacturer\'s instructions. Plasmid constructs {#s2d} ------------------ All cDNA clones used in this study were obtained by polymerase chain reaction (PCR) with specific primers. The DNA sequence of the insert as well as the flanking regions in each cDNA clone was verified by sequencing. ### Yeast two-hybrid constructs {#s2d1} Dmknδ5 cDNA was generated by PCR with the previously made pCEP4 construct as template [@pone.0017816-Toulza1], and cloned into the pCR2.1TOPO vector (Invitrogen). cDNAs of Dmknδ5, Dmknδ5-Nt (corresponding to exons 13 to 19 of Dmknδ5), and Dmknδ5-Ct (corresponding to exons 20 to 23 of Dmknδ5), were subcloned into the pGBKT7 vector (Clontech). Rab5a cDNA was PCR amplified using the pCMV-SPORT6-Rab5a (purchased from RZPD) as template and cloned into the pGADT7 plasmid (Clontech). ### Constructs for *in vitro* binding assays {#s2d2} Dmknδ5, Dmknδ5-Nt and Dmknδ5-Ct cDNAs were cloned into the pGEX-6P-1 expression vector (Amersham Biosciences). Wild-type Rab5b (wt) cDNA was amplified by PCR from the pGAD10 construct rescued from yeast two-hybrid screening, and cloned into pCR2.1TOPO. The previously described mutants Rab5S34N and Rab5Q79L [@pone.0017816-Stenmark2] were generated by site-directed PCR mutagenesis using Rab5bwt cDNA as template and specific primers harbouring the mutation concerned. Each Rab5b form was subcloned into the pGEX-6P-1 plasmid. The previously described "Rab5 binding domain" (R5BD) comprising the last 73 amino acids of rabaptin-5 [@pone.0017816-Stenmark3] was obtained by RT-PCR from total HeLa cells mRNA and cloned into the pGEX-6P-1 plasmid. ### Constructs for the localization studies {#s2d3} Dmknδ5 and Rab5b constructs were cloned into the pEGFP-C1 and pDsRed1-C1 vectors (Clontech) respectively. Recombinant proteins {#s2e} -------------------- The pGEX-6P-1 vectors were transformed into E. coli BL21-CodonPlus competent cells (Stratagene) and protein expression was induced with 1 mM isopropyl thio-β-D-galactoside (IPTG) for 2 hours at 37°C. Recombinant GST proteins were then extracted from bacteria cells and purified on a Glutathione Sepharose 4 Fast Flow column (Amersham Biosciences) according to the manufacturer\'s instructions. GST-Rab5bwt recombinant protein was further treated with PreScission Protease (GE Healthcare) to remove the GST moiety following the manufacturer\'s recommendations. Immunoblotting {#s2f} -------------- Proteins separated by SDS-PAGE were transferred to a Hybond-C extra membrane (GE Healthcare) and probed overnight at 4°C with primary antibodies. Bound antibodies were detected with horseradish peroxidase-conjugated secondary antibodies and developed using the Lumi-Light kit (Roche Applied Science). To determine relative protein amounts, three representative exposures for each sample were quantitated by densitometry analysis using the ImageJ free software. HeLa cell protein extract {#s2g} ------------------------- HeLa cells were harvested 36 hours after transfection in lysis buffer (25 mM Hepes-NaOH pH 7.4, 100 mM NaCl, 5 mM MgCl~2~, 1% NP40, 10% glycerol, 1 mM DTT, protease inhibitors). Extracts were incubated for 5 minutes on ice and clarified by centrifugation (10,000×*g*, 1 minute, 4°C). The supernatants were recovered and used for further pull-down assays. GST pull-down assays {#s2h} -------------------- 20 µg of glutathione sepharose (GS) beads were coated with 30 µg of GST-Dmknδ5 or GST alone for 1 hour at 4°C. After washing and equilibration, Hela cell protein extract or 10 µg of Rab5b recombinant protein were incubated for 1 hour at 4°C with coated beads. Interacting complexes were eluted with 10 mM Gluthation pH 8 and subjected to immunoblotting. In some cases, cleaved Rab5bwt recombinant protein was preloaded with 500 µM of GppNHp, a non-hydrolysable analogue of GTP or 500 µM of GDP (Jena Biosciences), overnight at 4°C, in the presence of 10 mM EDTA and 0.3% β-mercaptoethanol. The nucleotide binding reaction was stopped by adding 10 mM MgCl~2~. GppNHp-bound Rab5bwt and GDP-bound Rab5bwt were used for further pull-down assays. GTP-loaded Rab5 pull-down assay {#s2i} ------------------------------- The GST-R5BD pull-down assay was performed as previously described [@pone.0017816-Torres1]. Briefly, 80 µg of GS beads were coated with 100 µg of GST-R5BD. Beads were then incubated with fresh transfected HeLa cell protein extract for 1 hour at 4°C. Eluted interacting complexes were subjected to immunoblotting. Immunofluorescence {#s2j} ------------------ Hela cells were grown on glass coverslips for 24 hours and subjected or not to transient transfection. After 36 hours, the cells were fixed with methanol at −20°C for 2 minutes. For indirect immunofluorescence experiments, cells were immunostained with primary antibodies for 1 hour at 37°C. The respective AlexaFluor conjugated secondary antibodies were then incubated for 1 hour at room temperature. After extensive washing with PBS, the coverslips were mounted in Mowiol (Sigma-Aldrich) on glass slides and imaged on a Carl Zeiss confocal microscope LSM710. Final images were analysed using the Zen software (Carl Zeiss). Transferrin internalization assay {#s2k} --------------------------------- To deplete endogenous transferrin, HeLa cells transiently expressing GFP-Dmknδ5 or not were serum-starved for 2 hours at 37°C in internalization medium (IM) consisting of DMEM with 20 mM Hepes-NaOH (pH 7.4) and 2 mg/ml BSA added. Cells were then placed on ice, and incubated for 30 minutes at 4°C in IM containing 50 µg/ml Alexa Fluor-555 labelled transferrin (Invitrogen). After washing with ice-cold PBS, prewarmed IM was added to the cells to allow internalization of transferrin, followed by incubation at 37°C for the times indicated. The reaction was stopped by putting the cells back on ice and washing with ice-cold PBS. Cells were then fixed and processed for confocal microscopy analysis as described for the immunofluorescence experiments. Flow cytometry {#s2l} -------------- Internalization and recycling of transferrin were quantified by fluorescence-activated cell sorter (FACS), in HeLa cells transiently expressing GFP-Dmknδ5 or GFP alone. For these experiments, we used Alexa Fluor-647 labelled transferrin (Invitrogen) at the final concentration of 10 µg/ml in IM. *Internalization assay* was performed as described above except that, after stopping the reaction, the non-internalized transferrin was removed by washing with ice-cold 0.2 M acetic acid (pH 2.8) containing 0.5 M NaCl. Cells were then washed with ice-cold PBS and detached with ice-cold PBS containing 5 mM EDTA. After washing with ice-cold PBS cells were resuspended in FACS buffer (2% BSA in PBS). Alexa Fluor-647 labelled transferrin uptake was measured by flow cytometry and the percentage of transferrin that was internalized at each time-point was calculated by subtracting background (fluorescence of cells subjected to acid wash without allowing internalization) and then normalized by the total amount of transferrin prebound at +4°C. For *recycling experiments*, cells depleted of endogenous transferrin were incubated for 15 minutes at 37°C with prewarmed Alexa Fluor-647 labelled transferrin in IM. Cells were then placed on ice and washed with ice-cold PBS. Then, prewarmed IM was added followed by incubation at 37°C for the times indicated. Cells were then placed on ice, washed with ice-cold-PBS, and detached with trypsin. Harvested cells were washed in ice-cold-PBS and resuspended in FACS buffer. The amount of the fluorescent transferrin remained (non-released) in cells was measured by flow cytometry and expressed as the percentage of the initial intracellular transferrin amount detected in cells (100%, time 0 of recycling), in each experimental condition. Flow cytometry and data collection were performed on a FACSCalibur cell sorter (BD Biosciences). Data analysis was done using the WinMDI free software. Results {#s3} ======= The Dmknδ isoform family {#s3a} ------------------------ From the nine Dmknδ variants we previously cloned from human epidermis [@pone.0017816-Toulza1], we could deduce the sequence of 6 hypothetical proteins. All of them share a 123-amino-acid-length minimal sequence, and could be distinguished by 3 putative first methionine and additional sequences encoded by the alternative exons ([Figure 1](#pone-0017816-g001){ref-type="fig"} and [@pone.0017816-Toulza1]). Unlike the other Dmkn groups of transcripts, Dmknδ mRNAs were shown to be ubiquitously expressed [@pone.0017816-Toulza1]. The following experiments presented in this paper which were carried out in order to characterize Dmknδ function, were performed with the Dmknδ5 isoform. The Dmknδ5 protein displays the minimal sequence present in all the Dmknδ plus the amino acids encoded by the alternative exon 20. ::: {#pone-0017816-g001 .fig} 10.1371/journal.pone.0017816.g001 Figure 1 ::: {.caption} ###### Organization of human Dmknδ isoform family. Amino-acid sequences encoded by different exons are individualized by yellow or white boxes, and the corresponding exon number is indicated at the bottom. The amino-acid sequences encoded by the alternative exons 9, 11, 12 and 20 are in italic. The putative first methionine that can be at amino-acid position 1, 3 or 76, is indicated in blue. :::  ::: Identification of Rab5 as a binding partner for Dmknδ5 {#s3b} ------------------------------------------------------ The peptide sequences of the Dmknδ isoforms did not reveal any similarity with known functional domains. In order to gain an insight into Dmknδ function, we looked for potential partners using yeast two-hybrid analysis. We screened a human keratinocyte cDNA library with Dmknδ5 as a bait and obtained 5 clones growing on selective medium and positive for the β-galactosidase reporter gene assay. Four of them corresponded to full-length Rab5c and one to full-length Rab5b ([Figure 2A](#pone-0017816-g002){ref-type="fig"}). The small GTPase Rab5 having three isoforms that share 90% of sequence identity [@pone.0017816-Bucci1], we used the yeast two-hybrid system and found that the third isoform, Rab5a, was also able to interact with Dmknδ5 ([Figure 2A](#pone-0017816-g002){ref-type="fig"}). All further experiments were carried out using the Rab5b isoform. In order to confirm the interaction between Dmknδ5 and Rab5, we performed GST pull-down assays using bacterially expressed recombinant Dmknδ5. GST-Dmknδ5 was able to retain Rab5, either present in HeLa protein extract or produced as recombinant ([Figure 2B](#pone-0017816-g002){ref-type="fig"}). These data indicated that Dmknδ5 interacts with the endogenous Rab5 and confirmed that the interaction is direct. As we could not obtain Dmknδ specific antibody, we further analysed the localization of the endogenous Rab5 and the Dmknδ5 by confocal microscopy performed on HeLa cells expressing a GFP-tagged Dmknδ5. Rab5 was distributed throughout the cell body, with accumulation at the nuclear periphery ([Figure 2C](#pone-0017816-g002){ref-type="fig"}, *middle panel*) as previously described [@pone.0017816-Chavrier1]. GFP-Dmknδ5 was detected as diffuse in the cytosol as well as concentrated in puncta localized in the perinuclear region ([Figure 2C, *left panel*](#pone-0017816-g002){ref-type="fig"}) where it partially colocalized with endogenous Rab5 ([Figure 2C](#pone-0017816-g002){ref-type="fig"}, *arrowheads*). The size of these structures, 0.5 to 1 µm, is typical of endosomal vesicles [@pone.0017816-Stenmark2]. The expression of GFP alone only induced a diffuse cytosolic green labelling (data not shown), proving that the vesicle staining was related to Dmknδ5 expression. ::: {#pone-0017816-g002 .fig} 10.1371/journal.pone.0017816.g002 Figure 2 ::: {.caption} ###### Rab5 is a partner of Dmknδ5. *A.* By yeast two-hybrid screening, positive clones were identified as Rab5b and Rab5c. Rab5a, subsequently tested, was also able to grow on selective medium and was positive for the β--galactosidase filter assay. Three representative clones of each double transformant corresponding to Dmknδ5/Rab5b, -c or -a are shown. *B.* GST-Dmknδ5 fusion protein or GST alone were captured on glutathione-sepharose beads before loading HeLa protein extract (HeLa) or purified recombinant wild-type Rab5 (Rab5). Proteins initially loaded onto the column (input) or eluted from the column (output) were detected by immunoblotting with an antibody directed against Rab5. *C.* HeLa cells were transiently transfected with GFP-Dmknδ5 (green) and processed for immunofluorescence analysis using an anti-Rab5 antibody (red). Representative transfected cells are shown, where GFP-Dmknδ5 is found in punctate structures (arrowheads) partially colocalized with endogenous Rab5. :::  ::: The domain of Dmknδ5 responsible for the interaction with Rab5 resides in the N-terminus region of the protein {#s3c} -------------------------------------------------------------------------------------------------------------- To specify the domain of Dmknδ5 involved in the interaction with Rab5, we constructed cDNAs encoding the N-terminus (Nt) or the C-terminus (Ct) of the protein, encompassing amino-acid residues 1 to 76 and 77 to 137, respectively. These constructs were assayed for interaction with Rab5, first by using the yeast two-hybrid system. As shown in [figure 3A](#pone-0017816-g003){ref-type="fig"}, only the clones expressing Dmknδ5-Nt grew on selective medium and expressed an active LacZ reporter gene. Thus, Dmknδ5 interacts with the small GTPase via its first 76 amino-acid residues. We then investigated the subcellular localization of both Dmknδ5 regions by confocal microscopy analysis of HeLa cells co-transfected with GFP-Dmknδ5-Nt or -Ct and DsRed-Rab5wt ([Figure 3B](#pone-0017816-g003){ref-type="fig"}). We found that GFP-Dmknδ5-Nt was localized in large endosome*-*like structures that were also positive for DsRed-Rab5wt, whereas GFP-Dmknδ5-Ct showed a diffuse cytosolic pattern of expression and never colocalized with the DsRed-Rab5wt-positive large endosomes. These results are consistent with the yeast two-hybrid assay and suggest that the vesicular location of Dmknδ5 is associated with the interaction of its N-terminus with Rab5. ::: {#pone-0017816-g003 .fig} 10.1371/journal.pone.0017816.g003 Figure 3 ::: {.caption} ###### The N-terminal region of Dmknδ5 is responsible for the interaction with Rab5. *A.* δ5-Nt and δ5-Ct were tested for interaction with Rab5 using the yeast two-hybrid system. Expression of the reporter genes assay is shown for three representative clones of each double transformant (δ5-Nt/Rab5 and δ5-Ct/Rab5). *B.* HeLa cells were transiently transfected with DsRed-Rab5wt (red) and GFP-δ5-Nt or GFP-δ5-Ct (green) and observed by confocal microscopy. Bar, 5 µm :::  ::: Dmknδ5 appears to be involved early in the endocytic pathway {#s3d} ------------------------------------------------------------ To further characterize the nature of GFP-Dmknδ5 positive structures, we analysed its co-localization in HeLa cells with several well-characterized organelle markers of the endocytic pathway ([Figure 4](#pone-0017816-g004){ref-type="fig"}). We first checked that transient expression of GFP-Dmknδ5 had no impact on the subcellular localization of these proteins. We next found that GFP-Dmknδ5 colocalized with clathrin on vesicles ([Figure 4A](#pone-0017816-g004){ref-type="fig"}, *arrowheads*). Moreover, GFP-Dmknδ5 never colocalized with the Rab5 effector EEA1, suggesting that the GFP-Dmknδ5 vesicles positive for Rab5 are distinct from early endosomes ([Figure 4B](#pone-0017816-g004){ref-type="fig"}). We also tested the late endosomal marker Rab7, the lysosomal protein LAMP1 and the recycling endosomal Rab11 and never noted any colocalization with GFP-Dmknδ5 ([Figure 4C--E](#pone-0017816-g004){ref-type="fig"}). Dmknδ5 could thus play its role in the endocytic pathway, as early as the formation of clathrin-coated vesicles. ::: {#pone-0017816-g004 .fig} 10.1371/journal.pone.0017816.g004 Figure 4 ::: {.caption} ###### Characterization of Dmknδ5 positive vesicles. HeLa cells transiently transfected with GFP-Dmknδ5 (green) were processed for immunofluorescence analysis using antibodies directed against the endogenous organelle markers (red) clathrin (*A*), EEA1 (*B*), Rab11 (*C*), Rab7 (*D*) or LAMP1 (*E*). Cells were then visualized by confocal microscopy. Colocalization with GFP-Dmknδ5 was obvious only with clathrin as seen in the merged images (*A, arrowheads*). Bars, 5 µm :::  ::: In order to confirm these results, we performed pulse-chase experiments of Alexa-labelled transferrin in HeLa cells untransfected or transiently expressing the GFP-Dmknδ5 ([Figure 5A](#pone-0017816-g005){ref-type="fig"}). At the beginning of the chase (0 min), transferrin labelling was detected on the plasma membrane. Fluorescent transferrin was subsequently internalized and accumulated in big perinuclear puncta formed by the transient expression of GFP-Dmknδ5 in the cytoplasm ([Figure 5A](#pone-0017816-g005){ref-type="fig"}, [*4*](#pone-0017816-g004){ref-type="fig"} *min, arrowheads*). The staining of the membrane was no longer visible. After ten minutes of chase, colocalization of transferrin and GFP-Dmknδ5 strongly diminished ([Figure 5A](#pone-0017816-g005){ref-type="fig"}, *10 min, arrowheads*), and was no longer noticeable after 15 minutes of uptake. We next investigated whether the colocalization of GFP-Dmknδ5 with transferrin at early stages of endocytosis had an impact on the kinetic of transferrin uptake or recycling by fluorescence-activated cell sorter. No accelerated or delayed kinetics of internalization or recycling of transferrin was observed in HeLa cells expressing GFP-Dmknδ5 in comparison to HeLa cells expressing GFP alone ([Figure 5B, C](#pone-0017816-g005){ref-type="fig"}). Overall, these results confirm that Dmknδ5 plays its role upstream of early endosomes, probably during the clathrin-coated vesicle formation and/or transport to the sorting endosome, but does not modulate the kinetics of endocytosis or recycling of transferrin. ::: {#pone-0017816-g005 .fig} 10.1371/journal.pone.0017816.g005 Figure 5 ::: {.caption} ###### Dmknδ5 colocalizes early with endocytosed transferrin and does not influence transferrin uptake or recycling kinetics. *A*. HeLa cells transiently transfected with GFP-Dmknδ5 were incubated with AlexaFluor-555 conjugated-transferrin at 4°C (*0 min*). Transferrin uptake was then carried out for 4, 10, 15 or 30 min at 37°C, as indicated. Localization of GFP-Dmknδ5 (green) and Alexa Fluor-555 conjugated-transferrin (red) was then observed by confocal microscopy. Arrowheads show colocalization between GFP-Dmknδ5 and transferrin. Bars, 5 µm. *B, C.* Kinetics of endocytosis (*B*) and recycling (*C*) of transferrin in HeLa cells transiently expressing GFP-Dmknδ5 (black square) or GFP alone (open circle). For transferrin endocytosis, results are expressed as the percentage of internalized transferrin with respect to the prebound transferrin at +4°C (*B*). For recycling of intracellular transferrin, results are expressed as the percentage of initial (time 0, 100%) intracellular tranferrin (*C*). In *B* and *C*, the graphs are mean ± SD of three independent experiments. :::  ::: Dmknδ5 interacts *in vitro* with both inactive and active forms of Rab5, but preferentially colocalizes with inactive Rab5 *in vivo* {#s3e} ------------------------------------------------------------------------------------------------------------------------------------ Protein interaction with Rab5 is modulated according to the nucleotide status of the small GTPase. In order to determine whether Dmknδ5 interacts preferentially with the active or the inactive form of Rab5, we performed GST pull-down assays. Recombinant Rab5 was subjected to an exchange reaction to load it with either GppNHp or GDP and test its interaction with GST-Dmknδ5 immobilized on glutathione-sepharose beads. Both GppNHp and GDP-bound Rab5 were retained by GST-Dmknδ5 ([Figure 6A](#pone-0017816-g006){ref-type="fig"}). We then analysed the in vivo colocalization of Dmknδ5 with the constitutively active or inactive forms of Rab5. For this purpose, HeLa cells were co-transfected with GFP-Dmknδ5 and DsRed-Rab5Q79L, or DsRed-Rab5S34N. Expression of the constitutively active DsRed-Rab5Q79L induced the formation of giant early endosomes as previously described [@pone.0017816-Stenmark2]. These structures appeared negative for GFP-Dmknδ5 which localized to smaller vesicles ([Figure 6B](#pone-0017816-g006){ref-type="fig"}, *upper panel*). In contrast, GFP-Dmknδ5 colocalized with DsRed-Rab5S34N to a large extent ([Figure 6B](#pone-0017816-g006){ref-type="fig"}, *middle panel*). Interestingly, the forced expression of Dmknδ5 seemed to modify the morphology and the localization of the structures positive for the dominant negative Rab5 mutant. HeLa cells expressing GFP and DsRed-Rab5S34N formed typical tubulo-vesicular structures consistent with the inability of the mutant to promote membrane fusion [@pone.0017816-Stenmark2] ([Figure 6B](#pone-0017816-g006){ref-type="fig"}, *lower panel*). In contrast, GFP-Dmknδ5 and DsRed-Rab5S34N colocalized into big endosome-like vesicles (1--1.5 µm) scattered around the cytoplasm. Hence, Dmknδ5 is able to interact with both the active and the inactive forms of Rab5 in vitro, but preferentially localizes with the inactive mutant of Rab5 in vivo. ::: {#pone-0017816-g006 .fig} 10.1371/journal.pone.0017816.g006 Figure 6 ::: {.caption} ###### Dmknδ interacts with active and inactive Rab5. *A.* After capture of GST-Dmknδ5 fusion protein or GST alone on glutathione-sepharose beads, purified recombinant wild-type Rab5 (Rab5), incubated beforehand with GppNHp (active conformation) or GDP (inactive conformation), was loaded. Proteins initially loaded onto the column (input) or eluted from the column (output) were detected by immunoblotting with an antibody directed against Rab5. *B.* HeLa cells were transiently transfected with GFP-Dmknδ5 (green) and DsRed-Rab5Q79L or DsRed-Rab5S34N (red), respectively. Cells were then visualized by confocal microscopy. Bars, 5 µm :::  ::: Dmknδ5 modifies the balance between inactive/active Rab5 in HeLa cells {#s3f} ---------------------------------------------------------------------- Interestingly, we saw that, when GFP-Dmknδ5 was expressed in addition to DsRed-Rab5wt, it induced the enlargement of the DsRed-Rab5wt positive structures ([Figure 7A](#pone-0017816-g007){ref-type="fig"}). The diameter of these vesicles increased from 1--1.5 µm when DsRed-Rab5wt was coexpressed with GFP alone ([Figure 7A](#pone-0017816-g007){ref-type="fig"}, *lower panel)*, to 2--3 µm when it was coexpressed with GFP-Dmknδ5 ([Figure 7A](#pone-0017816-g007){ref-type="fig"} *upper panel*). The large to giant vesicles induced by the transient expression of GFP-Dmknδ5 together with DsRed-Rab5wt are reminiscent of the giant endosomes caused by the constitutively active form Rab5Q79L [@pone.0017816-Stenmark2]. This suggests that Dmknδ5 has an impact on the switch between the inactive (GDP-bound) and the active (GTP-bound) state of Rab5. To clarify this issue, we performed a GST pull-down assay based on the ability of the Rab5 binding domain (R5BD) of Rabaptin5, a Rab5 effector, to specifically link GTP-bound Rab5 [@pone.0017816-Stenmark3], [@pone.0017816-Zhu1]. We produced GST-R5BD recombinant protein and used it to pull down Rab5-GTP in HeLa protein extract. HeLa cells were transfected with DsRed-Rab5wt and either GFP, GFP-Dmknδ5, GFP-Dmknδ5-Nt or GFP-Dmknδ5-Ct. We checked that all these GFP-tagged proteins were efficiently expressed in transfected HeLa cells, as shown in [Figure 7B](#pone-0017816-g007){ref-type="fig"} (*bottom panel*). Using an anti-Rab5 antibody, we confirmed the presence of the DsRed-Rab5wt in HeLa protein extracts ([Figure 7B](#pone-0017816-g007){ref-type="fig"}, *middle panel*). We also detected, with the same antibody, the DsRed-Rab5-GTP retained by the R5BD-GST beads ([Figure 7B](#pone-0017816-g007){ref-type="fig"}, *upper panel*). After quantification by densitometry, we found significantly increased levels of active DsRed-Rab5wt among cells expressing GFP-Dmknδ5 (∼2.5 fold) and, to a lesser extent, in cells expressing GFP-Dmknδ5-Nt (∼1.5 fold) ([Figure 7C](#pone-0017816-g007){ref-type="fig"}). This is consistent with our previous observation that GFP-Dmknδ5-Nt co-expressed with DsRed-Rab5wt did not induce the formation of giant vesicles (see [Figure 3B](#pone-0017816-g003){ref-type="fig"}). In contrast, the C-terminal domain of Dmknδ5 had no effect on DsRed-Rab5wt GTP level ([Figure 7B, C](#pone-0017816-g007){ref-type="fig"}). We can thus conclude that Dmknδ5 is able to modulate Rab5 activity by promoting its GTP loading. Moreover, the N-terminal region of Dmknδ5 does not seem to be fully functional, although it is able to interact with Rab5. ::: {#pone-0017816-g007 .fig} 10.1371/journal.pone.0017816.g007 Figure 7 ::: {.caption} ###### Dmknδ expression in HeLa cells modifies the balance Rab5-GTP/Rab5-GDP. *A*. HeLa cells transiently transfected with DsRed-Rab5wt (red) and GFP-Dmknδ5 or GFP alone (green), were observed by confocal microscopy. Bars, 5 µm *B.* DsRed-Rab5wt was co-expressed in HeLa cells with GFP, GFP-Dmknδ5 (GFP-δ5), GFP-Dmknδ5-Nt (GFP-δ5Nt), or GFP-Dmknδ5-Ct (GFP-δ5Ct) as indicated. *Top,* detection of DsRed-Rab5wt-GTP amount retained by R5BD-GST with the anti-Rab5 antibody. *Middle*, total amount of DsRed-Rab5wt protein present in each HeLa protein extract used for the pull-down experiment as determined by immunoblot with the anti-Rab5 antibody. *Bottom*, expression analysis of the different GFP-tagged constructs immunoblotted with the anti-GFP antibody. *C.* Quantification of active DsRed-Rab5wt-GTP. The ratio of DsRed-Rab5wt-GTP over total DsRed-Rab5wt was determined for each condition. Western blots from two independent experiments were analysed by densitometry. Values are mean ± s.e.m. :::  ::: Discussion {#s4} ========== In this work, we describe for the first time the function of the ubiquitous Dmkn isoform, Dmknδ. By yeast two-hybrid and GST pull-down assays, we identified the small GTPase Rab5 as partner for Dmknδ5. We observed that Dmknδ5 expression in HeLa cells modified some vesicle features. Dmknδ5 is sufficient to relocalize the Rab5S34N dominant negative mutant to large vesicular structures, scattered around the cytosol. The expression of Dmknδ5 could provoke endosome fusion in spite of the inhibitory effect of Rab5S34N. Such a rescue from Rab5S34N-mediated inhibition of endosome fusion has been described in the case of BHK cells transiently expressing the Rab5GEF Rabex5 [@pone.0017816-Zhu1]. Another feature of Dmknδ5 is its ability to enlarge the vesicles induced by Rab5wt expression in HeLa cells, from 1--1.5 µm to more than 2 µm. This vesicle size is reminiscent of that observed when the Rab5-GTP mutant is expressed in HeLa cells [@pone.0017816-Stenmark2]. These data suggest that Dmknδ5 is able to promote early endosome fusion in vivo. Finally, we showed that Dmknδ5 activated Rab5 in vivo, by promoting GTP loading onto the small GTPase. This result is consistent with the observed Dmknδ5 impact on the morphology of Rab5 positive vesicles, and with Dmknδ5 preferential targeting of the GDP-bound Rab5 in vivo. Altogether, we found that Dmknδ5 activated Rab5 function and thus was involved in early endosome dynamics. The vesicles induced by Dmknδ5 expression in HeLa cells colocalized only partially with Rab5. We thus investigated other known organelle markers in order to identify these Dmknδ5 structures. We found colocalization only with clathrin, which is present on plasma membrane and clathrin-coated vesicles. The Dmknδ5-positive vesicles never contained the early endosomal marker EEA1. These results were supported by the analysis of fluorescently labelled transferrin endocytosis. Transferrin was present in the same structures as Dmknδ5, from 0 to 10 minutes of chase, which corresponds to the transport of transferrin from the plasma membrane to the early endosome. Colocalization was reduced after 10 minutes of uptake, a time reported to match the association of Rab5 with EEA1 positive endosomes [@pone.0017816-Yoon1]. Finally, colocalization was no longer visible 15 minutes after the beginning of the chase, a time corresponding to transferrin transfer into structures positive for Rab5 and Rab11 [@pone.0017816-Sonnichsen1]. Therefore, Dmknδ5 positive vesicles seem to transport transferrin from the plasma membrane to the sorting endosomes. However, we did not find any difference in the kinetics of the transferrin uptake or recycling between HeLa cells expressing GFP-Dmknδ5 or not. We also investigated the specificity of Dmknδ5 binding towards Rab5 nucleotide state. In vitro, we found that it could bind to either the active GTP-bound or the inactive GDP-bound form of Rab5. Similar biochemical properties have been reported in the case of ALS2, Varp or the RIN family [@pone.0017816-Kajiho1], [@pone.0017816-Otomo1], [@pone.0017816-Saito1], [@pone.0017816-Tall1], [@pone.0017816-Zhang1], all these proteins being identified as Rab5GEFs. We identified the three Rab5 isoforms as partners for the Dmknδ5. Currently, few data comparing these three proteins are available. In a large scale analysis of Rab protein distribution, Gurkan et al. found that, despite their high homology, Rab5a, b and c exhibited distinct tissue distribution [@pone.0017816-Gurkan1]. Moreover, although the three isoforms play their major role in the early endocytic pathway [@pone.0017816-Bucci1], they could be specifically regulated [@pone.0017816-Chiariello1], [@pone.0017816-Callaghan1] and differently involved in some processes [@pone.0017816-Wainszelbaum1], [@pone.0017816-Barbieri1]. Their specificity of function may also reside in particular affinity with some partners, like GEF. It has recently been shown that Rin1 and Gapex-5, two Rab5 GEFs, could bind to either a single Rab or indifferently the three isoforms, respectively [@pone.0017816-Chen1]. Dmknδ5, like Gapex-5, showed no significant specificity towards isoforms. Overall, these in vivo and in vitro experimental results strongly suggest that Dmknδ5 acts like a GEF for Rab5. However, its amino-acid sequence lacks the well characterized VPS9 domain which is common to all Rab5 GEFs and required for the nucleotide exchange reaction [@pone.0017816-Delprato1]. Such properties have recently been described for another Rab5 partner, the caveolin-1 [@pone.0017816-Hagiwara1]. The authors hypothesized that caveolin-1 may recruit Rab5-GEF or promote its function onto Rab5, by direct binding. Dmknδ5 might have the same properties. Our yeast two-hybrid screening did not allow us to identify a molecular partner for Dmknδ5 other than Rab5. However, this hypothesis is not excluded. Dmknδ5 binds to Rab5 via its N-terminal domain but coexpression of GFP-Dmknδ5-Nt and DsRed-Rab5wt in HeLa cells did not lead to formation of giant structures such as we observed when the full length GFP-Dmknδ5 was expressed instead of the truncated protein. Consistent results were found with the R5BD pull-down assay, showing that Rab5 activation by Dmknδ5 is reduced when only the N-terminal domain of the protein is expressed. Thus, the N-terminal part of Dmknδ5 is necessary but not sufficient for full function on Rab5 activation. Consequently, Dmknδ5 may interact, via its C-terminal domain, with another partner accounting for its modulation of Rab5 activation. Such an interaction may also be consistent with the vesicular location of Dmknδ5, which is not always correlated with the Rab5 one. In conclusion, we found that Dmknδ5 is a new actor of the early endocytic pathway. Its direct interaction with Rab5 leads to activation of the small GTPase. We also showed that Dmknδ5 is involved in endocytosis of transferrin. Further studies will help to determine which molecular mechanisms and other potential partners are involved in this process. We thank M. Ribouchon for her invaluable technical assistance and C. Serrurier for her contribution to yeast two-hybrid experiments. We thank the staff of INSERM-IFR150 (Toulouse) technical platforms, and more particularly H. Brun, C. Offer and L. Buisson from the sequencing platform, S. Allart from the confocal microscopy facility, and V. Duplan-Eche and F. L\'Faqihi-Olive from the flow cytometry facility. We wish to acknowledge M. Simon and C. Leprince for critically reviewing the manuscript. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This work was supported by grants from the \"Centre National de la Recherche Scientifique\" (CNRS) and Toulouse III University. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: EAL NJ MG. Performed the experiments: EAL LG NJ. Analyzed the data: EAL LG NJ MG GS. Wrote the paper: EAL LG NJ. Critically revised the manuscript: GS.

PubMed Central

2024-06-05T04:04:19.882763

2011-3-10

PMC3053397

Introduction {#s1} ============ Transition metals, including Cu, Mn and Zn, have essential functions in plant growth and development [@pone.0017814-White1]. However, when present at high concentrations, these metals, along with non-essential metals including Cd and Pb, become phytotoxic and must be prevented from interfering with cellular processes through compartmentalisation and exclusion [@pone.0017814-White1]-[@pone.0017814-Haydon1]. Numerous transmembrane proteins catalyse metal efflux from plant cells. These include P~1B~-ATPases, of which one group transports Cu/Ag and another transports Zn/Cd/Co/Pb [@pone.0017814-Mills1]. The most widely studied P~1B~-ATPase *in planta* is the plasma membrane protein HMA4 [@pone.0017814-White2], which has been shown to transport Zn and Cd in yeast [@pone.0017814-Mills1], [@pone.0017814-Mills2] as well as confer Zn, Cd and Co tolerance in *Arabidopsis thaliana* [@pone.0017814-Mills1], [@pone.0017814-Verret1]. HMA4 is thought to be involved in Zn homeostasis and Cd detoxification, *via* metal translocation from the root to the shoot [@pone.0017814-Mills1], [@pone.0017814-Mills2]--[@pone.0017814-Wong1]. At a subcellular level, the expression of *HMA4* has been shown to localise in the plasma membranes of *Arabidopsis thaliana* mesophyll protoplasts [@pone.0017814-Verret1]. At the tissue level, it has been localised to the pericycle cell layer of the root vasculature [@pone.0017814-Sinclair1]. In *hma4* knockout mutants, increased pericycle Zn accumulation, decreased Zn transport to the xylem parenchyma, and reduced shoot Zn accumulation have been observed [@pone.0017814-Sinclair1]. In *A. thaliana* shoots, *HMA4* expression has been localised in the phloem tissue, at the base of developing siliques, and in developing anthers, especially tapetum cells, to supply Zn to male reproductive tissue [@pone.0017814-Hussain1]. A small number of plant species have evolved that can tolerate and accumulate high concentrations of some metals in their aerial tissues under natural conditions, including Zn and Cd [@pone.0017814-Broadley1], [@pone.0017814-Krmer2]. It is thought that 10-20 species of angiosperms are Zn hyperaccumulators (\>∼0.3% Zn DW), with two of these also able to accumulate Cd to similarly high levels. In the Brassicaceae, the accumulation of high levels of Zn in shoot tissues occurs within *Noccaea* and its sister clade *Raparia* [@pone.0017814-Broadley1], [@pone.0017814-Macnair1], [@pone.0017814-Taylor1], but not in *Thlaspiceras* which contains Zn hypertolerant species (e.g. *Thlaspiceras oxyceras* (Boiss.) F.K. Mey; [@pone.0017814-Peer1]), and not in the non-Zn-hypertolerant *Microthlaspi* and *Neurotropis* clades, which are more distantly related. Within *Noccaea*, Cd hyperaccumulation occurs in a subset of *N. caerulescens* populations. *Arabidopsis halleri* is the only known Brassicaceae Zn/Cd hyperaccumulator occurring outside of the *Noccaea* genus [@pone.0017814-Broadley1], [@pone.0017814-Krmer2]. Thus, Zn/Cd hyperaccumulation may have arisen through two evolutionary events within the Brassicaceae. In *Arabidopsis halleri*, QTL involved with Zn and Cd tolerance co-localize with *HMA4* [@pone.0017814-Courbot1]. High expression of *HMA4* in the first back-cross (BC~1~) between *A. halleri,* and the non-hyperaccumulator, *A. lyrata* ssp. *petraea,* co-segregated with the *A. halleri HMA4* allele and with Cd tolerance [@pone.0017814-Courbot1]. Using RNA interference (RNAi), it was demonstrated that Zn and Cd hypertolerance were associated with *HMA4* expression in *A. halleri* [@pone.0017814-Hanikenne1]. These plants were sensitive to increased exogenous Zn and Cd treatments, translocated less Zn from the root to the shoot, and were phenotypically more similar to *A. thaliana* [@pone.0017814-Hanikenne1]. Conversely, expression of *AhHMA4* cDNA under its endogenous promoter in *A. thaliana* resulted in increased Zn concentrations in xylem parenchyma cells, resembling Zn distribution in *A. halleri* roots [@pone.0017814-Hanikenne1]. Subsequent sequencing and functional analyses of *AhHMA4* revealed that enhanced *HMA4* expression was the result of both tandem gene triplication and altered *cis* regulation [@pone.0017814-Hanikenne1]. For *N. caerulescens*, expression of a *NcHMA4* cDNA in yeast (*Saccharomyces cerevisae*) associated with enhanced Zn tolerance and increased Zn transport out of cells which supported a role for Zn efflux across plasma membranes *in planta* [@pone.0017814-Papoyan1]. In general, P~1B~-type ATPases are more highly expressed in the shoots of *N. caerulescens* than non-hyperaccumulating *Thlaspi arvense* [@pone.0017814-Hammond1] and *Arabidopsis thaliana* [@pone.0017814-Bernard1], [@pone.0017814-vandeMortel1]. Further studies characterising *N. caerulescens HMA4* transcripts found increased expression as exogenous Zn was applied at levels which were either deficient or toxic to non-hyperaccumulating species [@pone.0017814-Papoyan1], [@pone.0017814-Hammond1]. Despite circumstantial evidence for similar roles in Zn hyperaccumulation, genomic sequence data has not been published for *HMA4* in *Noccaea caerulescens.* The aim of this study was to test the hypothesis that tandem duplication and deregulation of *HMA4* expression, which occurs in *A. halleri* [@pone.0017814-Hanikenne1], also occurs in *N. caerulescens*. Results and Discussion {#s2} ====================== To test for tandem duplications of the *HMA4* locus in *N. caerulescens* required *de novo* sequence. To achieve this goal, the creation of a single copy genomic fosmid library coupled with high-throughput pyrosequencing were selected as appropriate strategies. Fosmid libraries yield large insert sizes, have high stability and reduced susceptibility to aberrent recombination, thereby ensuring maximum genomic sequence representation [@pone.0017814-Kim1], [@pone.0017814-Wang1]. By randomly shearing DNA fragments, these libraries also retain a wider selection of sequences than those based on traditional restriction digestion [@pone.0017814-Wild1]. Sequences were generated *via* Next Generation Genome Sequencer (NextGen GS) FLX 454 technology as it offered the greatest read length (350--450 bp) of current pyrosequencing technologies, and is routinely employed for *de novo* sequencing [@pone.0017814-Nyrn1]--[@pone.0017814-Pettersson1]. Construction and characterisation of a Noccaea caerulescens fosmid library {#s2a} -------------------------------------------------------------------------- The genomic fosmid library was constructed for the self-compatible Zn and Cd hyperaccumulator *Noccaea caerulescens* (J.&C. Presl) F.K. Mey., from a first generation accession from a geographically isolated population in Saint Laurent Le Minier, southern France (supplied by Guy Delmot, Saint Laurent le Minier, France, 43°55′48″ N, 3°40′12″ E) [@pone.0017814-Lochlainn1]. Such populations are self-compatible and highly inbred [@pone.0017814-Riley1]--[@pone.0017814-JimnezAmbriz1], and demonstrate low levels of heterozygosity and high inbreeding coefficients [@pone.0017814-JimnezAmbriz1]--[@pone.0017814-Dubois1]. The creation of a laboratory inbred line was not pursued, since this could result in an accumulation of mutations [@pone.0017814-Higgins1] leading to increased genetic load [@pone.0017814-Roze1] and reduced fitness, as well as gene copy number variation [@pone.0017814-Bikard1]--[@pone.0017814-SwansonWagner1] and perturbed sequencing results. To further prevent potential allelic perturbations in sequencing results, the library was constructed using leaf genomic DNA from a single plant (250 Mb), and cloned into 36,864 *Escherichia coli* EPI300^TM^-T1^R^ host cells containing highly stable, randomly sheered, ∼40 kb genomic inserts, representing ∼5.9 fold genomic coverage, while 454 sequencing reads returned \>20 fold coverage. Such sequencing strategies compare favourably with those adopted by [@pone.0017814-Hanikenne1] to robustly identify tandem triplication of *HMA4* in the self-incompatible *Arabidopsis halleri.* To elucidate the genomic sequence of *HMA4* in *N. caerulescens*, the library was probed with a radiolabelled *NcHMA4* specific sequence. Seven clones, N18P80, P6P46, N12P82, H2P47, B3P40, B22P20 and J12P81, were identified as containing *NcHMA4* sequences following PCR amplification using primers specific for the *NcHMA4* probe. Six of these fosmids demonstrated unique evidence of multiple copies of the *NcHMA4* locus following restriction digest fingerprinting ([Figure 1](#pone-0017814-g001){ref-type="fig"}). Initial pyrosequencing [@pone.0017814-Nyrn1] of a pool containing all seven fosmids returned 3 Mbp of sequence at 5-fold coverage per fosmid. Sequences were assembled into contigs and aligned to syntenic regions in the *A. thaliana* genome to confirm the presence of multiple *NcHMA4* copies. Individual copies were assigned to unique clones through PCR analyses using locus specific primers ([Figure 2](#pone-0017814-g002){ref-type="fig"}). Fosmids were then sequenced individually to improve the specificity and efficiency of prior pooled sequence assemblies, and returned 2.4 Mbp at \>20 fold coverage per fosmid. Two independent *HMA4* copies were identified in fosmids B3P40 (27,978 bp; *NcHMA4*-1 and *NcHMA4*-2) and P6P46 (31,521 bp; *NcHMA4*-3 and *NcHMA4*-4) ([Figures S1](#pone.0017814.s001){ref-type="supplementary-material"} & [S2](#pone.0017814.s002){ref-type="supplementary-material"}, [Data S1](#pone.0017814.s009){ref-type="supplementary-material"} & [S2](#pone.0017814.s010){ref-type="supplementary-material"}). Fosmid J12P81 (31,218 bp) contained *NcHMA4*-2 as well as two genes downstream to its 3′ end, whose sequences were homologous to the *A. thaliana* genes At2g19160 and At2g19170, and so demonstrated synteny with *Arabidopsis thaliana* ([Figure S3](#pone.0017814.s003){ref-type="supplementary-material"}, [Data S3](#pone.0017814.s011){ref-type="supplementary-material"}). Fosmid N18P80 (20,090 bp) contained 941 bp of the 5′ region of *NcHMA4*-3 in addition to four orthologues to At2g19060, At2g19070, At2g19080 and At2g19090, which were syntenic to this region in *A. thaliana* ([Figure S4](#pone.0017814.s004){ref-type="supplementary-material"}, [Data S4](#pone.0017814.s012){ref-type="supplementary-material"}). As indicated through locus specific PCR analysis ([Figure 2](#pone-0017814-g002){ref-type="fig"}), sequence data from fosmid H2P47 (20,258 bp) showed homology to *NcHMA4*-4 and its 5′ intergenic region, as well as the 5′ intergenic region of *NcHMA4*-1 ([Figure S5](#pone.0017814.s005){ref-type="supplementary-material"}, [Data S5](#pone.0017814.s013){ref-type="supplementary-material"}). Fosmid inserts, containing homologous sequences which demonstrated \>99% sequence identity along 5′ and 3′ ends of between 425 and 14,866 bp, were assembled into unique contiguous sequences. Consequently, fosmid H2P47 assembled both fosmids P6P46 (containing *HMA4*-3 and *HMA4*-4) and B3P40 (containing *HMA4*-1 and *HMA4*-2) into a unique locus ([Data S7](#pone.0017814.s015){ref-type="supplementary-material"},[S8](#pone.0017814.s016){ref-type="supplementary-material"}), flanked to its 5′ by N18P80, and to its 3′ by J12P81 ([Figure 3](#pone-0017814-g003){ref-type="fig"}). In support of this *HMA4* quadruplication, a genomic Southern illustrated hybridisation intensities for *Hind*III fragments, which were indicative of a 3:1 (1040--1050 bp fragment (representing *HMA4*-1, *HMA4*-3 and *HMA4*-4): 1.9 kb fragment (representing *HMA4*-2)) genomic ratio ([Figure 1](#pone-0017814-g001){ref-type="fig"}). ::: {#pone-0017814-g001 .fig} 10.1371/journal.pone.0017814.g001 Figure 1 ::: {.caption} ###### Montage of two gel blot autoradiographs of *NcHMA4* tandem repeats from *N. caerulescens* genomic DNA and genomic library fosmid insert DNA. All DNA was digested with *EcoR*I, *Hind*III or *Bam*HI corresponding to lanes **1**, **2** or **3** respectively, resolved on two 0.9% (w/v) agarose gels, blotted, and hybridized with the radiolabeled *NcHMA4* library probe (represented by darkened regions). Fosmids labelled with '+' contain tandem repeats of a *NcHMA4* insert. The DNA marker was a 1 kb DNA ladder (Hyperladder I, Bioline). Montage was prepared using CorelDraw Graphics Suite X3. :::  ::: ::: {#pone-0017814-g002 .fig} 10.1371/journal.pone.0017814.g002 Figure 2 ::: {.caption} ###### Agarose gel electrophoresis of PCR products from fosmid clones containing *N. caerulescens HMA4* sequences and *Noccaea* genomic DNA. Primers were specific for *NcHMA4*-1, *NcHMA4*-1 and 4-2, *NcHMA4*-3 and *NcHMA4*-4. Lanes were labelled according to fosmid clones or 'Genomic' *Noccaea caerulescens* DNA. The molecular ladder was a 1 kb DNA ladder (Hyperladder I, Bioline). Gel contained 1% (w/v) agarose. :::  ::: ::: {#pone-0017814-g003 .fig} 10.1371/journal.pone.0017814.g003 Figure 3 ::: {.caption} ###### Genomic organisation of *HMA4* in *N. caerulescens*. Five overlapping genomic fosmids (horizontal blue lines) represented a 101,480 bp single locus of *N. caerulescens*. Genes and their transcriptional direction are represented by arrows and given *A. thaliana* Genome Identifier appellations (brown arrows, genes syntenic with *A. thaliana HMA4* flanking regions; orange, quadruplicated *NcHMA4* genes). Restriction endonuclease site locations used for fosmid fingerprinting are indicated in red; *Hind*III, green; *Bam*HI, and teal; *EcoR*I. Numbers in brackets refer to genomic locations in base pairs. :::  ::: All five overlapping *N. caerulescens* fosmids spanned a single 101,480 bp locus in *N. caerulescens* and contained four *HMA4* tandem repeats (corresponding to At2g19110 in *A. thaliana*), compared to syntenic regions in *A. thaliana* and *A. halleri*, containing one and three copies respectively ([Figure 3](#pone-0017814-g003){ref-type="fig"}, [Data S6](#pone.0017814.s014){ref-type="supplementary-material"}). Sequences flanking *NcHMA4* tandem repeats remained essentially syntenic with *A. thaliana*. Analysis of NcHMA4 sequences {#s2b} ---------------------------- Within the deduced coding sequences, all four *NcHMA4* gene copies share between 87 and 99% nucleotide sequence identity, whilst introns demonstrated between 81 and 100% identity to consensus *NcHMA4* sequences ([Figures 4](#pone-0017814-g004){ref-type="fig"} & [S6](#pone.0017814.s006){ref-type="supplementary-material"}). The deduced coding sequences showed lower sequence identities with those of *AtHMA4* (between 76--78%) and of all three *AhHMA4* copies (between 62--66%), which may indicate that quadruplication was a relatively recent evolutionary event within *N. caerulescens* ([Figure S6](#pone.0017814.s006){ref-type="supplementary-material"}). *NcHMA4*-4 contained a truncation in exon 9 after amino acid (aa) 684 of the deduced protein sequence ([Figure 4](#pone-0017814-g004){ref-type="fig"}) and could indicate a functional but less efficient *in planta* Zn transporter, as recently reported for an *AtHMA4* which contained a comparable truncation after aa 713 [@pone.0017814-Mills3]. At the deduced amino acid level, *NcHMA4* share between 92 and 98% identity, but only between 72 and 83% identity with *AtHMA4* and between 74 and 84% identity with the three *AhHMA4* ([Figure S7](#pone.0017814.s007){ref-type="supplementary-material"}). ::: {#pone-0017814-g004 .fig} 10.1371/journal.pone.0017814.g004 Figure 4 ::: {.caption} ###### Genomic illustration of all four *NcHMA4* tandem repeats. Exons are represented by orange squares flanked by introns (blue lines). The *NcHMA4* library probe (yellow box) is illustrated at its site of hybridisation for each copy. Numbers above exons and below introns represent percentage sequence identities for each copy to a consensus *NcHMA4* sequence using Dot Matrix (Vector NTI 11). :::  ::: Within the first 2000 bp upstream of the translational start codon, *NcHMA4* sequences shared 59 and 98% identity, but between 44--49% and 41--51% identity with *A. thaliana* and *A. halleri* promoter sequences respectively ([Figure S8](#pone.0017814.s008){ref-type="supplementary-material"}). *AhHMA4* regions shared greater identity, 53--88% with *AtHMA4*, as previously reported [@pone.0017814-Hanikenne1]. Significant sequence divergence from *A. thaliana* and *A. halleri* in the 5′-flanking regions of *NcHMA4* genes indicates *cis* gene regulation may differ between species. In *A. halleri*, high *HMA4* expression was regulated in *cis* and amplified by a triplication in gene copy number [@pone.0017814-Hanikenne1]. Increased expression of *AhHMA4* correlated with enhanced Zn flux from the root symplasm into the xylem parenchyma as well as up-regulation of Zn deficiency response genes in roots supporting its role in Zn hyperaccumulation. Expression profile of NcHMA4 {#s2c} ---------------------------- To investigate the expression profile of *NcHMA4*, T~2~ *A. thaliana* plants (n = 30), transformed with *HMA4* promoters from *A. thaliana* (*AtHMA4*p::*GUS*, negative control), *A. halleri* (*AhHMA4*-3p::*GUS*, positive control) and *N. caerulescens* (*NcHMA4*-2p::*GUS*) fused to the *β-glucuronidase* (*GUS*) reporter gene, were analysed for *GUS* activity under identical nutrient replete conditions *in vitro,* 21 days after sowing (DAS). Lines bearing the *AtHMA4*p::*GUS* construct showed expression in root and stem tissue, although no staining was observed in leaf tissues ([Figure 5](#pone-0017814-g005){ref-type="fig"}). For both *NcHMA4*-2p::*GUS* and *AhHMA4*-3p::*GUS* constructs, transformed lines showed expression in most plant tissue including roots, shoots and stems ([Figure 5](#pone-0017814-g005){ref-type="fig"}). The *GUS* gene appeared to be similarly and more intensely expressed throughout plants when driven by either the *NcHMA4*-2 or the *AhHMA4*-3 promoters ([Figure 5](#pone-0017814-g005){ref-type="fig"}). ::: {#pone-0017814-g005 .fig} 10.1371/journal.pone.0017814.g005 Figure 5 ::: {.caption} ###### The spatial expression of *β-glucuronidase* (*GUS*) fused to *HMA4* promoter regions in *Arabidopsis thaliana*. The activity of *GUS* in whole leaves A--C and roots D--F from 21 day old *in vitro* cultured *A. thaliana* T~2~ transformants bearing pGWB3 constructs containing the *GUS* marker gene under the control of promoter sequences from **A, D**; *AtHMA4*, **B, E**; *NcHMA4*-2 and **C, F**; *AhHMA4*-3. Red bars represent 2 mm. :::  ::: Conclusion {#s2d} ---------- The aim of this study was to test the hypothesis that tandem duplication and deregulation of *HMA4* expression, which occurs in *A. halleri*, occurs in *N. caerulescens*. A fosmid library comprising 36,864∼40 kb inserts was developed, representing a potentially valuable resource for future map-based cloning and genome sequencing in *N. caerulescens*. Following *de novo* sequencing, there was compelling evidence of tandem quadruplication for *HMA4* in *N. caerulescens*. Whilst it is hypothetically feasible that allelic artefacts can occur, even in highly inbred populations, here, the sequencing of multiple fosmids (including long-reads of intergenic regions/non-coding repeats which are overlapping between fosmids) provides very strong support for tandem repeats of *NcHMA4*. This observation is strikingly consistent with a tandem *HMA4* triplication in the *A. halleri* genome [@pone.0017814-Hanikenne1]. *Noccaea caerulescens* and *A. halleri* last shared a common ancestor \>40 mya [@pone.0017814-Beilstein1], and the current study provides intriguing new evidence that parallel evolutionary pathways may underlie two occurrences of Zn/Cd hyperaccumulation in the Brassicaceae. Further detailed sequencing is now required in a wider number of species. An initial functional analysis was undertaken of *HMA4* using promoters from both *N. caerulescens* and *A. halleri* expressed in *A. thaliana*. Again, results were remarkably consistent between species, and the regulation of *HMA4* from both hyperaccumulator species appears to be distinct from endogenous *AtHMA4* regulation. *GUS* expression was driven more highly throughout the plant by *NcHMA4*-2p and *AhHMA4*-3p than by *AtHMA4*p, consistent with high levels of *HMA*4 transcripts in shoots and roots of *N. caerulescens* [@pone.0017814-Papoyan1]-[@pone.0017814-Bernard1] and *A. halleri* [@pone.0017814-Hanikenne1], [@pone.0017814-Talke1]. Throughout leaf tissue, promoters from both hyperaccumulator species drove enhanced *GUS* activity compared to *AtHMA4*p, which is consistent with a possible role in increasing Zn accumulation in leaf epidermal cells in *N. caerulescens* [@pone.0017814-Ma1], [@pone.0017814-Frey1] and mesophyll cells in *A. halleri* [@pone.0017814-Kpper1]. Moreover, expression of all three *AhHMA4* promoters in the xylem parenchyma and the cambium of leaves in both *A. thaliana* and *A. halleri* were hypothesised to be consistent with putative roles in Zn exclusion from particular cell types and metal distribution within the leaf blade [@pone.0017814-Hanikenne1]. In contrast to the relatively high, shared HMA4 sequence identities between all three genomes, *N. caerulescens HMA4* promoters (*NcHMA4*p) exhibited lower identities with those from both *A. thaliana* and *A. halleri*. We conclude that novel *cis* regulatory elements in *N. caerulescens* contribute to increased *NcHMA4* gene expression. Further elucidating these *cis* regulatory regions in hyperaccumulators could enable the manipulation of *HMA4* expression that may be exploited for use within crop systems to enhance Zn leaf accumulation for biofortification or phytoremediation strategies. Materials and Methods {#s3} ===================== Library Construction {#s3a} -------------------- DNA from a *Noccaea caerulescens* (J.&C. Presl) F.K.Mey (∼250 Mb genome, 2n = 2x = 14) from a population originating from Saint Laurent Le Minier, southern France [@pone.0017814-Lochlainn1] was used to construct a genomic fosmid library by Warwick Plant Genomic Libraries Ltd. (Warwick HRI, Warwick, UK). Preparation of Noccaea DNA and bacterial cells for filter arraying {#s3b} ------------------------------------------------------------------ DNA (2.5 µg), extracted from leaf tissue of a single *N. caerulescens* plant *via* the phenol -- chloroform procedure [@pone.0017814-Sambrook1], was randomly sheared to 40 kb fragments and end repaired to blunt, 5′-phosphorylated ends. Fragments were size resolved and purified from a low melting point (LMP) agarose gel (without exposure to UV irradiation), before ligating to 8 kb Cloning-Ready CopyControl pCC1FOS vectors and phage packaging (Epicentre Biotechnologies, Madison, W.I., USA). EPI300^TM^-T1^R^ plating strains were streaked on solid Luria-Bertani (LB) plates and grown for 12 h at 37°C. A starter culture (5 ml LB broth) was inoculated with a single colony and incubated on a shaker at 225 rpm. for a further 12 h at 30°C. 50 ml of LB broth +10 mM MgSO~4~+0.2% (w/v) maltose (20% filter sterilised stock) was inoculated with 1 ml of starter culture and shaken at 37°C for 2--3 h until an optical density at 600 nm (OD~600~) of 0.8--1.0 was reached. Bacteria was pelleted (500× g for 10 mins.), gently resuspended in 25 ml of 10 mM MgSO~4~, before being diluted to an OD~600~ of 0.5 with sterile 10 mM MgSO~4~. A 25 µl aliquot of this solution was mixed in a 2 ml microcentrifuge tube with 25 µl of the fosmid packaging reaction (diluted in phage dilution buffer according to library titre), and incubated for 30 mins at room temperature (RT). LB (200 µl) was added to each sample and incubated for 1 h at 37°C, shaking the tube gently once every 15 minutes. CopyControl fosmid clones were selected by pelleting samples (1 min. at 10,000 rpm.) and resuspending in fresh LB medium before spreading on LB agar plates supplemented with 12.5 µg ml^−1^ chloramphenicol and incubating at 37°C for 12 h. Colonies were then picked into 384 well plates using a Q-Pix II bench top colony picker (Genetix Ltd., New Milton, Hampshire, UK). The filters were constructed using a MicroGrid II high-throughput automated microarrayer, (BioRobotics Ltd., Cambridge, UK). Probing the N. caerulescens genomic library {#s3c} ------------------------------------------- *Noccaea caerulescens* library DNA fragments were cloned into 36,864 *E-coli* EPI300^TM^-T1^R^ host cells and stored in 96×384-well microtiter plates which were arrayed evenly onto two nitrocellulose filters (48 plates per filter). Each well contained duplicated DNA fragments, whose arrangement indicated the plate of origin for DNA that hybridised to the *HMA4* probe. Filters contained approximately 5% ribosomal and 15% chloroplast contamination. Synthesis and radiolabelling of DNA probes {#s3d} ------------------------------------------ Oligonucleotide were designed to amplify 421 bp fragment in the 3′ region of the publicly available *N. caerulescens* ecotype Prayon *HMA4* cDNA sequence, GenBank accession ID AY486001.1, (<http://www.ncbi.nlm.nih.gov/>) using forward: 5′-GCTAGGGAATGCTTTGGATG-3′, and reverse: 5′-CTTCTCTCGCAGAAGCAACA-3′, primer sequences (MWG Biotech, Ebersberg, Germany). DNA probes (50 ng) for library hybridisations were labelled with dCTP α-^32^P (0.4 MBq µl^−1^), by random priming using Ready-To-Go DNA Labelling Beads (-dCTP) kit (GE Healthcare, Buckinghamshire, UK) as described by the manufacturer. The labelled probe, dissolved in 50 µl of TE buffer, was separated from unincorporated nucleotides by passing through an Illustra Nick column (GE Healthcare, Buckinghamshire, UK) and heat denatured as described by the manufacturer. Radiolabelling and hybridisation of the HMA4 library probe {#s3e} ---------------------------------------------------------- Each pair of library filters were submerged for 4 h at 55°C in 250 ml prehybridisation solution, then incubated (55°C) overnight with the radiolabelled probe, before reducing radioactivity to 15--30 counts per minute through repeated washing in solutions of 2 X SSC +0.1% SDS [@pone.0017814-Sambrook1]. Hybridised filters were sealed in plastic and exposed to autoradigraphy film (Kodak X-Omat AR Film XAR-5, Sigma-Aldrich GmbH, Steinheim, Germany) at −80°C for 4--5 d. Positive hybridisations were localised and corresponding fosmids, grown and their plasmid extracted. Identification and 'fingerprinting' fosmids of interest {#s3f} ------------------------------------------------------- Fosmids containing genes of interest were confirmed initially by colony PCR using probe specific primers. Selected fosmids were then 'fingerprinted' through individual restriction digestion with *EcoR*I, *Hind*III and *Bam*HI (4 h at 37°C) (Promega, Southampton, UK) before running 10 µl of each digest in a 1% (w/v) agarose/TAE electrophoresis gel for 12 hrs at 0.5 V cm^−1^. Gels were blotted onto a pre-cut nylon membrane (12 h) and hybridised with the *HMA4* library probe [@pone.0017814-Sambrook1]. Genomic DNA extracted from *N. caerulescens* Saint Laurent Le Minier was used as a positive control to compare all observed hybridisation patterns. Sequencing N. caerulescens fosmid clones of interest {#s3g} ---------------------------------------------------- Pooled fosmid pyrosequencing and shotgun library preparation using a 454 Genome Sequencer FLX (454 GS-FLX) Next Generation (NextGen) platform with standard sequencing chemistry (∼250 bp read lengths; Roche Diagnostics GmbH) was carried out by Cogenics Genome Express (Cambridge, UK) while individual fosmid shotgun libraries and GS-FLX Titanium sequencing chemistry (350--450 bp read lengths, 20-fold coverage) with gap filling by Sanger dideoxy sequencing was performed by Eurofins Genetic Services Limited (82152 Martinsried, Germany). For all, sequencing and assembly of the shotgun data was performed using a standard whole genome sequencing assembly with the 454/Roche Newbler assembler V 1.1.02.15. [@pone.0017814-Meyer1], [@pone.0017814-Pettersson1]. Fosmids were extracted from bacterial suspension following the Maxiprep plasmid isolation protocol [@pone.0017814-Sambrook1]. Contig alignments of N. caerulescens fosmid sequences {#s3h} ----------------------------------------------------- Fosmid pCC1FOS™ vector sequences were isolated from *Noccaea caerulescens* inserts *in silico, via* the NCBI database Basic Local Alignment Search Tool (BLASTn) algorithm, against all available nucleotide sequences at default parameters (<http://blast.ncbi.nlm.nih.gov/Blast.cgi?CMD=Web&PAGE_TYPE=BlastHome>). Inserts were aligned to *A. thaliana* orthologous regions and assembled into one large contiguous sequence *via* AlignX and ContigExpress software using default gap settings (Vector NTI 11; Invitrogen, Paisley, UK). Overlapping insert regions were identified between fosmid end-sequences which aligned to identical *A. thaliana* regions and shared \>99% sequence identity. Consensus sequences were assigned to assemblies of repetitive regions and poly-A and poly-T stretches that showed variation between homologous fosmid sections. All protein and nucleotide sequence comparisons and percentage identities were calculated using Dot Matrix at a stringency of 30% and window of 5 (Vector NTI 11). Creating promoter::GUS fusion constructs {#s3i} ---------------------------------------- Primers specific for the promoter regions of *Arabidopsis thaliana* (L.) Heynh. Colombia (Col-0) (*AtHMA4*), *A. halleri* (L.) O\'Kane & Al-Shehbaz *ssp. halleri* (*AhHMA4*-3), and *N. caerulescens* (J.&C. Presl) F.K. Mey. Saint Laurent Le Minier (*NcHMA4*-2) were designed using Primer 3 Version 0.4.0 (<http://frodo.wi.mit.edu/primer3>). Promoter fragments were PCR amplified from plant DNA with Phusion® proofreading polymerase (Finnzymes, Finland) and ligated into the pCR8®/GW/TOPO® entry vector using the TA cloning system (Invitrogen, Paisley, UK.). Cloned promoter sequences were fused with *β-glucuronidase* (*GUS*) in pGWB3 Gateway-compatible destination vectors [@pone.0017814-Nakagawa1] *via* LR-mediated Gateway cloning technology as described by the manufacturer (Gateway® LR Clonase®; Invitrogen, Paisley, UK). Bacterial transformations {#s3j} ------------------------- Presence and orientation of promoters were confirmed in all constructs through Sanger sequencing. Plasmid extractions, antibiotic selection and transformations for chemical- (*E. coli* DH5α) and electro- (*Agrobacterium tumefaciens* GV3101 [@pone.0017814-Koncz1]) competent bacterial cells were performed as described [@pone.0017814-Sambrook1]. Analysis of GUS expression in T~2~ transgenic Arabidopsis thaliana {#s3k} ------------------------------------------------------------------ Histochemical detection of *GUS* activity [@pone.0017814-Jefferson1] was performed on T~2~ segregating transformed *Arabidopsis thaliana* Col-0 plants [@pone.0017814-Clough1], selected on agar-based medium (10 g l^−1^ sucrose, 8 g l^−1^ agar and 2.1 g l^−1^ Murashige and Skoog (MS) basal medium (M5524, Sigma-Aldrich, Poole, UK)), supplemented with 50 µg ml^−1^ kanamycin sulphate. Seven DAS healthy, green, actively growing plants were transferred under axenic conditions to translucent polycarbonate growth boxes containing 75 ml of un-supplemented agar-based media and cultured for a further 14 days (21 DAS) at 20±2°C, under 16 h photoperiod, at 50--80 µmol photons m^−2^ s^−1^ light intensity from 58 W white halophosphate fluorescent tubes (Cooper Lighting and Security, Doncaster, UK). A randomised block design comprising three replicates was employed, with three independent transformed lines for each of the three promoter constructs and one wild type (WT) line allocated at random within each replicate (n = 30). For each replicate, all samples from each line, including WT control, were placed into individual sterile glass universals (3 plants per bottle) containing 10 ml of *GUS* assay solution [@pone.0017814-Jefferson1] (5 mg of X-Gluc (5-bromo-4-chloro-3-indolyl-β-D-glucuronic acid; Melford Laboratories Ltd, Ipswich, UK) dissolved in 100 µl of dimethyl formamide (DMF), phosphate buffer (0.2 M NaH~2~PO~4~ plus 0.2 M Na~2~HPO~4~, pH 7.0), 0.5 M Na~2~EDTA, 10 mM K~3~Fe(CN)~6~, 10 mM K~4~Fe(CN)~6~.3H~2~O and 0.1% (v/v) Triton-X-100 (Sigma-Aldrich Co., Steinheim, Germany), and incubated in the dark at 37°C for 16 h. Chlorophyll was removed from each sample to aid later imaging of *GUS* staining from the histochemical treatment. Samples were suspended in acidified methanol (2 ml conc. HCl +10 ml methanol +38 ml H~2~O) for 15 min at 50°C with intermittent gentle shaking, before decanting and re-suspending in a neutralisation solution (7% NaOH in 60% ethanol) for 15 min at RT. Solutions were discarded and retained samples were rehydrated using a series of decreasing concentrations of ethanol (from 40, 20 and 10% v/v). Once fully rehydrated in milli-Q H~2~O, samples were mounted in 50% glycerol (v/v) and viewed under a stereo microscope for traces of indigo staining to indicate *GUS* activity. Primers employed {#s3l} ---------------- Sequences of primers employed to isolate *HMA4* promoter sequences from *A. thaliana*, *A. halleri* and *N. caerulescens* were from 5′ to 3′: ***NcHMA4*** **-2 promoter\_Fwd** CTCCTTCTGTAACGCCATTTCTGTA ***NcHMA4*** **-2 promoter\_Rev** CTCCTTCTGTAACGCCATTTCTGTA ***AtHMA4*** **promoter\_Fwd** ACTTACCGATCGGGTATGCCATG ***AtHMA4*** **promoter\_Rev** TTTCTCTTCTTCTTTGTTTTGTAACGCC ***AhHMA4*** **-3 promoter\_Fwd** GTGTTTGCTGGTGCTACTGTCTGA ***AhHMA4*** **-3 promoter\_Rev** TTTCTCTTCTTCTTTGTTTTGTGACGCC Supporting Information {#s4} ====================== Figure S1 ::: {.caption} ###### **Consensus of the genomic illustration of the fosmid B3P40.** Yellow bar represents the entire 27978 bp genomic insert. Green arrows illustrate both tandem repeats of *NcHMA4*-1 and *NcHMA4*-2 and their transcriptional direction. Blue script and lines highlight sites in the fosmid which were 100% specific for that primer. Image created through Vector NTI 11 (Invitrogen, Paisley, UK). (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### **Consensus of the genomic illustration of the fosmid P6P46.** Yellow bar represents the entire 31521 bp genomic insert. Green arrows illustrate both tandem repeats of *NcHMA4*-3 and *NcHMA4*-4 and their transcriptional direction. Blue script and lines highlight sites in the fosmid which were 100% specific for that primer. Image created through Vector NTI 11 (Invitrogen, Paisley, UK). (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S3 ::: {.caption} ###### **Consensus of the genomic illustration of the fosmid J12P81.** Yellow bar represents the entire 31218 bp genomic insert. Green arrow illustrates a single copy of *NcHMA4*-2 its transcriptional direction. Brown arrows illustrate flanking genes At2g19160 and At2g19170 and their transcriptional directions. Flanking genes are labelled according to their *A. thaliana* orthologues. Blue script and lines highlight sites in the fosmid which were 100% specific for that primer. Image created through Vector NTI 11 (Invitrogen, Paisley, UK). (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S4 ::: {.caption} ###### **Consensus of the genomic illustration of the fosmid N18P80.** Yellow bar represents the entire 20090 bp genomic insert. Green box illustrates a single copy of the 5′ end of *NcHMA4*-3. Brown arrows illustrate flanking genes At2g19060, At2g19070, At2g19080 and At2g19090 and their transcriptional directions. Flanking genes are labelled according to their *A. thaliana* orthologues. Blue script and lines highlight sites in the fosmid which were 100% specific for that primer. Image created through Vector NTI 11 (Invitrogen, Paisley, UK). (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S5 ::: {.caption} ###### **Consensus of the genomic illustration of the fosmid H2P47.** Yellow bar represents the entire 20258 bp genomic insert. Green arrow illustrates a single copy of *NcHMA4*-4 and its transcriptional direction. Blue script and lines highlight sites in the fosmid which were 100% specific for that primer. Image created through Vector NTI 11 (Invitrogen, Paisley, UK). (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S6 ::: {.caption} ###### ***HMA4*** **coding sequence identities.** A) Cladogram and B) Dot Matrix comparisons of coding sequences of *HMA4* orthologues from Ah: *Arabidopsis halleri*, At: *Arabidopsis thaliana* and Nc: *Noccaea caerulescens.* Tandem repeats are highlighted by "-". "Pra" and "Her" refer to publicly available sequence data from *N. caerulescens* ecotypes Prayon and Hérault. The cladogram was created for nucleotide sequences by the DNA Sequence Parsimony Method (DNApars), and was run at default settings in Phylip version 3.68. The Dot Matrix program was run at default settings and supplied by Vector NTI 11. Numbers represent percentage sequence identities. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S7 ::: {.caption} ###### **HMA4 Protein sequence identities.** A) Cladogram and B) Dot Matrix comparison of protein sequences of *HMA4* orthologues from Ah: *Arabidopsis halleri*, At: *Arabidopsis thaliana* and Nc: *Noccaea caerulescens.* Tandem repeats are highlighted by "-". "Pra" and "Her" refer to publicly available sequence data from *N. caerulescens* ecotypes Prayon and Hérault. The cladogram was created for amino acid sequences through Protpars, Protein Sequence Parsimony Method and was run at default settings and supplied by Phylip version 3.68. The Dot Matrix program was run at default settings and supplied by Vector NTI 11. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S8 ::: {.caption} ###### ***HMA4*** **promoter region sequence identities.** A) Cladogram and B) Dot Matrix comparisons of sequences 2000 bp upstream from the transcriptional start site of *HMA4* orthologues from Ah: *Arabidopsis halleri*, At: *Arabidopsis thaliana* and Nc: *Noccaea caerulescens.* Tandem repeats are differentiated by "-". The cladogram was created for nucleotide sequences by the DNA Sequence Parsimony Method (DNApars), and was run at default settings in Phylip version 3.68. The Dot Matrix program was run at default settings and supplied by Vector NTI 11. Numbers represent percentage sequence identities. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Data S1 ::: {.caption} ###### **Fosmid B3P40 insert sequence.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Data S2 ::: {.caption} ###### **Fosmid P6P46 insert sequence.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Data S3 ::: {.caption} ###### **Fosmid J12P81 insert sequence.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Data S4 ::: {.caption} ###### **Fosmid N18P80 insert sequence.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Data S5 ::: {.caption} ###### **Fosmid H2P47 insert sequence.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Data S6 ::: {.caption} ###### **Consensus sequence of the entire** ***NcHMA4*** **single genomic locus.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Data S7 ::: {.caption} ###### **Sequence alignment of overlapping regions of fosmids P6P46 and H2P47.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: Data S8 ::: {.caption} ###### **Sequence alignment of overlapping regions of fosmids H2P47 and B3P40.** (DOC) ::: ::: {.caption} ###### Click here for additional data file. ::: We gratefully acknowledge all assistance from staff at the University of Nottingham, specifically, from the Plant and Crop Sciences Division, Drs. Zsuzsanna Bodi, Silin Zhong and Katalin Kovács for their guidance with molecular genetic techniques, Karmeswaree Naiken and Dr. Ranjan Swarup for assisting with histochemical analyses, and Mike Beard and Laura Holt (Photograph Unit) for image preparation. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**Seosamh Ó Lochlainn was funded by a UK Biotechnology and Biological Sciences Research Council (BBSRC) Studentship (BBSSE200613215). Rothamsted Research is an institute of the BBSRC. The Scottish Crop Research Institute is funded by the Scottish Government Rural and Environment Research and Analysis Directorate. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: SOL HCB RGF JPH GJK PJW NSG MRB. Performed the experiments: SOL HCB. Analyzed the data: SOL. Contributed reagents/materials/analysis tools: HCB RGF JPH GJK PJW NSG MRB. Wrote the paper: SOL HCB RGF JPH GJK PJW NSG MRB.

PubMed Central

2024-06-05T04:04:19.885640

2011-3-10

PMC3053398

Introduction {#s1} ============ Cancer is a complex disease characterized by a number of genetic and epigenetic abnormalities. Patients associated with similar clinical and pathological parameters may have very different tumor profiles at the molecular level and may respond differently to treatment [@pone.0017845-Srlie1], [@pone.0017845-vandeVijver1], [@pone.0017845-vantVeer1], [@pone.0017845-Sotiriou1]. Genome-wide expression profiling of tumors has become an important tool to identify gene sets and gene signatures that can be used to predict clinical endpoints, such as survival and therapy response [@pone.0017845-Srlie1], [@pone.0017845-vandeVijver1], [@pone.0017845-vantVeer1], [@pone.0017845-Chang1], [@pone.0017845-Paik1], [@pone.0017845-Minn1], [@pone.0017845-Wang1], [@pone.0017845-Chi1], [@pone.0017845-Liu1], [@pone.0017845-Naume1], [@pone.0017845-Hu1]. A number of tumor classification algorithms based on gene expression profiles have been proposed, using clinical data or known biological class labels to build predictive models for outcome: e.g. the 70-gene signature MammaPrint® [@pone.0017845-vantVeer1], the 76-gene signature of Wang et al. [@pone.0017845-Wang1] and the Genomic Grade Index [@pone.0017845-Sotiriou1]. Several classification methods also utilize unsupervised methods to discover biological subgroups [@pone.0017845-Srlie1], [@pone.0017845-Chang1], [@pone.0017845-Chi1]. Published gene signatures that are predictive of clinical outcome in breast cancer are partly or completely based on different genes. Nevertheless, their predictions are often in good concordance in terms of assigning new patient samples into groups of poor and good outcome [@pone.0017845-Fan1], [@pone.0017845-Reyal1]. This indicates that some common biological processes overlap across those gene signatures [@pone.0017845-Reyal1], [@pone.0017845-Yu1], but the reported gene signatures are likely to capture various biological aspects of breast cancer. Hence, by combining information from multiple gene signatures, one would potentially increase the prediction power and bring out an overall picture of this disease. We therefore aim to develop an analytical framework that allows us to utilize the combined strength from individual gene signatures. Such a framework and the resulting model will be broadly applicable for survival prediction across heterogeneous tumor groups capturing a broad spectrum of biological aspects. Using the original gene signatures would generally require recalibration on new datasets. As a consequence, the original gene signatures will not always be maintained and used as originally intended. In this paper, the aim is not to compare or condition on the original gene signatures per se, but rather to focus on the genes themselves. Thus we rely only on the gene sets on which the gene signatures are based, and fit survival models based on these gene sets without depending on the quantitative specifics of the published gene signatures. In doing so, we are able to utilize what is arguably the most critical piece of information from the gene signatures, and analytically most challenging to derive: shortlists of genes related to breast cancer survival. It should be emphasized that the main aim of our study is to present a method for explorative analyses and for seeding gene selection algorithms with prior known gene sets, rather than a claimed method for producing optimized gene signatures competing with published counterparts. We use the gene sets of eleven published gene signatures to analyze breast cancer survival and relapse ([Table 1](#pone-0017845-t001){ref-type="table"}). We aim to build survival predictors on a common data set to achieve a fair comparison of the gene sets. For each gene set, we fit survival models to one dataset and apply them to another to obtain predictions of cancer relapse and death, resulting in a predictive index (PI) per patient. We then combine the PIs obtained for a patient by extracting a common signal using the first principal component, in order to improve the predictions obtained by each gene set separately. We illustrate the capacity of the proposed framework to lead to improved survival prediction. A flowchart of the analysis is shown in [Figure 1](#pone-0017845-g001){ref-type="fig"}; see specific sections under [Results](#s2){ref-type="sec"} for a detailed explanation of the different parts of the figure. ::: {#pone-0017845-g001 .fig} 10.1371/journal.pone.0017845.g001 Figure 1 ::: {.caption} ###### Flowchart of the analysis. (**A**) Construction of the gene-set predictor/gene signature for risk prediction. Input: A set of genes of interest (gene *1*, ..., *m*) which can be traced by the corresponding colors through out the diagram; gene expression data for training cohort and test cohort with genes placed in the rows and patients in the columns. *Step 1*. Gene identity mapping and extract expression matrix. *Step 2*. With available status of observing an event for the patients on the training set, a Cox model with L2 penalty is used to model the relationship of survival probability and gene expression pattern of the gene set. The coefficients or "gene weights" (*β~1~, ..., β~m~*) associated with individual genes are estimated from the Cox-ridge model. Size of the bubble in the gene weights matrix reflects the importance of the corresponding gene for survival prediction. *Step 3*. A *Prognostic Index* (PI), the predicted risk score for a test patient *i* (*i = 1, ..., n*) is calculated by the sum of weighted gene expression from test patient *i* using the estimated gene weights from step2. (**B**) Integration of multiple gene signatures by dimension reduction. Input multiple gene sets of interest together with their gene expression data. *Module 1*: For *j*th gene set (*j* = 1, *..., R*), the procedure described in panel A is used to predict a risk score PI for individual test patient. The resulting PI matrix is positioned in *R* by *n* dimension representing the risk prediction of the *n* test patients by each of the *R* gene sets. *Module 2:* Integrate predictions from multiple gene signatures by dimension reduction using principal components analysis (PCA). *Module 3:* Dichotomize the risk scores on PC1 by median (higher than median indicates high risk) resulting in two predicted risk groups for survival outcome. :::  ::: ::: {#pone-0017845-t001 .table-wrap} 10.1371/journal.pone.0017845.t001 Table 1 ::: {.caption} ###### Published gene sets included in the analysis. ::: {#pone-0017845-t001-1} Coding Full name Platform ----------- -------------------------------------------------------------------- --------------------------------- RS 16-gene-recurrence-score predictor [@pone.0017845-Paik1] Oncotype DX assay SD 26-gene stroma-derived prognostic predictor [@pone.0017845-Finak1] Agilent Human oligo microarrays LM 54-lung-metastasis-gene signature [@pone.0017845-Minn1] Affymetrix GeneChips AMST 70-gene predictor [@pone.0017845-vantVeer1] Agilent Human oligo microarrays ROT 76-gene predictor [@pone.0017845-Wang1] Affymetrix GeneChips Grade 97- histologic-grade-associated markers [@pone.0017845-Sotiriou1] Affymetrix GeneChips Robust 127-gene classifier [@pone.0017845-vanVliet1] Affymetrix GeneChips Hypoxia 168-hypoxia-gene signature [@pone.0017845-Chi1] Stanford 43k cDNA array Stem 186-invasiveness-gene signature [@pone.0017845-Liu1] Affymetrix GeneChips Intrinsic 306-intrinsic/UNC gene list [@pone.0017845-Hu1] Agilent Human oligo microarrays WR 512 wound response gene list [@pone.0017845-Chang1] Stanford 43k cDNA array ::: Results {#s2} ======= Cross-platform gene mapping {#s2a} --------------------------- Using the gene mapping procedure described in Methods, we were able to identify and map at least 80% of the genes from each of the originally published gene sets to the Stanford 43k cDNA array. [Table 2](#pone-0017845-t002){ref-type="table"} summarizes the number of genes mapped to the training set (MicMa [@pone.0017845-Naume1]) and test set (Ull [@pone.0017845-Langerd1]). The number of genes that overlap between the different gene sets is shown in [Table 3](#pone-0017845-t003){ref-type="table"}. The percentage of overlap ranges from 0 to 25% (between Robust and Grade) with a median overlap of 0.57%. ::: {#pone-0017845-t002 .table-wrap} 10.1371/journal.pone.0017845.t002 Table 2 ::: {.caption} ###### Acronyms for original gene sets and coverage on training & test set. ::: {#pone-0017845-t002-2} Gene set Mapped genes \# Coverage % ----------- ----------------- ------------------------------------------------ RS 15 15/16 = 94% SD 22 22/26 = 85% LM 48 48/54 = 89% AMST 57 57/70 = 81% ROT 66 66/76 = 87% Grade 111 111/128[§](#nt101){ref-type="table-fn"} = 87% Robust 114 114/127 = 90% Hypoxia 168 168/168 = 100% Stem 161 161/186 = 87% Intrinsic 290 290/306 = 95% WR 561 561/573[‡](#nt102){ref-type="table-fn"}  = 98% § 128 Affymetrix probe IDs were used instead of 97 gene symbols. ‡ 573 image clone IDs were used instead of 512 genes. ::: ::: {#pone-0017845-t003 .table-wrap} 10.1371/journal.pone.0017845.t003 Table 3 ::: {.caption} ###### Number of overlapping genes between gene sets. ::: {#pone-0017845-t003-3} RS SD LM AMST ROT Grade Robust Hypoxia Stem Intrinsic WR ----------- ---- ---- ---- ----------------------------------- ----- ------- -------- --------- ------ ----------- ---- RS 0 0 1 0 3 2 0 1 5 1 SD 0 0 1 0 0 1 1 0 0 1 LM 0 0 0 0 0 0 1 0 7 5 AMST 1 1 0 0 7 7 1 0 2 3 ROT 0 0 0 0[§](#nt103){ref-type="table-fn"} 7 2 1 0 4 3 Grade 3 0 0 7 7 45 0 0 11 12 Robust 2 1 0 7 2 45 2 4 11 10 Hypoxia 0 1 1 1 1 0 2 4 3 12 Stem 1 0 0 0 0 0 4 4 5 7 Intrinsic 5 0 7 2 4 11 11 3 5 19 WR 1 1 5 3 3 12 10 12 7 19 § There was 1 gene overlapping between ROT & AMST according to the previous report [@pone.0017845-Yu1]: *CCNE2* (GenBankID NM\_004702). However, in the newer version of the NCBI database: "NM\_004702.2 was permanently suppressed because the transcript has insufficient support and is a nonsense-mediated mRNA decay (NMD) candidate." (<http://www.ncbi.nlm.nih.gov/entrez/viewer.fcgi?db=nuccore&id=17318566>. Accessed on Mar. 18, 2009). ::: Penalized Cox regression {#s2b} ------------------------ For each gene set, a Cox model was used to relate time to systemic relapse to gene expression (see the second last section in [Results](#s2){ref-type="sec"} for analysis using breast cancer specific death as the clinical endpoint). A weight is thus assigned to each gene and a risk score (or *prognostic index, PI*) is found for each individual by adding together the weighted gene expressions. Due to the large number of covariates, ranging from 15 genes (RS; [@pone.0017845-Paik1]) up to 561 genes (WR; [@pone.0017845-Chang1]), we estimated a penalized partial likelihood, where the penalty is proportional to the Euclidean (L2) norm of the parameter vector. This has the effect of avoid over-fitting by reducing the variance of the estimator, at the cost of introducing a bias. A penalty parameter *λ* controls the trade-off between the goodness of fit (imposed by the partial likelihood) and low variance (imposed by the penalty term). As *λ*→0 the solution approaches that of ordinary Cox regression, while as *λ*→∞ the solution approaches that of a constant estimator that does not depend on the expression of any gene. In practice, a good choice for *λ* has to be determined empirically, and for this purpose we applied leave-one-out cross-validation (LOOCV) [@pone.0017845-Verweij1]. The cross-validation curves obtained for each of the gene sets are shown in [Figure S1](#pone.0017845.s001){ref-type="supplementary-material"}, and the *λ* that maximizes the cross-validation function can be found in [Table 4](#pone-0017845-t004){ref-type="table"}. For gene sets *SD* and *LM*, the cross-validation function did not have a maximum in the search range resulting in a null model that has no covariates. ::: {#pone-0017845-t004 .table-wrap} 10.1371/journal.pone.0017845.t004 Table 4 ::: {.caption} ###### Individual gene set prediction characteristics (optimal *λ* by LOOCV in model building, change in deviance on test set, standard deviation for PIs). ::: {#pone-0017845-t004-4} Prediction characteristics statistic ----------- -------------------------------------- ------- ------- RS 338 -1.4 0.109 SD \- 0 0 LM \- 0 0 AMST 227 0.3 0.195 ROT 1029 0.1 0.095 Grade 3580 -2.12 0.078 Robust 1705 -2.57 0.121 Hypoxia 392 -0.71 0.245 Stem 3623 -0.25 0.044 Intrinsic 1576 -0.88 0.137 WR 11261 -0.37 0.037 ::: Prediction of survival with the prognostic index {#s2c} ------------------------------------------------ Adding together the weighted gene expressions for a particular gene set, each patient in the test set was assigned a Prognostic Index (PI). The distributions of the PIs are shown in [Figure 2A](#pone-0017845-g002){ref-type="fig"}; observe that some gene sets discriminate the patients on a wider range of risk scores than others. The deviance, an indicator of the models\' goodness of fit, was calculated for each gene set. [Table 4](#pone-0017845-t004){ref-type="table"} shows the difference in deviance (ΔD) between the fitted models and a null model with no genes. The magnitude of ΔD indicated the prediction power gained by a gene-set predictor. For gene sets with positive ΔD, the corresponding gene-set models were likely to perform poorly in prediction. Since no optimal *λ* was found for gene set *SD* and *LM*, the corresponding ΔD was 0. The standard deviation of the empirical PI distributions can be found in [Table 4](#pone-0017845-t004){ref-type="table"}. ::: {#pone-0017845-g002 .fig} 10.1371/journal.pone.0017845.g002 Figure 2 ::: {.caption} ###### Boxplot of predicted PIs on test data. (**A**) Systemic recurrence: The figure shows that the predicted PIs across all the studied gene sets were roughly centered around 0, resulting from the standardization procedure of the expression matrix on both training and test set for individual gene set in the model building stage. The standard deviations of PIs for individual gene set are following: RS: 0.109; SD: 0; LM: 0; AMST: 0.195; ROT: 0.095; Grade: 0.078; Robust: 0.121; Hypoxia: 0.245; Stem: 0.044; Intrinsic: 0.137; WR: 0.037. Due to lack of convergence, the predicted PIs by gene set SD and LM was calculated by setting tuning parameter *λ,* at a large value. (**B**) Breast cancer specific death (BC specific death): Boxplot of predicted PIs on test set. Gene set LM failed to converge in the model training. :::  ::: Similarities between PIs {#s2d} ------------------------ The predicted PIs for the test patients by each of the individual gene sets in our study were stacked into a PI matrix where rows indicate the identity of the gene set, columns indicate the identity of the patient in the test set and each cell contains the predicted PI risk score for relapse of a specific patient by a specific gene set ([Figure 1B](#pone-0017845-g001){ref-type="fig"}). Hierarchical clustering of the patients based on the PI matrix revealed two risk groups with distinct clinical characteristics ([Figure 3A](#pone-0017845-g003){ref-type="fig"}) and associated with significantly different survival probabilities ([Figure 3B:](#pone-0017845-g003){ref-type="fig"} Logrank test: χ^2^ = 7.8, df = 1, *p* = 0.005). This is in line with the findings from the results on the training set ([Figure S2](#pone.0017845.s002){ref-type="supplementary-material"}; see [Methods S1](#pone.0017845.s010){ref-type="supplementary-material"} Part IV for calculation of the estimated PIs on the training set and Part V for a brief discussion on the results). A control sample included in the Ull set (DNR\_N\_100; marked in green in [Figure 3A](#pone-0017845-g003){ref-type="fig"}, see [Materials and Methods](#s4){ref-type="sec"}) was correctly classified to the low risk group. Clinical and molecular characteristics for each of the two clusters are summarized in [Table 5](#pone-0017845-t005){ref-type="table"}. For the 77 patients in the test set, the low risk group consisted of 44 patients; 14 of them had developed systemic recurrence within the follow-up time period. Furthermore, 29 out of 38 luminal A tumors (76%), 5 out of 13 basal-like tumors (39%), 33 out of 45 ER positive tumors (73%), 32 out 51 PR positive tumors (63%), 2 out of 20 *TP53* mutated tumors (10%) and all 6 Grade 1 tumors (100%) belonged to this group. The high risk group consisted of 33 patients of whom 20 experienced relapse within the follow-up time period with 53.9 month median survival time. Moreover, 9 out of 38 Luminal A tumors (24%); 8 out of 13 basal tumors (61%) and almost all *TP53* mutated tumors (18/20, 90%) belonged to the high risk group. ::: {#pone-0017845-g003 .fig} 10.1371/journal.pone.0017845.g003 Figure 3 ::: {.caption} ###### Hierarchical clustering of predicted PIs on test set and Kaplan-Meier analysis of the clusters. Results for systemic recurrence are in (A--B); for BC specific death in (C--D). In heatmaps (A, C), rows are notations for the gene sets. Columns are annotation for the patients; data outside of 1% quantile were trimmed. "Average" linkage based on Spearman correlation was used to construct the dendrograms. Figure A and C share legends for the clinical parameters. (**A**) Heatmap of predicted PIs on the test set for systemic recurrence from each gene sets. Two risk groups were observed from the hierarchical clustering; cluster I and cluster II. The control sample in Ull DNR\_N\_100, marked by green, was classified in the cluster associated with a lower risk (cluster II). (**B**) The Kaplan-Meier curves for the two clusters. A significant separation between the two groups was observed (χ^2^ = 7.8, df = 1, *p* = 0.005). (**C**) Heatmap of predicted PIs on the test set for BC specific death from each gene sets. Two risk groups were observed from the hierarchical clustering; cluster I and cluster II. The control sample in Ull DNR\_N\_100, marked by green, was classified in the cluster associated with a lower risk (cluster I). (**D**) A significant separation between the two Kaplan-Meier curves associated with the clusters was observed (χ^2^ = 5.996, df = 1, *p* = 0.014). :::  ::: ::: {#pone-0017845-t005 .table-wrap} 10.1371/journal.pone.0017845.t005 Table 5 ::: {.caption} ###### Clinical and molecular characteristics of the two risk groups from hierarchical clustering of the test patients based on the predicted PI matrix. ::: {#pone-0017845-t005-5} Low risk High risk ---------------------------- ------------ ------------ Number of patients 44 33 Number of events (Recurr.) 14 20 Median survival (month) \- 53.9 LumA (%) 29/38 (76) 9/38 (24) Basal (%) 5/13 (39) 8/13 (61) ER+ (%) 33/45 (73) 12/45 (27) PR+ (%) 32/51 (63) 19/51 (37) *TP53* mutated (%) 2/20 (10) 18/20 (90) Grade 1 (%) 6/6 (100) 0/6 (0) ::: Prediction by individual gene sets {#s2e} ---------------------------------- The concordance structure for survival prediction among the studied gene sets is shown in the heatmap of the Spearman correlation matrix on continuous PI scales ([Figure 4A](#pone-0017845-g004){ref-type="fig"}). The gene sets *SD* and *LM* were left out since no optimal tuning parameter *λ* could be found. It should be noted that *Hypoxia* showed weak correlations with other gene sets, whereas *Robust*, *Grade* and *RS* were highly correlated. ::: {#pone-0017845-g004 .fig} 10.1371/journal.pone.0017845.g004 Figure 4 ::: {.caption} ###### Correlation structure of predicted PIs from gene sets with convergence in model-building stage. Heatmap of Spearman correlation matrix of predicted PIs for corresponding survival endpoint from individual gene sets. (**A**) For systemic recurrence, nine gene sets that reached convergence during modeling building are displayed. (**B**) For BC specific death, ten gene sets that reached convergence during modeling building are displayed. Figure A and B share the same color legend. :::  ::: To increase the clinical applicability of PI scores, patients with a positive PI score were assigned to the high risk group and the remaining patients to the low risk group. The survival probabilities associated with the dichotomized risk groups were assessed by the logrank test. Kaplan-Meier plots for the predicted groups are shown in [Figure 5](#pone-0017845-g005){ref-type="fig"} for each of the individual gene sets. Three gene sets were found to be significant: *Grade* (*p* = 0.012), *Robust* (*p* = 0.02) and *RS* (*p* = 0.027). Previously, we observed that predictions by these three gene sets were highly correlated, which explains their similar performances on the dichotomous scale of PI. In addition, two gene sets were borderline significant: *ROT* (*p* = 0.058) and *WR* (*p* = 0.077). ::: {#pone-0017845-g005 .fig} 10.1371/journal.pone.0017845.g005 Figure 5 ::: {.caption} ###### Systemic recurrence: Kaplan-Meier plot of the PI-risk groups for each of the individual gene sets. The Kaplan-Meier curves and the associated logrank *p* values for dichotomized PI-risk groups from each of the 9 converged gene-set models. :::  ::: Using principal components analysis to obtain a combined risk predictor {#s2f} ----------------------------------------------------------------------- The concordance structure among the studied gene sets for survival prediction ([Figure 3A](#pone-0017845-g003){ref-type="fig"} and [Figure 4A](#pone-0017845-g004){ref-type="fig"}) indicated that the main signal related to survival should be well captured by a lower dimensional representation of the PI vector associated with a patient. Consider the PI matrix (rows indicate gene sets and columns indicate patients) consisting of the predicted PIs of the test patients from the 9 converged gene sets in the model training. We used principal components analysis to derive linear combinations of the original nine-dimensional PI vectors that captured most of the variability of the PIs. [Figure 6A](#pone-0017845-g006){ref-type="fig"} shows a scatter plot of the combined risk prediction scores for the test patients on the first two principal components. The first principal component (PC1) captured 64% of the total variation and a total of 76% cumulative proportion of variation was captured by the first two components ([Table S1](#pone.0017845.s007){ref-type="supplementary-material"}). An increasing proportion of relapses were observed at the higher end of PC1. We created two classes by dividing the PC1 risk scores at the median value (−0.45); thus any patient with a PC1 risk score larger than −0.45 was considered to be a high-risk patient, and any patient with a PC1 risk score less than or equal to −0.45 was considered to be a low-risk patient. The two risk groups were associated with significantly different survival probabilities ([Figure 6B](#pone-0017845-g006){ref-type="fig"}, logrank test *p* = 0.003). ::: {#pone-0017845-g006 .fig} 10.1371/journal.pone.0017845.g006 Figure 6 ::: {.caption} ###### Systemic recurrence: PCA of predicted PIs from converged gene sets and performance of the resulting groups from PC1. Results for systemic recurrence are in (A-B); for BC specific death in (C-D). (A) Scatter plot of predicted PIs from 9 converged gene-set models on the space of the top two leading PCs. Black circles indicate censored observations; red dots indicate patients with relapse. (B) The Kaplan-Meier curves for high and low risk groups are significantly different (χ^2^ = 8.76, df = 1, *p* = 0.003). (C) Scatter plot of predicted PIs from 10 converged gene-set models on the space of the top two leading PCs. Black circles indicate censored observations; red dots indicate patients with BC specific death; brown stars indicate death from other reasons. (D) The Kaplan-Meier curves for high and low risk groups are significantly different (χ^2^ = 10.26, df = 1, *p* = 0.001). :::  ::: Univariate comparison of predictors {#s2g} ----------------------------------- We performed univariate analysis for individual gene-set predictors as well as for clinical parameters including tumor size (pT1, pT2 and pT3-pT4), *TP53* mutation status, stage (1--4), node status (pN0, pN1, pN2-pN3 and pNx), ER status (positive versus negative), histological grade (1--3) and the Adjuvant! Online model (AOL), respectively. AOL is an established on-line breast cancer survival predictor; it calculates a 10-year survival probability based on the patient\'s age, tumor size, tumor grade, oestrogen-receptor status, and nodal status. Patients were assigned to the low risk group if their 10-year mortality risk was lower than 10% as predicted by Adjuvant! Online software. The dichotomized PI scores (positive scores indicate high risk and nonpositive scores indicate low risk) for the gene-set predictors were used in the univariate Cox model. The performance comparisons by using the likelihood ratio test, the deviance, the *proportion of variation explained* (PVE), the *concordance index* (C-index) and the *Hazard Ratio* (HR) are summarized in [Table 6](#pone-0017845-t006){ref-type="table"}, [Figure 7](#pone-0017845-g007){ref-type="fig"} and [Table S2](#pone.0017845.s008){ref-type="supplementary-material"} for the combined-PI risk predictor (where we dichotomize on PC1) and other included predictors. ::: {#pone-0017845-g007 .fig} 10.1371/journal.pone.0017845.g007 Figure 7 ::: {.caption} ###### Univariate comparison of predictors for systemic recurrence. Comparison of combined-PI risk predictor with clinical parameters and individual gene-set predictors using univariate Cox model. (**A**) Y axis indicates C-index associated with individual predictor and X axis indicates the p values (on minus log10 scale) from likelihood ratio test in univariate Cox model. C-index  = 0.5 and the significant level: α = 0.05 for the likelihood ratio test are indicated by the dotted line. The size and the color of the bubble indicate the PVE and the deviance in univariate Cox model, respectively. The combined-PI risk predictor had the most significant marginal effect for predicting systemic recurrence (*p* = 0.003). It was associated with the second highest C-index score (C = 0.75) following *TP53* mutation status (C = 0.76). It had the second highest deviance (8.61) following tumor size (9.36), and the combined-PI predictor alone explained 10.6% of the variability as indicated by PVE, following tumor size (11.7%) and stage (11.1%) (**B**) X axis indicates HR from the univariate Cox model and the 95% CIs are shown along with the point estimates. "LR test" stands for likelihood ratio test. Insignificant predictors (likelihood ratio test *p*\>0.05) are grayed out. To keep the results interpretable, only predictors with two levels are compared. The combined-PI risk predictor had the 2^nd^ largest HR (2.82 with 95% CI 1.37---5.80) following *TP53* mutation status (2.87 with 95% CI 1.42---5.83). :::  ::: ::: {#pone-0017845-t006 .table-wrap} 10.1371/journal.pone.0017845.t006 Table 6 ::: {.caption} ###### PCA-combined PI risk predictor in univariate and multivariate Cox regression. ::: {#pone-0017845-t006-6} Univariate Multivariate ---------------------- ------------------------------------------------ ------------ -------------- ------------- ------- ---------------------- ------- ------- -- Combined-PI risk (overall effect) 0.003 0.106 8.611 0.746 11.3% High (vs Low) 3.34 \[1.49, 7.51\] 0.004 AOL risk 0.131 0.033 2.283 0.734 *TP53* 0.006 0.092 7.47 0.759 Tumor size (overall effect) 0.009 0.117 9.358 0.706 5% pT2 (vs pT1) 2.59 \[0.86, 7.81\] 0.091 pT3-pT4 (vs pT1) 2.97 \[0.85, 10.42\] 0.089 Stage 0.058 0.111 7.501 0.698 Node 0.107 0.076 6.090 0.605 ER 0.634 0.003 0.227 0.403 Histological grade 0.372 0.025 1.979 0.480 RS 0.028 0.061 4.827 0.686 AMST 0.148 0.027 2.092 0.630 ROT 0.058 0.046 3.591 0.686 Grade 0.013 0.077 6.174 0.712 Robust 0.02 0.068 5.391 0.703 Hypoxia 0.303 0.014 1.06 0.626 Stem 0.734 0.001 0.115 0.535 Intrinsic 0.508 0.006 0.437 0.538 WR 0.079 0.039 3.093 0.640 *Multivariate model* Risk  =  Combined-PI + Tumor Size + strata(ER) 0.002 0.19 (R^2^) 0.705 For overall effect of the predictor in univariate Cox regression, Likelihood ratio test p value was reported. For individual levels within the predictor, Wald test *p* value was reported. \*PVE: proportion of variation explained in the outcome variable. § C: concordance index. ::: The combined-PI risk predictor was competitive in all the tested measurements. It showed the most significant effect on survival (*p* = 0.003) and it was associated with the second highest C-index score (C = 0.75) following *TP53* mutation status (C = 0.76). The information carried by the Deviance and PVE is highly consistent ([Figure 7A](#pone-0017845-g007){ref-type="fig"}). The combined-PI risk predictor had the second highest deviance (8.61) following tumor size (9.36), indicating good fit of the model. Furthermore, the predictor alone explained 10.6% of the variability as indicated by PVE, following tumor size (11.7%) and stage (11.1%); see [Table 6](#pone-0017845-t006){ref-type="table"}. The high risk group assigned by the combined-PI risk predictor had a hazard rate of 2.82 (95% CI 1.37---5.80; [Figure 7B](#pone-0017845-g007){ref-type="fig"}; [Table S2](#pone.0017845.s008){ref-type="supplementary-material"}) times higher than the low risk group. Among all the tested predictors, *TP53* mutation status was the only factor that gave a slightly higher HR (2.87 with 95% CI 1.42---5.83; [Figure 7B](#pone-0017845-g007){ref-type="fig"}; [Table S2](#pone.0017845.s008){ref-type="supplementary-material"}). The proportional hazards assumption, where the hazard ratios over time are constant, held in all the reported univariate Cox models, expect for ER status (*p* = 0.009; [Table S2](#pone.0017845.s008){ref-type="supplementary-material"}). Multivariate comparison of predictors {#s2h} ------------------------------------- A multivariate Cox model was used to simultaneously assess the combined-PI risk predictor and the traditional clinical and molecular parameters that yielded significant results in the univariate comparison (*TP53* mutation status and tumor size). Due to the known association between ER status and survival, we included ER status as stratification variable and each stratum is permitted to have a different baseline hazard function while the coefficients of the remaining covariates are assumed to be constant across the strata. We observed a high correlation between *TP53* mutation status and the combined-PI risk predictor (odds ratio 15.0 (95% CI 3.1---145.7), Fisher\'s Exact test *p*\<0.001). Analysis for model comparison showed that the combined-PI predictor added significant information to tumor size and *TP53* mutations (analysis of deviance *p* = 0.04; Akaike\'s Information Criterion (AIC) for model with and without combined-PI  = 191.8 and 194.01, respectively; [Table S2](#pone.0017845.s008){ref-type="supplementary-material"}). We used AIC in a hybrid stepwise strategy to build a final prognostic model where the combined-PI and tumor size were left as covariates and ER status as stratification variable in a Cox regression model (AIC: 191.46; [Table S2](#pone.0017845.s008){ref-type="supplementary-material"}. Likelihood ratio test *p* = 0.002; [Table 6](#pone-0017845-t006){ref-type="table"}). The final model exhibited proportional hazards (*p* = 0.85; [Table S2](#pone.0017845.s008){ref-type="supplementary-material"}). The combined-PI risk predictor had the most significant effect in the final multivariate model (*p* = 0.004). The HR for the combined-PI high-risk group was 3.34 (95% CI 1.49---7.51) compared with the low risk group. The partial PVE was calculated to indicate relative importance of the individual predictor in the multivariate setting: The combined-PI risk predictor captured 11.3% of the variability compared with 5% captured by tumor size. The C-index (C = 0.71) indicated satisfactory predictive discrimination ability for the final multivariate model. Breast cancer specific death as clinical endpoint {#s2i} ------------------------------------------------- The proposed framework was also applied with breast cancer specific death as the clinical endpoint and the results are presented in [Figure 2](#pone-0017845-g002){ref-type="fig"}-[](#pone-0017845-g003){ref-type="fig"} [4](#pone-0017845-g004){ref-type="fig"}, [6](#pone-0017845-g006){ref-type="fig"}. The predicted PI scores of the risk of dying from breast cancer by individual gene sets are summarized in the box plot in [Figure 2B](#pone-0017845-g002){ref-type="fig"}. The tuning parameters chosen by leave-one-out cross-validation are shown in [Figure S3](#pone.0017845.s003){ref-type="supplementary-material"}. The correlation structure indicates concordant predictions made by different gene sets ([Figure 4B](#pone-0017845-g004){ref-type="fig"}). Overall, the gene set SD had the weakest correlation to the other gene sets. Hierarchical clustering of the test set patients based on the predicted PI matrix defined to two risk clusters ([Figure 3C](#pone-0017845-g003){ref-type="fig"}) with distinct clinical characteristics and associated with significantly different survival probabilities (*p* = 0.014); see [Figure 3D](#pone-0017845-g003){ref-type="fig"}. Risk scores were combined as previously described into one risk score by projecting the PIs onto the first principal component ([Figure 6C](#pone-0017845-g006){ref-type="fig"}), in which 73% of the total variance was captured. The continuous score was further dichotomized into high risk and low risk using the median of the PC1 score. The difference in the associated survival probabilities of the two predicted risk groups was confirmed by the logrank test (*p* = 0.001; Kaplan-Meier plot in [Figure 6D](#pone-0017845-g006){ref-type="fig"}). For comparison, the univariate performances of individual gene sets for dichotomous prediction and clinical parameters as well as AOL are shown in [Figure S4](#pone.0017845.s004){ref-type="supplementary-material"} and [Methods S1](#pone.0017845.s010){ref-type="supplementary-material"}. The combined-PI risk predictor for breast cancer specific death was the most significant among all the tested predictors in the univariate setting (likelihood ratio test *p* = 0.001) with the second largest HR 3.36 (95% CI 1.5---7.4), following *TP53* mutation status (HR 3.46 with 95% CI 1.7---7.2; [Figure S4B](#pone.0017845.s004){ref-type="supplementary-material"}). Furthermore, it had the second largest C-index (C = 0.77) following *TP53* (C = 0.8; [Figure S4A](#pone.0017845.s004){ref-type="supplementary-material"}) and it was also highly ranked by PVE (12.4%) and deviance (10.2) following node status (PVE = 12.5%, Deviance = 10.3; [Figure S4A](#pone.0017845.s004){ref-type="supplementary-material"}). The combined-PI risk predictor was the most significant predictor (*p* = 0.005; [Methods S1](#pone.0017845.s010){ref-type="supplementary-material"}) with HR 4.62 (95% CI 1.58---13.55) in a multivariate Cox model containing combined-PI risk, *TP53* mutation status (HR = 2.6, 95% CI 0.86---7.78), tumor size and node status, stratified by ER status (See [Methods S1](#pone.0017845.s010){ref-type="supplementary-material"} for the AIC stepwise model selection). The correlation between the combined-PI predictor and *TP53* mutation status appeared to be significant (odds ratio 33.9 with 95% CI 4.8---1490.4, Fisher\'s Exact test *p*\<0.001; [Methods S1](#pone.0017845.s010){ref-type="supplementary-material"}) and the combined-PI added significant information to *TP53*, tumor size and node status (analysis of deviance *p* = 0.035; AIC for model with and without combined-PI  = 149.96 and 152.41, respectively; [Methods S1](#pone.0017845.s010){ref-type="supplementary-material"}). Except for the univariate Cox model for ER status, the proportional hazards assumption was met for the reported models ([Methods S1](#pone.0017845.s010){ref-type="supplementary-material"}). Assess robustness of the combined-PI model {#s2j} ------------------------------------------ To test the robustness of the proposed procedure, we performed the analyses by switching the test and training sets. To gain computational efficiency, we applied 10-fold CV instead of LOOCV during the model training procedure for each of the gene sets. We obtained similar results for prediction of BC specific death ([Figure S5E--H](#pone.0017845.s005){ref-type="supplementary-material"}). The combined-PI predictor and *TP53* mutation status was strongly correlated (odds ratio 11.2 with 95% CI 4.0---36.8, Fisher\'s Exact test *p*\<0.001). For prediction of systemic recurrence, only three out of the eleven tested gene sets achieved model convergence, which most likely contributed to a compromised combined-PI risk predictor built from insufficient signals ([Figure S5A--D](#pone.0017845.s005){ref-type="supplementary-material"}). Furthermore, we also checked the stability of the combined-PI prediction by correlating the predicted combined-PIs on the test set from one analysis with the combined-PIs on the training set from the switched analysis. The continuous PCA combined-PI scores were used instead of dichotomizing on PC1. We performed this screening for the prediction of BC specific death, and observed moderate correlations of the combined-PIs on the MicMa set (0.48) as well as on the Ull set (0.63). This indicates clearly that there is a fair correlation between the combined-PIs when trained on different data sets, but also that there is substantial variation. Discussion {#s3} ========== High-throughput genome profiling for genetic marker discovery is widely applied in the field of genomic and personalized medicine. For example, a gene signature characterizes certain biological aspects and/or may be used to predict disease outcome by mathematically combining the expression values of a set of genes. The methods generating the combined expression pattern of the biomarkers differ from study to study. A gene signature therefore consists of a group of gene identities together with a distinct classifier or a predictor to predict disease outcome. In the present work, we applied Cox regression to reconstruct a classifier from the same set of original genes as a number of published gene signatures. Only the associated gene identities from the original signature are retained, while the classifiers themselves are derived on the basis of the same modeling procedure and the same tumor data set. As a consequence, the resulting gene signatures are derived on the basis of the same clinical endpoint (systemic relapse or breast cancer specific death). The gene sets in our study were pre-selected based on prior knowledge in breast cancer. We are interested in survival prediction by the collective expression pattern of the whole gene set without additional gene selection. Furthermore, since the number of features (genes) outnumbers the number of samples in the training set, p *\> N*, for most of the gene sets of interest, direct estimation of the coefficients using standard Cox regression is unfeasible. Even for those gene sets with p *\< N,* we still have p large enough to render the coefficient estimates highly unstable and thus of doubtful value for prediction. Accordingly, we perform parameter estimation using a penalized partial likelihood criterion [@pone.0017845-Verweij1] that forces the solution to the estimation problem to have small L2-norm. Using an L2 penalty trades a little bias for a larger reduction in variance to reduce the prediction error on a new data set. This is the so-called "bias and variance tradeoff" [@pone.0017845-Trevor1]. Cox-ridge regression has been shown to be an effective model in survival prediction using gene expression data [@pone.0017845-Bvelstad1], [@pone.0017845-VanWieringen1]. There is no actual variable selection involved; while genes are generally down weighted, all the genes included in the Cox model will be present in the final model. We observed that two of the pre-selected gene sets failed for survival prediction at the model training stage: gene set LM for both tested clinical endpoints, and gene set SD for systemic relapse. The reason for their poor performance is likely related to the lack of comparability of the data and methods used to construct the original gene signatures and those used to construct ours. Gene set LM contains 54 genes that mediate risk of breast cancer metastasis to lung [@pone.0017845-Minn1]. The original risk index for lung metastasis was defined as a linear combination of gene expression values weighted by their estimated Cox regression parameters. The essential biological information captured in the original Minn et. al dataset [@pone.0017845-Minn1] was absent from the cohorts studied in the present work, which contain mainly early-stage breast cancers. Given the comparable methods in classifier construction, we believe the biological incomparability of data sources used for model building led to the poor performance of gene set LM in our study. Likewise, the gene set SD might not have been expected to perform well due to the intrinsic biological differences in the training sets used for the present study and the original study, where microdissected stroma from breast cancer specimens were used to identified these 26 stroma-derived prognostic genes [@pone.0017845-Finak1]. Not all gene sets that were left for the model evaluation achieved significant survival prediction in the studied test set ([Figure 5](#pone-0017845-g005){ref-type="fig"} and [Table 6](#pone-0017845-t006){ref-type="table"}). The inter-cohort heterogeneity most likely played a role. The studied training and test set showed borderline significantly different survival probabilities for the survival endpoints (systemic relapse *p* = 0.0765; breast-cancer specific death *p = *0.0564; [Figure S6](#pone.0017845.s006){ref-type="supplementary-material"}). Unless there exists a "gold-standard" dataset representing perfectly the underlying population for breast cancer, dealing with the heterogeneity feature is inevitable. The proposed framework demonstrates a straightforward yet effective approach to improve the survival prediction power by integrating multiple gene set predictors. Interestingly the combined-PI risk predictor correlates with but provides additional information to *TP53* mutation status. *TP53* mutation status has been shown to be one of the strongest single molecular prognostic markers in breast cancer. It is known as a key molecule involved in different pathways important in cancer. It is a molecular marker important to compare with other prognostic markers to gain insights about the underlying biology. It is a prime example of a single molecule not function alone, but rather involving many players in various networks. Restoring *TP53* activity is a potential therapeutic strategy [@pone.0017845-Wang2]. The results indicated that to some extent the generalization from MicMa set to Ull set was better than that from Ull to MicMa. We suspect the main reason may be the reduction of the training set sample size from 123 in the MicMa cohort to 80 in the Ull cohort. Since the training step is prone to over-fitting due to the large number of genes from which the models may be fitted, the ability to fit a reliable model strongly depends on having a sufficient sample size. As a potential extension of the framework of mining large collection of gene sets for survival prediction, an optional filter step could be added prior to the analysis to eliminate gene sets that are not significantly related to survival or not enriched in the training set. Our analysis framework does not restrain to the gene sets from gene signatures, as this is one of the many ways to provide input to the "combining power procedure". In conclusion, our proposed framework for improving survival prediction contains three analytical modules: (1) gene signature (gene-set prognostic model) construction, (2) dimension reduction and (3) risk prediction. Each module could be fine tuned or modified depending on the data under study. Our study showed that by aggregating the predictive strength from multiple gene sets we can improve the outcome prediction in breast cancer, and it can be broadly applicable to breast cancer survival risk assessment. Materials and Methods {#s4} ===================== Ethics statement {#s4a} ---------------- The studies included in the project were approved by the Regional Ethical Committee (REK: S97103 for MicMa and REK: 200401129-1 for Ull). All samples were obtained with written informed consent approved by the ethical committee. Tumor samples and patients {#s4b} -------------------------- The training set (MicMa) comprise published gene expression and clinical data on 123 human breast cancer cases, mainly stage I and II [@pone.0017845-Naume1]. Patients were treated for localized breast cancer and included between 1995 to 1998 [@pone.0017845-Wiedswang1]. The prognostic model for relapse was trained using 118 patients with available endpoint status. All the 123 patients were available for training the prognostic model for breast cancer specific death. The test set (Ull) mainly consisted of early stage breast cancers from which gene expression and clinical data were available [@pone.0017845-Langerd1]. Patient samples were sequentially collected from 1990 to 1994. In this data set, together with eighty tumor cases a normal breast tissue sample coded as "DNR\_N\_AO100" was included as control. Descriptive aspects of the training and test sets are given in [Table S3](#pone.0017845.s009){ref-type="supplementary-material"}. More details are provided in [Methods S1](#pone.0017845.s010){ref-type="supplementary-material"}. Microarray {#s4c} ---------- Both the training set (MicMa) and test set (Ull) were obtained using the Stanford 43k cDNA microarrays. Using the protocols described in [Methods S1](#pone.0017845.s010){ref-type="supplementary-material"}, total RNA was isolated, amplified, labeled and hybridized to arrays containing around 42,000 features representing 23169 unique cluster IDs (UniGene Build Number 215) produced at the Stanford Functional Genomics Facility (<http://www.microarray.org/sfgf/jsp/home.jsp>). All procedures are available at Stanford Genomics Breast Cancer Consortium Portal (<http://genome-www.stanford.edu/breast_cancer/>) for the MicMa set and at the referred web site (<http://www.stanford.edu/group/sjeffreylab/>) for the Ull set. All data is MIAME compliant. Raw data for MicMa set has been deposited in the NCBI\'s Gene Expression Omnibus database (GEO; <http://www.ncbi.nlm.nih.gov/geo/>) with accession number GSE3985. Both datasets are accessible through Stanford Microarray Database (SMD; for the MicMa set: <http://smd.stanford.edu/cgi-bin/publication/viewPublication.pl?pub_no=833>; for the Ull set: <http://smd.stanford.edu/cgi-bin/publication/viewPublication.pl?pub_no=629>). Preprocessing {#s4d} ------------- For the MicMa set, normalized (loess with print tip stratification) log2-transformed gene expression ratios were retrieved from SMD (<http://smd.stanford.edu/>), filtered for spot intensity over background at least 1.5 in both sample and reference, and finally, filtered for those genes that fulfilled the spot filter criteria in at least 85% of the experiments. For the Ull set, probes were filtered for spot quality and included in the analysis if the pixels within a spot showed a regression correlation of at least 0.6 or if the signal intensity of both sample and reference were at least 1.5 over background; data were further normalized by the default total Intensity Normalization in SMD (<http://smd.stanford.edu/help/results_normalization.shtml>). For both MicMa and Ull, k-nearest neighbor (KNN) imputation [@pone.0017845-Troyanskaya1] with k = 10 was applied to impute missing values in the filtered expression dataset. Gene sets {#s4e} --------- An overview of the 11 published gene sets studied in this report are listed in [Table 1](#pone-0017845-t001){ref-type="table"}. The abbreviated coding for the individual gene sets are used instead of the full name throughout this paper. The annotation files for the gene sets were downloaded from the web sites indicated in the original publications or were requested directly from the authors. Brief discussions of various gene sets are presented in the [Methods S1](#pone.0017845.s010){ref-type="supplementary-material"}. Adjuvant! Online model {#s4f} ---------------------- Adjuvant! [@pone.0017845-Ravdin1] (<https://www.adjuvantonline.com>) uses patient age, comorbidity level, ER status, tumor grade, tumor size and number of positive lymph nodes to predict 10-year risk for breast cancer mortality. It also predicts the benefit of adjuvant therapy for women with early-stage breast cancer. A description of the Adjuvant! Online model and the details for computing risk scores of the test patients can be found in [Methods S1](#pone.0017845.s010){ref-type="supplementary-material"}. Low risk is defined in this paper as lower than 10% 10-year risk of breast cancer mortality. A total of 72 patients in the Ull data set with suitable characteristics for Adjuvant! Model were entered into the comparison study. Molecular subtype assignment {#s4g} ---------------------------- Tumors were assigned to a subtype using Pearson correlation to the expression centroids as previously described [@pone.0017845-Langerd1], [@pone.0017845-Srlie2]. Cross-platform gene identity mapping {#s4h} ------------------------------------ Most of the gene sets in the study were developed from microarray platforms different from Stanford 43k cDNA array (see [Table 1](#pone-0017845-t001){ref-type="table"}). For gene sets developed from Stanford 43k cDNA arrays, clone IDs were used as the "linker" to link the gene identifiers in the gene set to cDNA. For gene sets developed from different platforms, gene identifiers in the gene set were mapped to the cDNA clones using the following linkers: UniGene, gene symbol and gene alias, where UniGene had the highest priority while gene alias had the lowest priority. When matches occurred by using multiple linkers, we took the matches from the highest ranked linker. Furthermore, we collapsed the matched clones representing the same gene by their mean expression value ([Figure 1A](#pone-0017845-g001){ref-type="fig"}). Annotations for Stanford 43k cDNA array were retrieved from SMD SOURCE (<http://smd.stanford.edu/cgi-bin/source/sourceSearch>) under UniGene Build Number 215. And annotations for the individual gene sets were retrieved either from manufacture chip annotation files or from SMD. Penalized Cox regression for survival prediction based on individual gene sets {#s4i} ------------------------------------------------------------------------------ We assume that at any given time, a breast cancer patient has a certain risk of experiencing a specific event, which in our case is either relapse after primary surgery or breast cancer related death. We furthermore assume that for a number of patients we have measured the time to this event. Patients that experience none of these events for the duration of the study are labeled as censored. For such patients, the recorded time is simply the last point of observation and carries a different interpretation than a time to an event. To associate the risk of an event to observed features, we consider a Cox\'s proportional hazards model (*Cox model*) with the expression of selected genes as the covariates. Suppose that for a patient we have observed a total of *p* expression values . In principle, we may model the risk of relapse/death as a function of all measured expression values, as in the Cox model , where the interpretation is that the instantaneous risk (also known as the *hazard*) of experiencing the event at time *t* is a product of two functions, the first depending only on the time point (and not on the particular patient at hand), and the second depending only on the expression values of the patient (and not on the time point). In a Cox model, the relative hazard between two individuals with expression vectors and respectively is expressed by the quantity Accordingly, even though the hazard fluctuates over time for any one individual, the relative hazard between individuals is constant over time. Let denote a subselection of genes (variables) from the complete list of *p* genes. Considering the risk of relapse/death as a function of the selected genes only, we obtain the Cox model . In this paper, we want to model risk of relapse/death as a function of R = 11 gene sets of individual sizes , each corresponding to a particular subselection of genes. Thus, we consider the Cox modelswhere *j = 1,2,...,R* indexes the gene sets, is the selected genes (variables) in the *j*th gene set, is the vector of coefficients (weights) associated with the genes in the *j*th gene set, and is the (common) baseline hazard function. Since the response of interest is a possibly censored survival time, for a given gene set, a Cox proportional hazards model was used to describe the risk of a patient experiencing an event in response to expression of gene covariates. The expression values for each gene were mean centered and scaled to unit standard deviation in both the training and the test datasets. All Cox models in the gene signature construction part of the analysis were fitted using an L2-penalized partial Cox likelihood, whereas the Cox model in the multivariate comparison was fitted using the ordinary partial Cox likelihood. To determine the penalty parameter, we applied the leave-one-out cross validation procedure proposed by Verweij and van Houwelingen [@pone.0017845-Verweij1] (See [Methods S1](#pone.0017845.s010){ref-type="supplementary-material"}). A total of 118 MicMa patients with available information on systemic recurrence status were used to develop a prediction model using Cox-ridge regression for a gene set. For the *j*\'th gene set (*j* = 1, ..., 11), the optimal gene-set specific tuning parameter was found by the leave-one-out cross validation procedure [@pone.0017845-Verweij1] using the training set; we then estimated the coefficient vector associated with individual genes in gene set *j* by the Cox-ridge model (1) ([Figure 1A](#pone-0017845-g001){ref-type="fig"}). The predicted prognostic index for the *j*th gene set and the *i*th patient in the test data set was calculated as which is sum of weighted gene expression of the test patient *i* ([Figure 1A](#pone-0017845-g001){ref-type="fig"}). The weights for individual genes in the *j*th gene set were the corresponding estimated coefficients from the training set using the Cox-ridge model (1) ([Figure 1A](#pone-0017845-g001){ref-type="fig"}). See [Methods S1](#pone.0017845.s010){ref-type="supplementary-material"} for details. Prediction performance evaluation for individual gene set {#s4j} --------------------------------------------------------- Spearman correlations were used to show the concordance structure for risk prediction among the gene sets. The change in deviance (ΔD) is an indicator of predictive performance of a model obtained from a training dataset on a novel test dataset and is given by where is the optimal log-likelihood with all the genes in the gene set included, and is the optimal log-likelihood with no genes included. The continuous PI scores were dichotomized into high- and low-risk groups using zero as cutoff, where a positive PI score indicates high risk. The resulting PI risk groups for individual gene set were tested in a univariate Cox model and the difference of the Kaplan-Meier survival curves of the groups were also tested by logrank test. Hierarchical clustering {#s4k} ----------------------- As a graphical illustration of the relationship among the gene sets, the predicted PIs for all patients in the test set (Ull) using each of the gene sets were clustered using hierarchical clustering with Spearman correlation and average distance. The Kaplan-Meier survival curves were plotted for the resulting groups and the differences in clinical indications among the clusters were tested by a logrank test. Dimension reduction by principal components analysis {#s4l} ---------------------------------------------------- Principal components analysis (PCA) was used to project the PIs from each of the individual gene sets onto a lower dimensional space ([Figure 1B](#pone-0017845-g001){ref-type="fig"}). The values off the first principal component were used as the combined risk scores, and further dichotomized by median cut, where a patient with a score on PC1 higher than median values on PC1 was considered to be high risk. A logrank test was carried out to assess the significance of the differences in survival probabilities associated with resulting dichotomous risk groups ([Figure 1B](#pone-0017845-g001){ref-type="fig"}). The method of choice was PCA based on the covariance matrix since we expected PIs with higher variance to carry more information than PIs with less variance. In analyses of relapse, however, this ended up dominated by Hypoxia although this correlated little with other PIs, and was therefore replaced by PCA based on the correlation matrix. Univariate comparison of predictors {#s4m} ----------------------------------- Univariate Cox models were used to compare the effects of the *combined-PI risk predictor,* individual gene signatures (dichotomized by splitting at 0), clinical parameters (Tumor size and Histological grade, *TP53*, Node status, ER status and Stage), as well as the Adjuvant! Online model. A *likelihood ratio test* was used to assess the significance of the overall effect of a predictor in the univariate models. In addition, *deviance* was used to check the goodness of the model fit. The marginal contribution by a single predictor in the univariate setting was evaluated using the *proportion of variation explained* in the outcome variable (PVE) [@pone.0017845-Schemper1]. PVE , comparable with the R^2^ in regression modeling is an indicator that quantifies the importance of covariates in the Cox model. The *Hazard Ratio* (HR) was used as an accuracy measure for the risk group prediction for different predictors. In the univariate setting, HR is a summary of the risk difference between patient groups defined by the predictor. To keep the results interpretable and comparable, we presented the HRs for the predictors with two risk groups (excluded Tumor size, Histological grade, Node status and Stage). The larger the HR, the better is the discrimination between the groups of the patients, such as low- and high-risk. The *concordance index* (C-index) [@pone.0017845-Harrell1], an analogy to area under the receiver operating characteristic (ROC) curve in survival analysis, was computed to assess the predictive discrimination ability of each of the predictors in the corresponding univariate Cox model ([Methods S1](#pone.0017845.s010){ref-type="supplementary-material"}). It measures the probability of concordance between the predicted and observed responses in terms of lengths of time to event of any two patients. The larger the C-index, the better is the predictability of a survival model. A value of 0.5 indicates no predictive discrimination and a value of 1 indicates perfect separation of patients with different outcomes [@pone.0017845-Harrell1]. Multivariate comparison of predictors {#s4n} ------------------------------------- The significant predictors in the univariate analysis were included in a multivariate Cox model. Model selection was carried out using Akaike\'s Information Criterion (AIC) [@pone.0017845-Akaike1] and analysis of deviance. The relative importance of a covariate in a multivariate Cox model was measured by the partial PVE, which was calculated as the difference between for the full model and for a model with a factor of interest excluded. See [Methods S1](#pone.0017845.s010){ref-type="supplementary-material"} for details. Software {#s4o} -------- All analyses were performed in R (version 2.11.1), which is available at <http://cran.r-project.org/>. The R package "penalized" [@pone.0017845-Goeman1] was to perform penalized Cox regression. R code for the procedures described in this paper are available from the correspondence author. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### **Systemic recurrence: Cross-validated likelihood profile on** ***λ*** **grid.** The dotted line indicates the location of the optimal *λ* value *λ*opt for each gene set. RS: 338; SD: Inf; LM: Inf; AMST: 227; ROT: 1029; Grade: 3580; Robust: 1705; Hypoxia: 392; Stem: 3623; Intrinsic: 1576; WR: 11261. Modeling gene set SD and LM did not reach convergence by the specified criteria in the study. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### **Systemic recurrence: Hierarchical clustering of estimated PIs for systemic recurrence on training set and the resulting risk clusters in a Kaplan-Meier plot.** (**A**) Heatmap of estimated PIs on training set for systemic recurrence from each gene sets. Rows are notations for the gene sets. Columns are annotation for the patients; data outside of 1% quantile were trimmed. "Average" linkage based on Spearman correlation was used to construct the dendrograms. Two risk clusters I and II, were observed from the hierarchical clustering with distinct clinical characteristics: a total of 30 out of 49 Luminal A tumors (61%) were clustered in the low risk group, and 19 luminal A tumors (39%) were found in the high risk group. (**B**) The Kaplan-Meier curves for cluster I and cluster II. A significant separation between the two clusters was observed (χ2 = 49.7, df = 1, *p*\<0.001). (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S3 ::: {.caption} ###### **BC specific death: Cross-validated likelihood profile on** ***λ*** **grid.** The purple dotted line indicates the location of the optimal *λ* value *λ*opt for each gene set. RS: 343; SD: 249; LM: Inf; AMST: 230; ROT: 379; Grade: 1337; Robust: 898; Hypoxia: 1823; Stem: 3866; Intrinsic: 1639; WR: 8317. Modeling gene set LM did not reach convergence by the specified criteria in the study. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S4 ::: {.caption} ###### **Univariate comparison of predictors for BC specific death.** (**A**) Y axis indicates C-index associated with individual predictor and X axis indicates the p values (on minus log10 scale) from likelihood ratio test in univariate Cox model. C-index  =  0.5 and the significant level: α  =  0.05 for the likelihood ratio test are indicated by the dotted line. The size and color of the bubble indicates the PVE and deviance in univariate Cox model, respectively. The combined-PI risk predictor for BC specific death was the most significant one among all the tested predictors (likelihood ratio test *p*  =  0.001). It had the second largest C-index (C  =  0.77) following *TP53* (C  =  0.8). And it was also highly ranked by PVE (12.4%) and deviance (10.2) following node status (PVE  =  12.5%, Deviance  =  10.3). (**B**) X axis indicates HR from the univariate Cox model and the 95% CIs are shown along with the point estimates. "LR test" stands for likelihood ratio test. Insignificant predictors (likelihood ratio test *p* \> 0.05) are grayed out. The combined-PI risk predictor had the second largest HR 3.36 (95% CI 1.5---7.4), following *TP53* mutation status (HR 3.46 with 95% CI 1.7---7.2). (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S5 ::: {.caption} ###### **Summary of switched analysis by using Ull as training set and MicMa as test set.** Results for systemic recurrence are in (A-D); for BC specific death in (E-H). (**A**) Boxplot of predicted PI for systemic recurrence. Only three out of eleven gene sets had converged model in model training stage. (**B**) Spearman correlation structure among gene sets with convergence. (**C**) Projection of the predicted PIs on space formed by PC1 (captured 81% variability) and PC2. (**D**) Kaplan-Meier curves associated with the two risk groups by median-cut of PC1 value (logrank *p*  =  0.191). (**E**) Boxplot of predicted PI for BC specific death. (**F**) Spearman correlation structure among gene sets with convergence. (**G**) Projection of the predicted PIs on space formed by PC1 (captured 65% variability) and PC2. (**H**) Kaplan-Meier curves associated with the two risk groups by median-cut of PC1 value (logrank *p*  =  0.028). (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S6 ::: {.caption} ###### **Survival curves for training set (MicMa) and test set (Ull).** The logrank test showed that the training cohort and test cohort had borderline significant survival curves for the survival endpoint. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S1 ::: {.caption} ###### **Results summary of PCA on predicted PIs for systemic recurrence from 9 converged gene sets.** (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S2 ::: {.caption} ###### **Results summary of univariate and multivariate analysis of combined PI-risk predictor for systemic recurrence.** (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S3 ::: {.caption} ###### **Molecular and clinicopathological characteristics of the tumor material in MicMa and Ull datasets.** (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: Methods S1 ::: {.caption} ###### **Supplementary methods, results, notes, and references.** (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This study was supported by Norwegian Research Council (NFR) FUGE program, Grant 175240 and NFR Cancer program, Grant 193387. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: ALBD OCL XZ VNK. Performed the experiments: TS AL. Analyzed the data: XZ OCL EAR. Contributed reagents/materials/analysis tools: ALBD BN. Wrote the paper: XZ OCL EAR TS ALBD. Contributed to manuscript and discussions: AF VNK.

PubMed Central

2024-06-05T04:04:19.888463

2011-3-10

PMC3053399

Introduction {#s1} ============ *Bordetella pertussis* is a human-specific pathogen that is the etiologic agent of whooping cough, an acute respiratory disease that is often particularly severe in infants [@pone.0017797-Mattoo1]. Universal immunization programs have contributed to a significant reduction in morbidity and mortality of pertussis, especially in infants and children; however, the incidence of pertussis has increased in several countries despite high vaccination coverage [@pone.0017797-Andrews1]--[@pone.0017797-Tanaka1]. Since the 1980s, a considerable genetic transition has been observed between *B. pertussis* vaccine strains and circulating clinical strains in many countries [@pone.0017797-Bottero1]--[@pone.0017797-vanAmersfoorth1]. Genetic variations have been found in the loci encoding the major *B. pertussis* virulence factors: pertussis toxin S1 subunit (*ptxA*), pertactin (*prn*) and fimbriae 3 (*fim3*). Among circulating *B. pertussis* strains, vaccine-type alleles (*ptxA2*, *prn1* and *fim3A*) have been replaced mainly with nonvaccine-type alleles (*ptxA1*, *prn2* and *fim3B*). It has been speculated that adaptation of the bacterial population to vaccine-induced immunity has produced this genetic shift, and is one possible explanation for the resurgence of pertussis [@pone.0017797-Borisova1]--[@pone.0017797-Mooi1]. However, there have been few reports of the exact mechanism underlying this phenomenon. *B. pertussis* expresses various virulence factors, including adhesins and toxins, which function to establish and maintain host infection. Several virulence factors such as filamentous haemagglutinin (FHA) and pertussis toxin (PT) are expressed under the control of the BvgAS two-component regulatory system [@pone.0017797-Mattoo1], [@pone.0017797-Cummings1], [@pone.0017797-Shrivastava1]. The BvgAS system also positively regulates virulence factor secretion via the type III secretion system (T3SS) [@pone.0017797-Mattoo2], [@pone.0017797-Yuk1]. T3SS is highly conserved among a number of Gram-negative bacteria and functions as an injector of virulence molecules (i.e., effectors) into the host cell through a needle-like injection apparatus [@pone.0017797-Coburn1], [@pone.0017797-Ghosh1]. In *B. pertussis*, T3SS plays a role in subverting the protective innate and adaptive immunity of the host. Three T3SS-secreted proteins, BopN, BopD and Bsp22, have been identified so far [@pone.0017797-Fennelly1]. In the animal pathogen *Bordetella bronchiseptica*, BopN is involved in the up-regulation of cytokine IL-10 [@pone.0017797-Nagamatsu1], while Bsp22 polymerizes to form a flexible filamentous structure at the tip of the needle structure and associates with the pore component BopD [@pone.0017797-Medhekar1]. The Bsp22 translocon is expressed in a significant proportion of *B. pertussis* clinical isolates but not in Tohama and Wellcome 28, the common laboratory-adapted vaccine strains [@pone.0017797-Fennelly1]. Genomic differences between *B. pertussis* clinical strains and the vaccine strain Tohama have been investigated. The comparative genomics profiling revealed that the genome of *B. pertussis* Tohama differs from clinical isolates in four regions (RD11 to RD14) [@pone.0017797-Caro1]. In contrast, progressive gene loss mediated by homologous recombination between IS*481* insertion sequence elements has been observed among recently circulating strains of *B. pertussis* isolates [@pone.0017797-Heikkinen1], [@pone.0017797-King1]. IS*481* is present in multiple copies on the *B. pertussis* chromosome, and it plays a critical role in *B. pertussis* evolution through genomic rearrangement. Proteomic analysis has been widely applied to comparisons of protein expression among different strains, and information accumulated from genomic studies of *Bordetella* spp. facilitates comparative proteomic approaches to the investigation of *B. pertussis* clinical strains [@pone.0017797-Bottero1], [@pone.0017797-Vidakovics1]. In the present study, a proteomic approach was employed to identify the protein(s) involved in the genetic shift from vaccine-type to nonvaccine-type in *B. pertussis* strains. The protein profile analyses identified one differentially expressed protein, the T3SS effector BteA (alias BopC) [@pone.0017797-Panina1], [@pone.0017797-Kuwae1], between the strain types. BteA is a 68 kDa cytotoxic effector that has been identified in *B. bronchiseptica* but not in the *B. pertussis* human pathogen. Here we studied the differential expression of BteA protein in *B. pertussis* clinical strains and identified a specific IS*481* insertion in the 5′ untranslated region (5′-UTR) of *bteA* in vaccine-type strains. Results {#s2} ======= Identification of BteA in *B. pertussis* nonvaccine-type strain {#s2a} --------------------------------------------------------------- A comparative proteomic analysis of two clinical strains was performed to investigate the shift of *B. pertussis* strains from vaccine-type to nonvaccine-type. [Figure 1](#pone-0017797-g001){ref-type="fig"} shows 2-dimensional electrophoretic (2-DE) maps of total protein expressed in the nonvaccine-type clinical strain BP235 and the vaccine-type BP233. Among \>600 protein spots detected on the 2-DE gel, one was notably absent in the 2-DE map of BP233. The protein spot was observed in other nonvaccine-type strains (BP157, BP159, BP162 and BP228), but not in other vaccine-type strains (BP155, BP156, BP232 and BP243). The protein represented by the spot was identified by LC-MS/MS analysis using tryptic digests. The MS/MS of the protein digests provided four peptide sequences (RPDEFAAR, FDALR, ITALNLR and TQTQLLALQR) that matched the *B. pertussis* hypothetical protein BP0500 (NCBI accession: NP\_879352). Hypothetical protein BP0500 was identified as the T3SS effector BteA, since the sequence is highly conserved with 98% amino acid identity to the BteA (BopC) of *B. bronchiseptica* [@pone.0017797-Panina1], [@pone.0017797-Kuwae1]. ::: {#pone-0017797-g001 .fig} 10.1371/journal.pone.0017797.g001 Figure 1 ::: {.caption} ###### Comparative proteomic analysis of *B. pertussis* nonvaccine-type and vaccine-type strains. Total protein (10 µg) from the nonvaccine-type and vaccine-type clinical strains was separated by 2-D gel electrophoresis and silver stained. The left upper panel shows the protein profile of the nonvaccine-type BP235. The right upper panel shows the protein profile of the vaccine-type BP233. The red-boxed areas are enlarged (lower panels). The arrow in left lower panel indicates the spot that was identified as type III effector BteA by LC-MS/MS analysis. :::  ::: High expression of BteA protein in nonvaccine-type strains {#s2b} ---------------------------------------------------------- Immunoblots of *B. pertussis* clinical strains using anti-BteA antiserum detected high levels of a protein of ∼68 kDa in all nonvaccine-type clinical strains (BP157, BP159, BP162, BP228 and BP235), whereas BteA expression was greatly reduced in the vaccine-type clinical strains (BP155, BP156, BP232, BP233 and BP243). Additional products of \>200 kDa were also detected in the nonvaccine-type clinical strains. These high molecular mass signals appear to be the protein bands that have been reported as a multimeric complex of BteA in *B. bronchiseptica* [@pone.0017797-Panina1], [@pone.0017797-Kuwae1] (see [Figure S1](#pone.0017797.s001){ref-type="supplementary-material"}). T3SS function in the nonvaccine-type strains was confirmed by using whole cell protein extracts for immunoblots of BtcA (the BteA chaperone) [@pone.0017797-Panina1], [@pone.0017797-French1] and BopD (the T3SS translocon) [@pone.0017797-Nogawa1]. BtcA and BopD polypeptides were detected in both strain types, but the BtcA signals produced by the nonvaccine-type strains were apparently lower than those of the vaccine-type strains ([Figure 2](#pone-0017797-g002){ref-type="fig"}). The reason for the different expression is not clear. In contrast, adenylate cyclase toxin (ACT), another *Bordetella* spp. virulence factor, was detected at similar levels in both strain types. ::: {#pone-0017797-g002 .fig} 10.1371/journal.pone.0017797.g002 Figure 2 ::: {.caption} ###### Expression of BteA, BtcA, BopD and ACT proteins in *B. pertussis* nonvaccine-type and vaccine-type strains. The nonvaccine-type clinical strains (BP157, BP159, BP162, BP228 and BP235) and vaccine-type clinical strains (BP155, BP156, BP232, BP233 and BP243) were cultured in modified SS medium for 18 h. Total protein extracted from bacterial cells was subjected to immunoblot analysis with anti-BteA, anti-BtcA, anti-BopD or anti-ACT antiserum. For BteA detection, 10 µg of total protein was loaded in each lane. :::  ::: In order to confirm BteA secretion by *B. pertussis* strains, BteA polypeptide in the culture supernatants (CS) was subjected to immunoblot analysis. BteA was detected in secreted proteins from the nonvaccine-type clinical strain BP159 at 12, 24 and 48 h, whereas the signal was very low in the vaccine-type clinical strain BP155 over the 48-h time period ([Figure 3](#pone-0017797-g003){ref-type="fig"}). Conversely, signals corresponding to PT-S1 subunit and FHA polypeptides were detected in the supernatants of both cultures throughout the sampling period, although silver staining revealed small differences in their protein profiles after 24 h in culture. ::: {#pone-0017797-g003 .fig} 10.1371/journal.pone.0017797.g003 Figure 3 ::: {.caption} ###### BteA secretion from *B. pertussis* nonvaccine-type and vaccine type strains. Strains BP235 (nonvaccine-type) and BP233 (vaccine-type) were cultured in modified SS medium, and the culture supernatants (CS) were collected at 12, 24 and 48 h. Protein samples prepared by precipitation with 10% trichloroacetic acid were separated by 12.5% SDS-PAGE followed by silver staining (left panel). BteA, FHA and PT secretions were analyzed by immunoblots using anti-BteA, anti-FHA or anti-PT antiserum (right panels). For BteA detection, the equivalent of 200 µl of CS was loaded in each lane. :::  ::: Transcription of *bteA* {#s2c} ----------------------- *bteA* gene expression in *B. pertussis* strains was investigated with conventional RT-PCR and quantitative RT-PCR. *bteA* was transcribed in both the nonvaccine-type (BP157, BP159, BP162, BP228 and BP235) and vaccine-type (BP155, BP156, BP232, BP233 and BP243) clinical strains ([Figure 4A](#pone-0017797-g004){ref-type="fig"}). Similarly, *btcA* transcripts were detected in both strain groups. RT-PCR experiments lacking reverse transcriptase showed no specific product for *bteA* amplification, confirming negligible genomic DNA contamination in the RNA preparations. Quantitative RT-PCR (qRT-PCR) showed an average *bteA* transcript level of 0.146 (±1SD range, 0.107 to 0.184) in nonvaccine-type strains and 0.095 (±1SD range, 0.076 to 0.113) in vaccine-type clinical strains, a difference that was not statistically significant (*P* = 0.11) ([Figure 4B](#pone-0017797-g004){ref-type="fig"}). ::: {#pone-0017797-g004 .fig} 10.1371/journal.pone.0017797.g004 Figure 4 ::: {.caption} ###### RT-PCR analysis of *bteA* transcript in *B. pertussis* nonvaccine-type and vaccine-type strains. \(A) RT-PCR with primers specific for *bteA* and *btcA*. cDNA made from total RNA of nonvaccine-type (BP157, BP159, BP162, BP228 and BP235) and vaccine-type (BP155, BP156, BP232, BP233 and BP243) clinical strains was used as templates for PCR. Genomic DNA (G) from *B. pertussis* strain Tohama was used as a positive control. A mock reaction for *bteA* (−RT) consisted of reactions lacking reverse transcriptase. (B) Quantitative RT-PCR analysis of *bteA* transcript levels in the nonvaccine-type and vaccine-type clinical strains listed in (A). The *recA* transcript was used as a reference. Each point represents one strain and vertical bars indicate standard deviations. :::  ::: IS*481* insertion in the *bteA* 5′-UTR in vaccine-type strains {#s2d} -------------------------------------------------------------- Sequencing of the *bteA* 5′-UTR of the five vaccine-type strains (BP155, BP156, BP232, BP233 and BP243), revealed a 1,043-bp insertion sequence (IS*481*) −136 bp upstream of the *bteA* start codon ([Figure 5A](#pone-0017797-g005){ref-type="fig"}). IS*481a*, which is newly identified in *B. pertussis*, showed 99% nucleotide sequence identity with IS*481* of *B. pertussis* Tohama. The CCTAAC sequence in the *bteA* 5′-UTR is an insertion site of IS*481a* and is duplicated by the insertion, although the 6-bp consensus recognition sequence of IS*481* has been reported as NCTAGN [@pone.0017797-Stibitz1]. IS*481* insertions were not found in the nonvaccine-type clinical strains, which had nucleotide sequences that were 99% identical to that of *B. pertussis* Tohama. In the *bteA* 5′-UTR of the nonvaccine-type strains (BP157, BP159, BP162 BP228 and BP235), one single nucleotide polymorphism (A→G) was observed at 207 bp upstream of the *bteA* translation start site ([Figure 5B](#pone-0017797-g005){ref-type="fig"}). ::: {#pone-0017797-g005 .fig} 10.1371/journal.pone.0017797.g005 Figure 5 ::: {.caption} ###### Physical maps of the *btcA−bteA* region of *B. pertussis* vaccine-type and nonvaccine-type strains. \(A) The vaccine-type clinical strain BP155. The location of IS*481a* is represented by a gray arrow on the physical map. The recognition sequence of IS*481a* is underlined. The two mapped transcriptional start sites (P1 and P2) of *bteA* are shown by arrows. Region amplified by qRT-PCR to determine the IS*481a*-promoter (P2) and total (P1 + P2) transcripts are shown by two-headed arrows below the physical map. (B) The nonvaccine-type clinical strain BP159. The mapped transcriptional start sites of *bteA* (P1) and *btcA* \[P (*btcA*)\] are shown by arrows. The single nucleotide polymorphism (A→G) at −207 bp from the *bteA* translation start codon is indicated by an asterisk. :::  ::: The *bteA* 5′-UTR was PCR-amplified from chromosomal DNA of other *B. pertussis* strains to confirm insertion of IS*481*. Among 61 vaccine-type clinical strains, 60 (98%) produced amplicons of ∼3.1 kb, a size indicative of an IS*481* insertion in the *bteA* 5′-UTR. One strain (BP121) had a product of ∼2.1 kb, corresponding to the predicted size of the native 5′-UTR (data not shown). Of the 23 nonvaccine-type strains examined, all generated ∼2.1 kb amplicons, confirming the absence of the IS*481* insertion. Determination of the *bteA* transcription start site {#s2e} ---------------------------------------------------- 5′-RACE mapping was used to identify the *bteA* transcription start site in vaccine-type strain BP155. Nucleotide sequences of the 5′-RACE PCR products revealed two transcription start sites, P1 and P2, located −68 and −147 bp from the *bteA* translation start codon ([Figure 5A](#pone-0017797-g005){ref-type="fig"}). The P1 start site (+1) was located within the *bteA* 5′-UTR, whereas the P2 start site (−79) was located within IS*481a*. Only the P1 start site was also found in the nonvaccine-type strain BP159 ([Figure 5B](#pone-0017797-g005){ref-type="fig"}). IS*481* contains an outward-facing promoter at one end that is responsible for transcription of the flanking catalase gene (*katA*) in *B. pertussis* [@pone.0017797-DeShazer1]. However, the P2 start site is different from the *katA* transcription start site. The transcription start site of *btcA*, also determined by 5′-RACE, was mapped to a T residue 31 bp upstream of the *btcA* translation start codon in both the vaccine-type and nonvaccine-type strains ([Figure 5B](#pone-0017797-g005){ref-type="fig"}). Primer extension analysis was also performed in an attempt to resolve the *bteA* transcription start sites. However, the start sites could not be ascertained, probably due to low amounts of *bteA* transcript in *B. pertussis*. IS*481a*-promoter transcript is the major *bteA* transcript in the vaccine-type strain {#s2f} -------------------------------------------------------------------------------------- Expression of the IS*481a*-promoter transcript (P2 transcript) in *B. pertussis* vaccine-type strain BP155 was analyzed by qRT-PCR with TaqMan probes ([Figure 5A](#pone-0017797-g005){ref-type="fig"}). The P2 transcript and total *bteA* (P1 + P2) transcripts were determined individually and the ratio of P2 transcript to total *bteA* transcript was calculated. Based on four independent experiments, the ratio (P2 transcript/P1 + P2 transcripts) was estimated to be 0.88 (±1SD range, 0.70 to 1.09), indicating that the P2 transcript is the major *bteA* transcript in the vaccine-type strain (data not shown). BteA expression in *B. pertussis* BteA mutants {#s2g} ---------------------------------------------- To clarify the effect of the IS*481* insertion on BteA expression, four BteA mutants (Δ*bteA*-BP155, ΔIS*481*-BP155, Δ*bteA*-BP157 and +IS*481*-BP157) were constructed from *B. pertussis* BP155 (vaccine-type) and BP157 (nonvaccine-type) by homologous recombination ([Figure 6A](#pone-0017797-g006){ref-type="fig"}). The Δ*bteA*-BP155 and Δ*bteA*-BP157 mutants had a 178-bp deletion in the 5′ region of *bteA*. In the ΔIS*481*-BP155 mutant, a 2.2-kb insertion containing an intact *bteA* 5′-UTR (derived from *B. pertussis* Tohama) replaced the native *bteA* 5′-UTR + IS*481a* gene. In contrast, +IS*481*-BP157 mutant had a 3.2-kb insertion containing a *bteA* 5′-UTR + IS*481a* (derived from *B. pertussis* BP155) instead of its own *bteA* 5′-UTR. Consequently, ΔIS*481*-BP155 had an IS*481a* deletion from the *bteA* 5′-UTR, whereas the +IS*481*-BP157 mutant had an IS*481a* insertion in the *bteA* 5′-UTR. The *btcA−bteA* region of the mutants was verified by DNA sequence analysis. ::: {#pone-0017797-g006 .fig} 10.1371/journal.pone.0017797.g006 Figure 6 ::: {.caption} ###### Construction and characterization of *B. pertussis* BteA mutants. \(A) Physical map of the *btcA−bteA* region of BteA mutants derived from *B. pertussis* BP155 (vaccine-type) and BP157 (nonvaccine-type). WT, wild-type; Δ*bteA*, a 178-bp deletion around the 5′ region of *bteA*; ΔIS*481*, IS*481a* deletion from the *bteA* 5′-UTR; +IS*481*, IS*481a* insertion in the *bteA* 5′-UTR. (B) Expression of BteA protein in the BteA mutants. The mutants were cultured in modified SS medium for 24 h. Total protein from the bacterial cells (Cell) and culture supernatants (CS) was analyzed with immunoblot using anti-BteA antiserum. :::  ::: BteA expression in the bacterial cells and CS after 24 h in culture was analyzed by immunoblot with anti-BteA antiserum ([Figure 6B](#pone-0017797-g006){ref-type="fig"}). In ΔIS*481*-BP155 bacterial cells and CS, BteA polypeptide(s) corresponding to ∼68 kDa and \>200 kDa were detected at the same level as was observed in the BP157 wild-type strain. In contrast, the signals of BteA polypeptide(s) from +IS*481*-BP157 mutant were very low in both bacterial cells and CS. Similarly, BteA polypeptide(s) were not detected in either Δ*bteA*-BP155 or Δ*bteA*-BP157. These results clearly showed that BteA protein expression is down-regulated by the IS*481* insertion in *B. pertussis*, and that the anti-BteA antiserum is highly specific to BteA. Discussion {#s3} ========== The BteA effector (alias BopC) is required for the induction of necrotic cell death during *B. bronchiseptica* infections, and is thought to play a pivotal role in T3SS-mediated cell death [@pone.0017797-Panina1], [@pone.0017797-Kuwae1], [@pone.0017797-Kozak1]. BteA is also involved in dephosphorylation of tyrosine-phosphorylated proteins (PY) of host cells [@pone.0017797-Kuwae1], and its 130-amino acid N-terminal sequence is associated with target lipid rafts [@pone.0017797-French1]. BteA is the only cytotoxic effector that has been identified in *Bordetella* spp. In *B. pertussis*, low-passage clinical strains have an ability to express a functionally active T3SS; however, BteA protein had not been detected in the clinical and common laboratory-adapted strains by MALDI-TOF mass spectrometry [@pone.0017797-Fennelly1]. Here we demonstrate that BteA protein is highly expressed in *B. pertussis* nonvaccine-type strains but not in the vaccine-type strains, and that BteA protein expression is down-regulated by IS*481a* insertion in the vaccine-type strains. We provide the first evidence that BteA protein expression is type-dependent due to the IS*481a* insertion in *B. pertussis* clinical strains. In Japan, *B. pertussis* circulating strains began to change from vaccine-type to nonvaccine-type in the mid-1990s [@pone.0017797-Kodama1], and the reported incidence of adult cases of pertussis has dramatically increased since 2002 [@pone.0017797-Han1]. The genetic divergence in *B. pertussis* circulating strains has also been observed in many other countries. A possible explanation for the genetic divergence is that the type shift is a result of vaccine-driven evolution [@pone.0017797-Borisova1]--[@pone.0017797-Mooi1]. More recently, Mooi et al. [@pone.0017797-Mooi2] reported that expansion of *B. pertussis* strains with increased PT production has contributed to the resurgence of pertussis in the Netherlands. Here we showed prominent expression of the T3SS effector protein BteA in the nonvaccine-type strains, and that PT and ACT (important virulence factors of *B. pertussis*) are expressed at the same level in both the nonvaccine and vaccine-type strains. Besides vaccine-driven evolution, our findings could provide another possible explanation for the type shift from vaccine-type to nonvaccine-type, i.e., the augmented expression of BteA protein in *B. pertussis* nonvaccine-type strains may be involved in the type shift. *B. bronchiseptica* BteA has *in vitro* cytotoxic activity against cultured mammalian cells [@pone.0017797-Mattoo2], [@pone.0017797-Fennelly1], [@pone.0017797-Panina1], [@pone.0017797-Kuwae1]. In this study, we determined the cytotoxicity of *B. pertussis* BteA mutants by measuring the release of lactate dehydrogenase (LDH) from L2 rat lung epithelial cells, J774 mouse macrophage-like cells, or HeLa cells. However, even BteA-expressing strains (ΔIS*481*-BP155 and wild-type BP157) showed low cytotoxicity (\<10%), and consequently no statistically significant differences in cytotoxicity were observed among the wild-type and mutant strains. *B. pertussis* is known to have a lower *in vitro* cytotoxicity than *B. bronchiseptica* [@pone.0017797-Mattoo2], [@pone.0017797-Fennelly1], which is consistent with the extremely low secretion of BteA in *B. pertussis* as compared to *B. bronchiseptica* ([Figure S1](#pone.0017797.s001){ref-type="supplementary-material"}). Therefore, a more sensitive and quantitative assay is required to determine the BteA-dependent cytotoxicity of *B. pertussis*. IS*481* belongs to the recently defined IS*481* family [@pone.0017797-Craig1], and 238 copies of IS*481* are present in the *B. pertussis* Tohama genome [@pone.0017797-Parkhill1]. In *B. pertussis* clinical strains, IS*481* is also present in multiple copies on the chromosome and it plays a critical role in *B. pertussis* evolution [@pone.0017797-Heikkinen1], [@pone.0017797-Caro2]. Many IS elements have been shown to activate the expression of neighboring genes. IS*481* contains an outward-facing promoter that is located in close proximity to the left terminal inverted repeat, and this promoter is responsible for the transcription of *katA* in certain *B. pertussis* strains [@pone.0017797-DeShazer1]. Here we identified an IS*481a* insertion in the *bteA* 5′-UTR in *B. pertussis* vaccine-type clinical strains and detected a high level of *bteA* transcripts from the IS*481a* promoter (P2) compared with its own promoter (P1). However, the vaccine-type strains showed a low level of BteA protein expression, suggesting that insertion of IS*481a* represses P1 promoter activity, and that P2 transcript has a low translational efficiency from the additional nucleotide sequence (79 nucleotides) at its 5′ end. Use of a cell-free coupled transcription-translation system revealed that the additional nucleotide sequence is involved in down-regulation of transcription and/or translation ([Figure S2](#pone.0017797.s002){ref-type="supplementary-material"}). The 5′-UTR of bacterial mRNAs can bear regulatory elements that are involved in down- or up-regulation of translation [@pone.0017797-Kaberdin1]. The regulatory mechanisms in this region are controlled by RNA-binding proteins, small noncoding RNAs and structural rearrangements with the 5′-UTR. In addition, a 5′ stem-loop structure that sequesters the ribosomal binding site has been shown to be involved in translational regulation. Bioinformatic analysis uncovered a predicted stem-loop structure in the *bteA* 5′-UTR of P2 transcript ([Figure S2](#pone.0017797.s002){ref-type="supplementary-material"}). In this study, the 5′-UTRs of five *B. pertussis* vaccine-type clinical strains were sequenced; all had an insertion of an IS*481a* in the *bteA* 5′-UTR, both transcribed in the same direction. In one of the vaccine-type strains, BP155, the major *bteA* mRNA was transcribed from P2 in the IS*481a*-promoter. These observations raise the possibilities that (i) the P2 transcript is translated into BteA under certain environmental conditions, and (ii) the P2 transcript is translated into another novel protein by translational frameshifting. BteA is known to be regulated by the BvgAS system and an extracytoplasmic function (ECF) sigma factor BtrS in *B. bronchiseptica* [@pone.0017797-Mattoo2], [@pone.0017797-Panina1]. In *B. pertussis*, it has been suggested that expression of the T3SS translocon Bsp22 is blocked by post-transcriptional regulation [@pone.0017797-Mattoo2]. However, the molecular details of the regulatory mechanism are still unclear. Further studies are needed to determine the down-regulation of BteA protein in *B. pertussis* vaccine-type clinical strains. In conclusion, *B. pertussis* vaccine-type strains have been replaced with the nonvaccine-type strains in many countries, and the resurgence of pertussis has been observed in several nations. In Japanese *B. pertussis* clinical strains, the T3SS effector BteA is highly expressed in nonvaccine-type strains as compared with the vaccine-type strains. Our findings indicate that augmented expression of BteA protein in *B. pertussis* circulating strains could play a key role in the type shift. However, it is unclear whether BteA protein is implicated in the resurgence of pertussis. Further studies are needed to determine the expression of BteA protein in *B. pertussis* circulating strains on a global scale. Materials and Methods {#s4} ===================== Bacterial strains and growth conditions {#s4a} --------------------------------------- *B. pertussis* clinical strains were selected from the laboratory collection of the National Institute of Infectious Diseases, Tokyo, Japan. The selection criteria included the time and geographic location of isolation, and their *ptxA* and *prn* alleles. A total of 10 clinical strains from 2002 to 2004 in Japan were included. Of the 10 clinical strains, 5 harbored *ptxA1* and *prn2* alleles (BP157, BP159, BP162, BP228 and BP235; nonvaccine-type strains), while the others carried *ptxA2* and *prn1* (BP155, BP156, BP232, BP233 and BP243; vaccine-type strains). All strains were cultured on Bordet-Gengou agar (BG agar, Difco) supplemented with 1% glycerol and 15% defibrinated horse blood or in modified Stainer-Scholte (SS) medium [@pone.0017797-Pradel1]. Two-dimensional gel electrophoresis (2D-PAGE) {#s4b} --------------------------------------------- 2D-PAGE was performed based on the O\'Farrell method [@pone.0017797-OFarrell1] with minor modifications. *B. pertussis* clinical strains grown on BG agar plates were resuspended in casamino acid solution (1% casamino acid, 0.6% NaCl, pH 7.1). Bacterial cells were precipitated by centrifugation (12,000 × *g*, 10 min) and resuspended in SDS-lysis buffer (62.5 mM Tris-HCl, 1% SDS, 10% glycerol, 5% 2-mercaptoethanol, pH 6.8) by sonication. Total protein was extracted by boiling for 3 min, followed by centrifugation. A portion (10 µg, approximately 2 µl) of the protein solution was mixed with 20 µl of sample buffer \[8.5 M urea, 2% Nonidet P-40, 2% Ampholine (pH 3.5 to 10)\], and applied to an isoelectric focusing tube gel (2.0 mm inside diameter by 12.0 cm) containing 4% polyacrylamide, 8.5 M urea, 2% Nonidet P-40, and 2% Ampholine (pH 5 to 7 and pH 3.5 to 10 in a ratio of 1∶4). Proteins were focused at 10°C for 17 h (1 h at 200 V, 2 h at 400 V, and 14 h at 800 V) with 10 mM H~3~PO~4~ (anolyte) and 20 mM NaOH (catholyte). In the second dimension, the electrofocused tube gel was electrophoresed in 12% SDS-PAGE. The separated polypeptides were visualized by silver staining and analyzed with the PDQuest 2-D Analysis Software (Bio-Rad, Hercules, CA). The Lowry assay was used to measure protein concentrations in a trichloroacetic acid (TCA) pellet (resuspended in 1 N NaOH) using bovine serum albumin as a standard. Protein identification {#s4c} ---------------------- 2D-PAGE gels were stained with silver nitrate without glutaraldehyde fixation [@pone.0017797-Shevchenko1], and protein spots of interest were excised. Proteins were reduced with 10 mM DTT, alkylated with 55 mM iodoacetamide, and digested with sequencing grade-modified trypsin (Promega, Madison, WI). Digested peptides were separated on a C18 capillary column (0.2 by 50 mm, Michrom Bioresources, CA) equipped with a Chorus 220 solvent delivery system and an HTC PAL auto-sampler system (CTC Analytics AG, Zwingen, Switzerland). Separated peptides were analyzed by the Finnigan LCQ^TM^ Deca XP ion trap mass spectrometer (Thermo Fisher Scientific Inc., MA) with electrospray ionization (ESI) interface using the Nanosprayer FS (GL Sciences Inc., Japan). To identify peptides, data files were generated from the MS/MS scans by Bioworks 3.0 using the SEQUEST algorithm (threshold, 10^5^; minimum group scan 2, Xc \>1.0, Thermo Fisher Scientific) and searched against the complete amino acid database derived from the *B. pertussis* Tohama genome database. Antibody production against recombinant BteA, BtcA and ACT {#s4d} ---------------------------------------------------------- The BteA gene (NCBI accession: NP\_879352) was amplified by PCR from *B. pertussis* Tohama DNA using BteA-F and BteA-R primers, and cloned into the XmnI/HindIII sites of pMal-c2X (New England Biolabs, Beverly, MA) to generate a maltose binding protein (MBP) fusion with BteA. Production of this fusion protein was induced in *E. coli* BL21 with 0.5 mM isopropyl-β-D-thiogalactopyranoside (IPTG) and subsequently purified using amylose resin (New England Biolabs) and Resource Q (Amersham Pharmacia Biotech, Uppsala, Sweden) columns. A two-step PCR was carried out to amplify recombinant BtcA (NCBI accession: NP\_879351). The first PCR was performed using the BtcA-BteA-F3 and BtcA-BteA-R3 primers ([Table S1](#pone.0017797.s003){ref-type="supplementary-material"}), which amplified the region between positions 165122 and 167190 of the *B. pertussis* Tohama genome (GenBank accession: BX640412). In the second PCR, *btcA* was amplified from the first PCR product with the 5-BtcA and 3-BtcA primers ([Table S1](#pone.0017797.s003){ref-type="supplementary-material"}) and cloned into the NdeI/HindIII sites of pCold II DNA (TAKARA Bio Inc.). His-tagged BtcA was induced in *E. coli* BL21 with 0.5 mM IPTG at 15°C and purified using the HisTrap FF Crude Kit (GE Healthcare UK Ltd.). A recombinant catalytic domain of *B. pertussis* adenylate cyclase toxin (ACT) was a gift from Mineo Watanabe. Antibodies against MBP-BteA, BtcA and ACT were generated in mice at Nippon Biotest Laboratories, Inc. (Tokyo, Japan). The MBP-BteA antiserum was pre-absorbed with MBP2 protein (New England BioLabs) and the resulting antiserum was used. Immunoblot analysis {#s4e} ------------------- *B. pertussis* clinical strains were inoculated in modified SS medium with a starting optical density of 0.2 at 600 nm, and further cultured with shaking at 36°C. Total protein was extracted with SDS-lysis buffer, and culture supernatant (CS) proteins were precipitated with 10% TCA. Protein samples were subjected to SDS-PAGE, transferred to nitrocellulose membranes (Bio-Rad) and incubated with anti-BteA, anti-BtcA, anti-BopD [@pone.0017797-Nogawa1], anti-ACT, anti-FHA, or anti-PT antiserum. Antigen-antibody complexes were visualized using horseradish peroxidase (HRP)-conjugated secondary antibody (Bio-Rad, Hercules, CA) and ECL Western Blotting Detection Reagents (GE Healthcare). DNA sequencing {#s4f} -------------- The region between the *btcA* and *bteA* gene corresponding to positions 165122 to 168021 of *B. pertussis* Tohama (GenBank accession: BX640412) was amplified in vaccine-type and nonvaccine-type clinical strains with the appropriate primers and sequenced. Sequencing reactions were carried out with the BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems, Foster City, CA), and the products were sequenced on an ABI PRISM 3130*xl* Genetic Analyzer (Applied Biosystems). Transcriptional analyses {#s4g} ------------------------ Total RNA was isolated using the RNeasy Mini Kit (QIAGEN, Hilden, Germany) and treated with RNase-free DNase (Promega) to degrade contaminating DNA. Reverse transcriptase-PCR (RT-PCR) was performed with bteA RT-R and btcA RT-R primers ([Table S1](#pone.0017797.s003){ref-type="supplementary-material"}) using the TAKARA One Step RNA PCR Kit (AMV, TAKARA Bio Inc.). PCR was performed with the following conditions: one cycle of 50°C for 30 min, 95°C for 2 min; 25 cycles of 95°C for 30 s, 58°C for 30 s, 72°C for 1 min; and a final incubation at 72°C for 10 min. Primer sets, bteA RT-F/bteA RT-R and btcA RT-F/btcA RT-R, were used for *bteA* and *btcA* amplification, respectively ([Table S1](#pone.0017797.s003){ref-type="supplementary-material"}). Products were analyzed by electrophoresis on a 1.5% agarose gel. Reverse transcriptase was omitted from the negative control reaction mixtures. For quantitative RT-PCR (qRT-PCR), 5 µg of RNA was reverse transcribed into cDNA using the SuperScript First-Strand Synthesis System (Invitrogen, Carlsbad, CA) with random hexamer primers. Relative levels of total *bteA* and *recA* transcripts were determined using TaqMan probes (bteA- and recA-probes, [Table S1](#pone.0017797.s003){ref-type="supplementary-material"}) and *Premix Ex Taq*TM (Perfect Real Time, TAKARA Bio Inc.) with the ABI PRISM 7500 Sequence Detection System (Applied Biosystems). The qRT-PCR conditions were 30 s at 95°C, followed by 40 cycles of 95°C for 15 s and 60°C for 1 min. The expression of *recA* was used as an internal control [@pone.0017797-Stefanelli1]. All samples were run in triplicate and *bteA* transcript (P1 + P2 transcripts) was normalized to the *recA* transcript for each sample. The *bteA* IS*481a*-promoter transcript (P2 transcript) was determined using a TaqMan probe (IS481-bteA probe). The qRT-PCR conditions were 30 s at 95°C, followed by 40 cycles of 95°C for 15 s and 55°C for 1 min. The ratio of P2 transcript to total *bteA* transcript (P2 transcript/P1 + P2 transcripts) was estimated from four independent experiments. The regions amplified by qRT-PCR are shown in [Figure 6A](#pone-0017797-g006){ref-type="fig"}. Mapping transcriptional start sites {#s4h} ----------------------------------- 5′ rapid amplification of cDNA ends (5′-RACE) was performed using 5′-Full RACE Core Set (TAKARA Bio Inc.) according to the manufacturer\'s instructions. Reverse transcription was executed at 55°C using a 5′ phosphorylated RT primer (bteA-RT, [Table S1](#pone.0017797.s003){ref-type="supplementary-material"}). The first PCR used primers bteA-S1 (S1) and bteA-A1 (A1) primers, and bteA-S2 (S2) and bteA-A2 (A2) for the second ([Table S1](#pone.0017797.s003){ref-type="supplementary-material"}). PCR products were cloned into the pT7Blue T-vector (Novagen, Madison, Wis.) and transformed into *E. coli* XL1-Blue, which were plated on LB agar plates. Several clones were sequenced. The transcription start site of *btcA* was located using 5′-RACE with five primers, btcA-RT (5′ phosphorylated primer), btcA-S1 (S1), btcA-A1 (A1), btcA-S2 (S2) and btcA-A2 (A2) ([Table S1](#pone.0017797.s003){ref-type="supplementary-material"}). Generation of BteA mutants {#s4i} -------------------------- Four BteA mutants, Δ*bteA*-BP155, Δ*bteA*-BP157, ΔIS*481*-BP155 and +IS1*481*-BP157, were constructed by homologous recombination as described previously with minor modifications [@pone.0017797-Kuwae1] ([Figure 6A](#pone-0017797-g006){ref-type="fig"}). *BteA-deficient mutants*: A 2.2-kbp DNA fragment containing a 5′ portion of the *bteA* gene was amplified by PCR with the B1-bteA and B2-bteA primers ([Table S1](#pone.0017797.s003){ref-type="supplementary-material"}) using the *B. pertussis* Tohama genomic DNA as the template. The PCR product was cloned into the pDONR221 vector (Invitrogen) to obtain pDONR-*bteA* by means of adaptor PCR and site-specific recombination techniques with the Gateway Cloning System (Invitrogen). Inverse PCR was then carried out with R1-bteA and R2-bteA primers ([Table S1](#pone.0017797.s003){ref-type="supplementary-material"}) using circular pDONR-*bteA* as the template. The R1-bteA and R2-bteA primers contained a BamHI site. The resulting PCR product was digested with BamHI and self-ligated to obtain pDONR-Δ*bteA*, which contained a 178-bp deletion around the 5′ region of *bteA*. pDONR-Δ*bteA* was mixed with pABB-CRS2 [@pone.0017797-Sekiya1] to obtain pABB-Δ*bteA* using the Gateway Cloning System. pABB-Δ*bteA* was then introduced into *E. coli* SM10λ*pir* and transconjugated into streptomycin (SM)-resistant *B. pertussis* BP155 (vaccine-type) and BP157 (nonvaccine-type) clinical strains. The resultant mutant strains were designated Δ*bteA*-BP155 and Δ*bteA*-BP157. *IS*481*-deletion mutant*: pABB-*bteA* was constructed from pDONR-*bteA*. pABB-*bteA* was introduced into *E. coli* SM10λ*pir* and transconjugated into SM-resistant *B. pertussis* vaccine-type BP155. The resultant mutant strain was designated ΔIS*481*-BP155. *IS*481*-insertion mutant*: a 3.2-kbp DNA fragment (*bteA*+IS*481*) containing the *bteA* 5′-UTR and IS*481a* was amplified with the B1-bteA and B2-bteA primers ([Table S1](#pone.0017797.s003){ref-type="supplementary-material"}) using *B. pertussis* BP155 genomic DNA as the template. pABB-*bteA*+IS*481* was constructed from pDONR-*bteA*+IS*481* and transconjugated into SM-resistant *B. pertussis* nonvaccine-type BP157 via *E. coli* SM10λ*pir*. The resultant mutant strain was designated +IS*481*-BP157. Statistical analysis {#s4j} -------------------- The Student\'s *t*-test was employed. A value of *P*\<0.05 was considered statistically significant. Nucleotide sequence accession number {#s4k} ------------------------------------ The IS*481a* sequence was deposited in the DDBJ/EMBL/GenBank nucleotide sequence databases under accession number AB473880. Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### **High secretion of BteA protein in** ***Bordetella bronchiseptica*** **.** *B. bronchiseptica* (BB R05), *B. pertussis* BP155 (vaccine-type) and BP157 (nonvaccine-type) were cultured in modified SS medium for 24 h. Total protein extracted from the bacterial cells (Cell) and culture supernatants (CS) was separated by SDS-PAGE followed by silver staining (left panel). Immunoblots were incubated with anti-BteA, anti-BtcA or anti-BopD antiserum (right panel). For BteA detection, 0.5 µg of total protein (for Cell) and 5 µl of CS were loaded in the indicated lanes. The amount of total protein loaded was one-twentieth of that in [Figure 2](#pone-0017797-g002){ref-type="fig"}, and the loaded CS volume was one-fortieth of that in [Figure 3](#pone-0017797-g003){ref-type="fig"}. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Figure S2 ::: {.caption} ###### ***In vitro*** **transcription-translation analysis of a** ***bteA*** **5′-UTR deletion series.** (A) *bteA* 5′-UTR deletion genes were PCR-amplified using *B. pertussis* BP155 (vaccine-type) as the template. Proteins were synthesized using the WakoPURE System (Wako Pure Chemical Industries, Ltd.). The 5′-UTR deletion genes harbored the T7 promoter at their 5′ end. (B) Expression of BteA protein in an *in vitro* transcription-translation system (WakoPURE System). The synthesized product was analyzed with immunoblots using anti-BteA antiserum. NC, negative control. (C) A predicted stem-loop structure in the 5′-UTR of *bteA* mRNA (P2 transcript). The RNA secondary structure was analyzed by CentroidFold (<http://www.ncrna.org/centroidfold>). The schematic shows a simplified map. TIR, translation initiation region. (TIF) ::: ::: {.caption} ###### Click here for additional data file. ::: Table S1 ::: {.caption} ###### **Primers and probes in this study.** (XLS) ::: ::: {.caption} ###### Click here for additional data file. ::: We would like to thank Yuko Sasaki for her assistance with the LC-MS/MS analysis and Jun-ichi Wachino for his technical advice in 5′-RACE mapping. We also thank Mineo Watanabe (Kitasato University) for his kind gift of recombinant ACT protein. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**This research was supported by grants for Research on Emerging and Re-emerging Infectious Diseases (09158691 and 09158699) from the Ministry of Health, Labor and Welfare of Japan. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: H-JH KK. Performed the experiments: H-JH KK. Analyzed the data: H-JH KK. Contributed reagents/materials/analysis tools: AK AA YA. Wrote the paper: KK. [^2]: ¤ Current address: National Fisheries Research and Development Institute, Busan, Republic of Korea

PubMed Central

2024-06-05T04:04:19.895183

2011-3-10

PMC3053400

Introduction {#s1} ============ The foundations of decision research, and hence its contemporary shape, have been strongly influenced by thinking from disciplines like economics. Human investors adjust their decisions according to partners on the basis of expected pay-offs. They are supposed to make rational decisions and to revise their decisions in order to optimize satisfaction [@pone.0017801-Nofsinger1]. Animals can also maximize pay-offs. When individuals exploit an environment where resources are distributed in patches, they can leave the patch and search for a new one when the rate of pay-off falls below the average rate for the entire area [@pone.0017801-Charnov1], [@pone.0017801-Stephens1]. Rational strategies are then defined as those increasing fitness and are an outcome of natural selection [@pone.0017801-McNamara1]. The theory of biological markets in particular assumes that living beings can adjust their investment based on the offers potentially provided by several partners [@pone.0017801-No1]. In non-human primates, individuals may vary their rates of grooming in exchange to access for commodities [@pone.0017801-Barrett1], [@pone.0017801-Fruteau1]. They are able to invest, that is, to avoid immediately consuming some goods with the intent of winning more [@pone.0017801-Drapier1], [@pone.0017801-Lefebvre1]. We lack evidence, however, about their abilities to adjust quantitatively their investment to expected pay-offs. Monkeys and great apes appear to possess many of the skills required to perform successful investments in various contexts. They can make inferences, categorize objects and understand tertiary relations [@pone.0017801-Tomasello1]. They are also able to make 'more' and 'less' value judgments about discrete quantities [@pone.0017801-Anderson1]--[@pone.0017801-Boysen1]. Numerous studies showed that monkeys are good at recognizing magnitudes for values under 8. For instance, rhesus macaques reliably prefer the larger amount in choices of one versus two items, two versus three, and three versus four [@pone.0017801-Hauser1], [@pone.0017801-Wood1]. Monkeys can also discriminate between larger numerical values when high ratios are involved [@pone.0017801-Hauser2]. Rhesus macaques can learn to select the stimulus with the larger number of dots when pairs of numerical values between 1 and 9 are presented [@pone.0017801-Brannon1], [@pone.0017801-Cantlon1]. Similar results are found in squirrel monkeys and tufted capuchin monkeys with discrimination between discrete quantities of one to nine food items [@pone.0017801-Addessi1]--[@pone.0017801-vanMarle1]. Non-human primates are also able to combine discrete quantities, which can allow them to adjust their investment quantitatively. When presented with two trays, each tray containing two separate sets of food items, chimpanzees and capuchin monkeys select the greater total, indicating that they consider the sum of items [@pone.0017801-Addessi2]--[@pone.0017801-Rumbaugh1]. Both great apes and monkeys succeed in tasks where they have to choose between two covered sets of food items to which an experimenter visibly adds or removes items in unequal numbers (capuchin monkeys: [@pone.0017801-Beran3]; chimpanzees: [@pone.0017801-Beran1], [@pone.0017801-Beran4], [@pone.0017801-Beran5]; orangutans: [@pone.0017801-Call1]; rhesus macaques: [@pone.0017801-Hauser1]). Monkeys can differentiate between different contingencies in discrimination learning task where they have to distinguish between two cues to gain rewards [@pone.0017801-Tomasello1]. They can also discriminate between experimenters who behave in different ways towards them. For instance, capuchins and macaques preferentially indicate a food location to the most cooperative partners [@pone.0017801-Mitchell1], and they recognize those providing the higher pay-off [@pone.0017801-Chen1]--[@pone.0017801-Brosnan1]. Monkeys can thus use potential partners as a tool to gain more. On the other hand, while monkeys may instrumentalize conspecifics, they may limit this to anticipating their behavior and not their intentions. It should be emphasized that in most experiments they appear unable to recognize actions in term of goals contrary to great apes [@pone.0017801-Povinelli1]--[@pone.0017801-Hare1] (but see [@pone.0017801-Flombaum1]). With regard to future-oriented behaviors, several experiments show that apes and monkeys can accept to lose an immediate benefit to gain more later; they postpone gratification from some seconds to a few minutes in tasks where they are given a choice between an immediately available but less preferred reward, and a delayed but more preferred one [@pone.0017801-Amici1]--[@pone.0017801-Stevens1]. They sustain similar delays of gratification when presented with food items accumulating at regular time intervals [@pone.0017801-Beran6], [@pone.0017801-Evans1]. Non-human primates also maximize their pay-offs in tests requiring them to exchange with an experimenter. Chimpanzees and capuchin monkeys can learn to attribute values to non-edible tokens and exchange them for food [@pone.0017801-Brosnan2]--[@pone.0017801-Westergaard1]. However, this set up implies training monkeys to understand the value of the tokens. Monkeys and great apes can also give food items to receive a qualitatively more desirable one [@pone.0017801-Drapier1], [@pone.0017801-Lefebvre1], [@pone.0017801-Lakshminaryanan1]. In that case, the value of the food is directly measured by consumption. In a study where capuchin monkeys were allowed to eat part of an item before returning it, individuals were seen to nibble most of a food item before attempting to exchange the remains for a larger reward with a human experimenter [@pone.0017801-Drapier1], [@pone.0017801-Ramseyer1]. Also, non-human primates can wait longer for a return if the expected quantity of food is larger [@pone.0017801-Ramseyer1], [@pone.0017801-Dufour1]. Decision-making in primates relies on skills requiring them to take into account several factors involving evaluation of discrete quantities, physical or temporal cost, and partner\'s reliability to maximize their pay-offs. In this study, we tested tufted capuchin monkeys, Tonkean macaques and long-tailed macaques in an exchange task where each subject initially received four food rewards that they could either consume or give back. To maximize the pay-off, subjects had to adapt the amount of food items they gave initially -- the investment -- to the food amounts to be returned by two different experimenters. We investigated whether the subjects could invest differentially depending on the experimenter\' qualities in term of income. One experimenter gave back a reward twice the amount of the subjects\' initial investment (doubling partner, providing 0, 2, 4, 6 or 8 rewards if subjects returned respectively 0, 1, 2, 3 or 4 rewards), whereas the other always gave back a constant amount regardless of the subjects\' initial investment (fixed partner, always providing 8 rewards regardless of the amount initially returned). To maximize food income, subjects had to respond in different ways to each experimenter, offering a maximal amount to the first one and a minimal amount to the second ([Table 1](#pone-0017801-t001){ref-type="table"}). ::: {#pone-0017801-t001 .table-wrap} 10.1371/journal.pone.0017801.t001 Table 1 ::: {.caption} ###### Number of rewards obtained from both experimenters and subjects\' net income according to the number of raisins returned by subjects. ::: {#pone-0017801-t001-1} Doubling partner Fixed partner --- ------------------ --------------- --- ---- 0 0 4 0 4 1 2 5 8 11 2 4 6 8 10 3 6 7 8 9 4 8 8 8 8 Within one session, the subjects\' net income, i.e. the amount of raisins non-invested by the subject plus those received after return. The subject maximises its gain by giving more (4 raisins, net income 8) to the doubling partner, and less to the fixed one (1 raisin, net income 11). ::: Results {#s2} ======= When giving less than four raisins to experimenters, subjects exhibited different ways to remove raisins from the initial amount. They either ate some and returned the remaining ones (Pis, Arn, Lad, Pao), put all of them in their mouth and spat some back (Sha, Rav, Lad, Syb, Sam), or shared the four raisins between both hands keeping the content of one and returning the content of the other (Kin, Sad, Syb). Each subject consistently used the same way across the different phases of study (except for Lad and Syb who alternated their removal procedures; they mainly used the second procedure but sometimes used the first one for Lad, or the third one for Syb). Phase 1 {#s2a} ------- In this phase, 20 subjects failed to adapt the amount of given raisins according to partners\' quality during 21 sets of two sessions ([Figure 1](#pone-0017801-g001){ref-type="fig"} and [S1](#pone.0017801.s001){ref-type="supplementary-material"}). Among subjects, seventeen consistently gave all four raisins to the doubling and fixed partners. Two other subjects (Sam, Pao) gave 1--3 raisins to both partners. A third subject (Arn) initially gave all four raisins, but after the 16^th^ set of two sessions, he learned to give 1--2 raisins to both partners. Comparing the performances of subjects according to partners\' quality in the last 10 sets of sessions did not yield significant differences (fixed partner: mean number of raisins ± sd  = 3.55±0.49, doubling partner: m = 3.56±0.36, n = 20 subjects, T = 53.0, p = 0.642). ::: {#pone-0017801-g001 .fig} 10.1371/journal.pone.0017801.g001 Figure 1 ::: {.caption} ###### Number of raisins returned by seven subjects in Phases 1, 2 and 3. In Phases 1 and 3, subjects were tested with both doubling and fixed partners. In Phase 2, subjects were tested with the fixed partner only. Six subjects successfully modified their strategy in Phase 2 except for Arn who already changed of behavior at the end of Phase 1. In Phase 3, Rav returned 1 raisin then stopped exchanging with the doubling partner. Each plot represents the mean number of raisins returned in one session of six trials, along with standard errors. :::  ::: One subject was able to adjust his behavior according to experimenters\' quality. This Tonkean macaque (Sha) was tested during 24 sets of two sessions. From the 17^th^ set, he gave a decreasing numbers of raisins (three to one) to the fixed partner while consistently returning 3--4 raisins to the doubling partner ([Figure 2](#pone-0017801-g002){ref-type="fig"}). Comparing his performances according to partners\' quality during the last 10 sets of sessions yielded a statistically significant difference (fixed: m = 2.51±1.35, doubling: m = 3.60±0.91, n = 10 sets, T = 4.0, p = 0.016). ::: {#pone-0017801-g002 .fig} 10.1371/journal.pone.0017801.g002 Figure 2 ::: {.caption} ###### Number of raisins returned by the subject Sha in Phases 1 and 3. :::  ::: Phase 2 {#s2b} ------- In Phase 2, the 20 subjects that had previously failed to differentiate between partners\' quality were tested in sessions involving a single fixed partner. Phase 2 was run to counterbalance the tendency of most subjects to return all 4 raisins in Phase 1. Among the 20 subjects, 13 maintained the main strategy used in Phase 1 (see [Figure S1](#pone.0017801.s001){ref-type="supplementary-material"}). The other seven subjects altered their behavior in the course of sessions. They learned to give 1--2 raisins to the fixed partner ([Figure 1](#pone-0017801-g001){ref-type="fig"}). Phase 3 {#s2c} ------- The seven subjects who reduced the number of raisins they gave in Phase 2 were tested again in sessions involving two different experimenters ([Figure 1](#pone-0017801-g001){ref-type="fig"}). Among them, six continued to give 1--2 raisins to both partners as in Phase 2. Comparing the performances of subjects according to partners\' quality did not yield significant differences (fixed: m = 1.05±0.10, doubling: m = 1.10±0.15, n = 6 subjects, T = 15.0, p = 0.144). A seventh subject (Rav) started to stop exchanging with the doubling partner, consuming the four raisins. Yet, he kept on giving one raisin to the fixed partner. The analysis showed that he responded differently to both partners\' qualities (fixed: m = 1.02±0.13, doubling: m = 0.17±0.39, n = 10 sets, T = 55.0, p = 0.005). The subject Sha, having differentiated between partners\' quality in Phase 1, was tested in phase 3 with two new experimenters in order to confirm his response ([Figure 2](#pone-0017801-g002){ref-type="fig"}). His behavior progressed during the sessions. At first, he gave about 1--2 raisins to the fixed partner while generally giving 2--3 raisins to the doubling one. After several sessions, he gave 2--3 raisins to the fixed partner and four to the doubling. In the last sessions, he gave a minimal number (one) to the fixed partner and a maximal number (four) to the other partner. Analyzing his performances showed that he adopted contrasting strategies according to partners\' quality (fixed: m = 1.23±1.26, doubling: m = 3.63±0.72, n = 10 sets, T = 55.0, p = 0.005). Net incomes in Phases 1 and 3 {#s2d} ----------------------------- By experimental design the subjects\' net income should differ according to experimenters\' quality. We checked that it was larger with the fixed than with the doubling partner in the last 10 sets of sessions in Phase 1 for Sha (fixed: m = 9.49±1.01, doubling: m = 7.60±0.34, n = 10 sets, T = 55.0, p = 0.005) and other subjects (fixed: m = 8.42±0.77, doubling: m = 7.61±0.84, n = 20 subjects, T = 41.0, p = 0.001), and also in Phase 3 for Sha (fixed: m = 10.17±0.87, doubling: m = 7.63±0.35, n = 10 sets, T = 55.0, p = 0.005), Rav (fixed: m = 10.98±0.13, doubling: m = 4.18±0.39, n = 10 sets, T = 55.0, p = 0.005) and other subjects (fixed: m = 10.95±0.20, doubling: m = 5.10±0.27, n = 6 subjects, T = 21.0, p = 0.028). In Phase 1, Sha received a total of 2414 raisins (1046 raisins with the doubling partner; 1368 raisins with the fixed partner; difference: 322 raisins). For other subjects, the total mean of raisins was of 2187 (983 raisins with the doubling partner; 1204 raisins with the fixed partner; mean difference: 221 raisins). In Phase 3, Sha earned a total income of 2478 raisins (1059 raisins with the doubling partner; 1419 raisins with the fixed partner; difference: 360 raisins). Rav had a total income of 2071 raisins (435 raisins with the doubling partner; 1636 raisins with the fixed partner; difference: 1201 raisins). For other subjects, the total mean of raisins was of 1855 (584 raisins with the doubling partner; 1271 raisins with the fixed partner; mean difference: 687 raisins). Discussion {#s3} ========== A single subject (Sha) could maximize pay-offs by following different rules according to experimenters\' quality. Most capuchins and macaques were not able to adapt the invested amount of food items to the potential returns from each experimenter. In Phase 1, most individuals consistently gave a maximal amount to both. Such strategy maximized pay-off with the doubling partner, but was inappropriate with the fixed one. Fewer subjects showed the reverse response pattern, giving a minimal number of raisins by the end of this phase. This strategy maximized pay-offs with the fixed partner, but not with the doubling. In Phase 2, subjects had to exchange only with the fixed partner. One third of them succeeded in maximizing pay-off and learned to give a minimal amount. Among these seven subjects, only one (Rav) discriminated between partners\' quality in Phase 3 but failed to understand the rule that would bring him optimal benefits with the doubling partner. The others subjects maintained the same strategy as in Phase 2 and did not adapt their investment strategy according to partners\' quality. It might be argued that the experimental set-up did not provide time enough for subjects to adjust their behavior, but the fact that Sha learned to modify his behavior after some trials weakens this interpretation. An alternative explanation is that most subjects may have been unable to differentiate between experimenters\' quality according to the food amounts that they returned. However, this explanation is also unlikely since it is known that monkeys are able to discriminate two experimenters behaving differently [@pone.0017801-Mitchell1]--[@pone.0017801-Brosnan1]. Moreover, most subjects sometimes gave back a different number of raisins to experimenters, thus getting an opportunity to learn that experimenters did not respond in the same manner. It should be emphasized that the net income differed according to experimenters\' quality, since no subjects always gave 0 or 4 raisins. In Phase 1, subjects experienced a difference of close to one raisin between experimenters; and in Phase 3, the subjects\' net income with the fixed partner was more than twice than with the doubling partner. Still, they did not adjust their behavior according to the partner\' quality. Moreover, former studies have shown that monkeys succeed in tasks requiring them to discriminate between quantities [@pone.0017801-Anderson1]--[@pone.0017801-vanMarle1]. When required to trade tokens for rewards with two different experimenters, tufted capuchins were able to select the one providing the higher pay-off [@pone.0017801-Chen1]. Here, subjects had to do more than just choosing between two options, they had to draw different decision rules from the contrasting conduct of two different human partners. From previous work on discrimination learning, we know that it is quite demanding for animals who learned in a training phase to select one cue in a two-choice discrimination task to learn, in a following reversal phase, that the second cue is then rewarded [@pone.0017801-Tomasello1], [@pone.0017801-Rumbaugh2]. Our experimental situation was even more challenging since it required subjects to respond in a different way at each partner\'s quality change. It is therefore not surprising that most subjects failed to regularly alternate their decision rule in this repeated conditional discrimination task. In Phase 2, seven subjects -- three macaques and four capuchin monkeys -- sized the opportunity to remove some raisins from the initial amount in order to maximize pay-off. This corroborates results previously found in a study where capuchins were observed nibbling part of the initial item before returning it [@pone.0017801-Drapier1], [@pone.0017801-Ramseyer1]. Therefore, failure to learn to keep some of the food was not what hindered success in this experiment. The fact that only a third of the monkeys succeeded is not too surprising. Indeed, monkeys were rewarded regardless of the number of raisins invested - even no exchange whatsoever rewarded them with the 4 raisins they kept. Therefore, there was no negative reinforcement for giving one quantity or another. Although we aimed to test whether monkeys could learn to differentiate between two experimenters\' qualities, we did not want to condition them to do so. In each phase, it was up to them to realize the differences in the rewards obtained according to the quality of the experimenter they were interacting with. Previous studies have shown that monkeys could recognize when experimenters subtracted several items from a given number of incentives [@pone.0017801-Hauser3], [@pone.0017801-Sulkowski1]. In the present study, some subjects consumed some of the raisins and gave the remaining to the experimenter, whereas others first gave some raisins and then ate the remaining ones. In both cases, subjects were able to remove 2--3 items from the total amount before giving 2-1 items to the experimenter. We propose that subjects\' decisions rested on their ability to recognize magnitudes, albeit in an imprecise way [@pone.0017801-Cordes1], [@pone.0017801-Gallistel1]. From the seven individuals who started giving back minimal amounts in Phase 2, one behaved differently with each partner\'s quality of Phase 3; while he consistently gave one raisin to the fixed partner, he eventually stopped exchanging with the doubling one. Thus, this capuchin monkey was able to maximize pay-offs with the fixed partner and could recognize that the doubling partner might respond in a less satisfactory way. Still, he failed to understand which rule would bring him optimal benefits with this second partner. It must be emphasized that one Tonkean macaque succeeded in optimizing pay-offs with both experimenters. He followed different decision rules with each experimenter\'s quality in Phase 1, and did it again with two new experimenters in Phase 3. By the end of each phase, he invested a maximal amount with the doubling partner and he removed most items before investing with the fixed one. To our knowledge, this represents the only example of decision-making by drawing different rules based on combination of discrete quantities in monkeys, and maybe even in great apes. The fact that only one of over 21 subjects could maximize benefits in adapting investment according to experimenters\' quality indicates that such a task is difficult for monkeys, albeit not impossible. Cognitive limits can underpin the present results, but we cannot exclude that different factors related to the design of the task concurred to create additional difficulties. First, one may argue that in Phase 3, monkeys only gave a minimal amount because they had been trained to do so in Phase 2, and that training at this stage would have also included training with the doubling partner. It is likely that training with only the fixed partner influenced their response greatly. However, training in Phase 2 was carried out to counterbalance the very strong tendency of the subjects to systematically give four raisins in Phase 1. If we had exposed them to both contingencies again, this could have forbidden the outcome "give as little as possible". Thus, in the actual set up Phase 3 was run with the knowledge that the seven subjects involved had all been capable of both responses, giving either a maximum (Phase 1) or a minimum (Phase 2) number of raisins. Second, albeit statistically significant, the weak difference of net income between different experimenter\'s qualities (1 raisin), experienced by individuals in Phase 1 could be insufficient for monkeys to detect that they were not maximising rewards. It is known that monkeys can distinguish between weak differences of items [@pone.0017801-Hauser1], [@pone.0017801-Wood1]. Moreover, in Phase 3 this difference was two-fold between experimenters. Nevertheless, subjects did not adjust their return according to experimenters\' qualities. Finally, it is possible that some individuals may require more exposure to each partner\'s quality in order to learn how to adjust their return. Whenever individuals showed unstable strategies in each phase, additional sessions were run to allow for such learning to occur. This however did not lead to successful learning. Still, sufficient learning time is probably a critical requirement for the adequate mastering of such complex cognitive decision-making by most subjects. In humans, being able to follow multiple directions or to switch between decision rules develops slowly during childhood [@pone.0017801-Blaye1], [@pone.0017801-Zelazo1]. Providing that sufficient learning time is allowed, and that monkeys can pay attention to differences in partners\' quality, maximizing pay-off using opposite decision rule is within the reach of these species. In the present experiments we reduced the complex interactions commonly addressed by behavioral biology and economics to a simple dyadic situation in which subjects interact with a human experimenter. This is a current procedure in experimental cognition. Further research should attend more specifically to those additional factors -- whether ecological, social or cognitive -- liable to facilitate such learning in non-human primates and other animals. Trading with multiple partners following different rules is characteristic of human economics; individuals make decisions based on their expectations regarding partners\' responses. Here, monkeys had to adjust the amount to be returned according to their expectations about the behavior of two different experimenters. Our results may have implications regarding how non-human primates manage their relationships with conspecifics. The ability to adapt pay-offs according to the gains potentially brought by each partner could be related to the ability of individuals to invest more in one mate or another [@pone.0017801-Brosnan3]--[@pone.0017801-Silk1]. Future studies should compare monkeys and great apes to investigate whether the development of such abilities would have preceded the rise of economical transactions in humans. Methods {#s4} ======= Ethics Statement {#s4a} ---------------- Animals were given *ad libitum* access to food and water. All procedures complied with the recommendations of the Weatherall report. The research was conducted under license 67--100 from the French Agricultural Department (Préfecture du Bas-Rhin). Subjects {#s4b} -------- The subjects were maintained at the Primatology Center of the Strasbourg University. Their age and sex are presented on [Table 2](#pone-0017801-t002){ref-type="table"}. We tested eight tufted capuchins (*Cebus apella*) belonging to a group of 18 individuals housed in an indoor-outdoor enclosure composed of several compartments totaling 78 m^2^. Four Tonkean macaques (*Macaca tonkeana*) belonged to a group of seven individuals housed in an indoor-outdoor enclosure composed of several compartments totaling 35 m^2^. Two other Tonkean macaques belonged to a group of 16 individuals raised in a 1-acre wooded area including a shelter and a 40-m^2^ wire-mesh fenced enclosure used for experiments. Three long-tailed macaques (*Macaca fascicularis*) were housed together in an enclosure of 10 m^2^ composed of several compartments and located in an indoor room. Four other long-tailed macaques were individually housed in the same room in cages of 125×80×80 cm. Animals were fed with commercial monkey diet. They were never deprived of food. ::: {#pone-0017801-t002 .table-wrap} 10.1371/journal.pone.0017801.t002 Table 2 ::: {.caption} ###### Subjects participating in the study. ::: {#pone-0017801-t002-2} Subjects Age (yrs) Sex Rearing conditions ---------------------- ----------- -------- -------------------------------- Tufted capuchins Kin 16 female group-living, indoor-outdoor Ali 9 female group-living, indoor-outdoor Pao 7 female group-living, indoor-outdoor Arn 10 male group-living, indoor-outdoor Pis 7 male group-living, indoor-outdoor Pop 7 male group-living, indoor-outdoor Rav 6 male group-living, indoor-outdoor Sam 5 male group-living, indoor-outdoor Tonkean macaques Syb 5 female group-living, indoor-outdoor Rim 6 male group-living, indoor-outdoor She 5 male group-living, indoor-outdoor Sim 5 male group-living, indoor-outdoor Lad 11 female group-living, semifree-ranging Sha 5 male group-living, semifree-ranging Long-tailed macaques Lou 11 male group-living, indoor Ram 16 male group-living, indoor Sad 12 male group-living, indoor Cas 12 male separated, indoor Don 16 male separated, indoor Jac 15 male separated, indoor Joe 11 male separated, indoor ::: Testing Procedure {#s4c} ----------------- Subjects had been trained to exchange food items with humans prior to experiments [@pone.0017801-Drapier1], [@pone.0017801-Pel1]. Most subjects had been involved in a delay-of-gratification task where they had to keep a piece of biscuit in their hand for a given amount of time before returning it for a better or larger reward. All subjects succeeded in waiting for more than 10 seconds in this task. The present study, by comparison, was based on an immediate exchange and imposed a lower need for self-control in all subjects. They were also involved in daily training sessions over a 3-month period where they had to give several Zante raisins to obtain twice the number of raisins. Another experiment gave subjects some background in discriminating between values of 6 and 18 food items [@pone.0017801-Steelandt1]. Group-living subjects were temporarily separated from their mates into individual compartments and later released back into their group. The experimenter sat in front of the wire mesh and laid four cups containing four potential rewards on the ground in full view of the subject. The number of potential rewards shown depended on the quality of the experimenter running the trial. A test started when the experimenter showed to the subject four raisins on a teaspoon for 2 s. Then she gave them to the subject. After 3 s, the experimenter held out a hand, palm open, in front of the subject requesting them back. When the subject gave one or more raisins, the experimenter rewarded the subject by supplying him/her with a corresponding, larger, number of raisins from one of the four potential cups ([Figure 3](#pone-0017801-g003){ref-type="fig"}). If the subject did not give raisins, the trial ended. We waited for 2 min after the end of food consumption before starting another trial. ::: {#pone-0017801-g003 .fig} 10.1371/journal.pone.0017801.g003 Figure 3 ::: {.caption} ###### Exchange sequence between capuchin monkey and experimenter. \(A) The experimenter presents four raisins on a spoon, (B) The subject is allowed to take the raisins, (C) The subject is requested to return the raisins, (D) The subject drops the raisins in the hand of the experimenter, (E) The subject receives eight raisins in a cup. :::  ::: Experimental Design {#s4d} ------------------- Two different experimenters familiar to the subjects were involved in the testing phase. A first one, the doubling partner, always returned a number of raisins twice those given by the subjects. Therefore, potential rewards consisted in cups presenting either two, four, six or eight raisins. The second experimenter, the fixed partner, always returned eight raisins, regardless of the number of raisins given by subjects (one to four). Thus, potential reward consisted in one cup among four, each cup presenting eight raisins. The subjects\' net income, i.e. the amount of raisins non-invested by the subject plus those received, could vary depending on which partner they interacted with ([Table 1](#pone-0017801-t001){ref-type="table"}). For training, a first 2 day-period was run where subjects were trained to give several raisins. Subjects were submitted to one daily session of six trials. A different experimenter from the two partners described above initially provided the subject with either one or four raisins and requested subjects to give them all to obtain eight raisins. Three trials were run in a random order for each condition. We did not require learning criteria for this step. In a second 2-day training period, subjects were familiarized to the doubling and fixed partners. They were exposed once (in a single trial) to the doubling partner and once to the fixed. In order for them to experience the difference in the reward amount, subjects had to give at least one raisin to each partner. If they failed, a second trial was run. Subjects needed between 2 and 4 trials to reach this criterion. With regard to the testing phase, we first tested subjects in successive sets of two sessions (one session per partner) in a random order. There was no more than one session of six trials per half-day. The subjects\' net income could vary within one session from 24 to 48 raisins with the doubling partner, and from 24 to 66 with the fixed partner. The partners\' role differed and was counterbalanced across subjects; the doubling partner for 11 subjects was the fixed one for the remaining ten. Because subjects failed to adapt their strategies according to the quality of the partner they were tested with, we ran them in a second phase involving the fixed partner only. We aimed to detect whether subjects could maximize their gain in a simplified version of the task. Phase 2 was run to counterbalance the tendency of subjects to return all 4 raisins in Phase 1. Indeed, during the training phase, all subjects had learned to return a maximum of raisins, which was the main behavior observed in Phase 1. In Phase 2 the goal was therefore to reinforce any subject who would start "giving less". When subjects did choose the best strategy in the second phase (giving only one raisin to obtain eight ones), we tested them in a third phase, which replicated the procedure of Phase 1. Phase 3 was then run with the knowledge that the seven subjects involved had all been capable of both responses, giving either a maximum (Phase 1) or a minimum number of raisins (Phase 2). A single subject (Sha) directly passed from Phase 1 to Phase 3 because of success in Phase 1. Each phase involved different experimenters. Whenever the strategy adopted by subjects was not stable at the end of each phase, and to ascertain that no learning trend was occurring, we added testing sessions until the performances\' curve flattened. In Phase 1, subjects were tested in 21 sets of two sessions with the doubling and fixed partners; the first set was a learning period. One subject (Sha) was tested in 24 sets of two sessions. Phase 2 was composed of 20 sessions with the single fixed partner. Four subjects (Pao, Kin, Pis, Rav) were tested in 40 sessions. In Phase 3, subjects were tested in 20 sets of two sessions with both partners\' qualities. We conducted 25 sets with Rav and 24 sets with Sha. Trials when subjects did not return any raisins (2.1% of trials) were discarded from data processing. To test whether subjects responded differently to the fixed and doubling partners, we compared their performances at the individual or at the group level in the last part of each testing phase, i.e. the last 10 sets of sessions, using a Wilcoxon matched-pairs test (exact procedure, [@pone.0017801-Mundry1]) with SPSS software version 16 (SPSS Inc., Chicago IL, U.S.A.). Supporting Information {#s5} ====================== Figure S1 ::: {.caption} ###### **Number of raisins returned by 13 subjects in Phases 1 and 2.** In Phase 1, each set is composed of one session with the doubling partner and another with the fixed one. In Phase 2, subjects were tested with the fixed partner only. They did not modify their strategy in this phase. Each plot represents the mean number of raisins returned in one session of six trials. Errors bars represent standard errors of the mean for each session. The subject Pao was tested for a larger number of sessions than others to ascertain that no learning trend occurred in its performances. (PDF) ::: ::: {.caption} ###### Click here for additional data file. ::: We are grateful to C. Arnaud, F. Ayari, M. Gallardo Ruiz, A. Leclerc-Imhoff, N. Moratscheck, L. Pelletier, G. Ruetsch, M. Schaaf, N. Soulaigre and P. Vuarin for valuable assistance with experiments, and to M. Bowler for proofreading. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**The research was supported by a grant from the Agence Nationale de la Recherche (ANR-08-BLAN-0042-01). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: BT SS VD. Performed the experiments: SS. Analyzed the data: SS. Wrote the paper: SS BT VD MHB.

PubMed Central

2024-06-05T04:04:19.898601

2011-3-10

PMC3053401

Introduction {#s1} ============ Liver cirrhosis is often complicated by an impaired renal capacity of maintaining water and sodium balance. Splanchnic arterial vasodilatation due to an increased release of endogenous vasodilators leads to compensatory activation of the endogenous vasoconstrictor systems: the sympathetic nervous system, the renin-angiotensin system and the non-osmotic release of vasopressin [@pone.0017891-Gines1]. This causes renal sodium/water retention and renal vasoconstriction eventually leading to the hepatorenal syndrome. In many patients, water retention can be controlled by sodium restriction and diuretics. However, diuretic therapy often entails deterioration of renal function [@pone.0017891-Gines1]. There is an urgent clinical need for alternative pharmacological approaches. The adenosine system with its four receptors (A~1~, A~2A~, A~2B~ and A~3~) is involved in several key functions of both liver and kidneys [@pone.0017891-Linden1]. Decreasing the portal flow, as it is the case in cirrhosis, results in an activation of the hepatic adenosine system [@pone.0017891-Ming1]. Adenosine, via A~1~ receptors (A~1~R), serves as a mediator for triggering the hepatorenal reflex leading to renal water and sodium retention [@pone.0017891-Ming2], [@pone.0017891-Ming3]. However, the exact localization of adenosine receptors within the liver remains unclear. In the kidney A~1~R are highly expressed in the preglomerular microcirculation, but also on the proximal tubules and other renal structures [@pone.0017891-Jackson1]. Short-term infusion of selective A~1~R antagonists inhibits the tubuloglomerular feedback and causes diuresis and natriuresis [@pone.0017891-Wilcox1], [@pone.0017891-Vallon1]. Blockade of hepatic and renal A~1~R could therefore provide a new therapeutic option in conditions with sodium and water retention such as liver cirrhosis [@pone.0017891-Gottlieb1]. Thioacetamide has been widely used to induce chronic liver injury in animal models since it mimics the human disease closely. Rats treated with thioacetamide present with typical cirrhotic liver damage and an impaired ability of excreting sodium and water [@pone.0017891-Ming4]. The first part of the present study evaluated the receptor binding affinity of the novel adenosine A~1~R antagonist SLV329 - a pyrimidine derivative - and confirmed its *in vivo* action on diuresis and natriuresis in healthy animals. As main hypothesis of this study, we tested whether SLV329 might exert diuretic effects without impairing renal function in animals with thioacetamide-induced liver cirrhosis. Results {#s2} ======= *In vitro* selectivity profile of SLV329 {#s2a} ---------------------------------------- In receptor binding experiments using cloned human receptors SLV329 behaved as a potent (pK~i~ 9.2) and selective A~1~R ligand. The affinity of SLV329 for the other adenosine receptors was at least 100-fold lower ([Table 1](#pone-0017891-t001){ref-type="table"}). The receptor-binding affinities and enzyme inhibitory properties of SLV329 were evaluated in a series of 94 receptors and 6 phosphodiesterases including adenosine transporters and a wide range of adrenergic, muscarinic, nicotinic, dopaminergic, serotoninergic, histaminergic, glutamatergic, opioid, angiotensin, bradykinin and neuropeptide receptors as well as ion channels and transporter sites. Only the significant affinities of SLV329 are summarized in [Table 1](#pone-0017891-t001){ref-type="table"}. Apart from adenosine receptors, significant binding was measured only for the high-affinity rolipram-binding site on phosphodiesterase 4. This effect was consistent with the inhibition of this enzyme by SLV329. However, phosphodiesterase 4 inhibition was much less potent than A~1~R binding. The ratio of the K~i~ value for phosphodiesterase 4 inhibition by SLV329 to that of the A~1~R binding was 3170. The activities on other phosphodiesterases were at least 10-fold lower. ::: {#pone-0017891-t001 .table-wrap} 10.1371/journal.pone.0017891.t001 Table 1 ::: {.caption} ###### Receptor binding affinities and enzyme inhibitory properties of SLV329. ::: {#pone-0017891-t001-1} Assay Cell/tissue Ligand SLV329 Affinity -------------------------------------- -------------- --------------- ----------------- Adenosine A~1~ CHO cells ^3^H-DPCPX 9.2±0.2 Adenosine A~2A~ HEK293 cells ^3^H-CGS21680 7.2±0.2 Adenosine A~3~ HEK293 cells ^3^H-AB-MECA 6.9±0.1 Adenosine A~2B~ HEK293 cells ^3^H-DPCPX 6.3±0.1 Phosphodiesterase 4 rolipram binding Total brain ^3^H-Rolipram 6.1±0.2 Phosphodiesterase 4 enzyme U-937 cells ^3^H-cAMP 5.4±0.1 Phosphodiesterase 6 enzyme Retina ^3^H-cAMP 4.3±0.1 Only significant affinities and enzyme inhibition of SLV329 are shown. Cells and tissues used to obtain receptors/enzymes and radioactive ligands of the respective assays are listed. Cells and tissues were provided by Cerep (Celle l\'Evescault, France) Results are expressed as pK~i~ for radioligand affinity assays, and as pIC~50~ for enzyme inhibition. Mean ± standard deviation of at least 3 determinations. ::: The half-maximally inhibitory concentration (IC~50~) value of SLV329 is 3.2 µg/l. Effects of SLV329 on renal function in healthy rats {#s2b} --------------------------------------------------- Five different intravenous doses of SLV329 were tested in healthy anesthetized rats. [Table 2](#pone-0017891-t002){ref-type="table"} shows that SLV329 treatment resulted in a marked dose-dependent increase of diuresis by up to 3.4-fold and sodium excretion by up to 13.5-fold compared to the time-matched vehicle control group (both p\<0.0001). The half-maximally effective concentration (EC~50~) values for diuresis and natriuresis were calculated to be 26 µg/l and 17 µg/l respectively. In contrast, the rate of potassium excretion was only modestly affected (up to only 1.7-fold vs. vehicle). The creatinine clearance was not affected by SLV329, even at the highest dose. SLV329 caused a dose-dependent increase in urinary adenosine excretion by up to 4.2-fold. ::: {#pone-0017891-t002 .table-wrap} 10.1371/journal.pone.0017891.t002 Table 2 ::: {.caption} ###### Effects of different doses of SLV329 in anesthetized rats. ::: {#pone-0017891-t002-2} ------------------------------------------------------- ------------------------------------------------------ ---------------------------------------------- ------------------------------------------------ ------------------------------------------------ ------------------------------------------------- ------------------------------------------------ ------ SLV329 bolus (µg/kg) Vehicle 18 54 180 540 945 infusion (µg/kg[\*](#nt107){ref-type="table-fn"}min) Vehicle 0.37 1.1 3.7 11 19.3 Plasma SLV329 (µg/l) 0±0 11±2.6 31±4.4 54±12 86±21 182±63 Diuresis (fold vs. vehicle) 1±0.1 1.4±0.1[\*](#nt107){ref-type="table-fn"} 2.5±0.3[\*\*\*\*](#nt110){ref-type="table-fn"} 2.3±0.3[\*\*\*](#nt109){ref-type="table-fn"} 3.4±0.4[\*\*\*\*](#nt110){ref-type="table-fn"} 2.8±0.4[\*\*\*\*](#nt110){ref-type="table-fn"} Diuresis (ml/kg[\*](#nt107){ref-type="table-fn"}h) 1±0.1 1.4±0.1[\*](#nt107){ref-type="table-fn"} 2.5±0.3[\*\*\*\*](#nt110){ref-type="table-fn"} 2.3±0.3[\*\*\*](#nt109){ref-type="table-fn"} 3.4±0.4[\*\*\*\*](#nt110){ref-type="table-fn"} 2.8±0.4[\*\*\*\*](#nt110){ref-type="table-fn"} Natriuresis (fold vs. vehicle) 1±0.2 5.0±0.8[\*\*\*](#nt109){ref-type="table-fn"} 7.6±0.9[\*\*\*\*](#nt110){ref-type="table-fn"} 7.9±1.3[\*\*\*\*](#nt110){ref-type="table-fn"} 13.5±1.3[\*\*\*\*](#nt110){ref-type="table-fn"} 8.3±1.4[\*\*\*\*](#nt110){ref-type="table-fn"} Natriuresis (µg/kg[\*](#nt107){ref-type="table-fn"}h) 65±16 324±52[\*\*\*](#nt109){ref-type="table-fn"} 497±62[\*\*\*\*](#nt110){ref-type="table-fn"} 518±84[\*\*\*\*](#nt110){ref-type="table-fn"} 882±87[\*\*\*\*](#nt110){ref-type="table-fn"} 544±94[\*\*\*\*](#nt110){ref-type="table-fn"} Kaliuresis (fold vs. vehicle) 1±0.1 1.6±0.2 1.4±0.1 1.7±0.1[\*\*](#nt108){ref-type="table-fn"} 1.3±0.1 1.6±0.2 Kaliuresis (µg/kg[\*](#nt107){ref-type="table-fn"}h) 106±13 167±21 146±13 175±16[\*\*](#nt108){ref-type="table-fn"} 136±12 170±22 Creatinine clearance (ml/min) 1.9±0.3 3.2±1.1 2.0±0.3 1.6±0.2 2.6±0.8 2.4±0.7 Adenosine excretion (nmol/h) 2.4±0.7 3.1±0.4[\*](#nt107){ref-type="table-fn"} 4.9±0.8[\*\*](#nt108){ref-type="table-fn"} 5.9±1.5[\*](#nt107){ref-type="table-fn"} 10.2±2.3[\*\*\*](#nt109){ref-type="table-fn"} 9.8±1.9[\*\*\*](#nt109){ref-type="table-fn"} Body weight (g) 363±3 360±2 356±3 362±2 364±2 357±2 Systolic blood pressure (mmHg) 107±6 108±2 104±3 110±5 113±7 116±6 Heart rate (beats/min) 313±19 317±23 328±21 350±10 294±23 301±30 ------------------------------------------------------- ------------------------------------------------------ ---------------------------------------------- ------------------------------------------------ ------------------------------------------------ ------------------------------------------------- ------------------------------------------------ ------ A bolus of SLV329 followed by an intravenous infusion over 3 hours produces a strong dose-dependent stimulation of renal water, sodium and adenosine excretion without any effect on creatinine clearance. Diuresis and electrolyte excretion is expressed as fold stimulation vs. time-matched vehicle controls. Data are means ± standard error of the mean. n = 9-13 animals per dose. \*p\<0.05, \*\*p\<0.01, \*\*\*p\<0.001, \*\*\*\*p\<0.0001 vs. time-matched vehicle controls. ::: Effects of SLV329 on renal function in thioacetamide-induced liver cirrhosis {#s2c} ---------------------------------------------------------------------------- Descriptive data, results from histological evaluation and Western blots are summarized in [Table 3](#pone-0017891-t003){ref-type="table"}. Representative immunoblots for hepatic and renal adenosine receptors are shown in [Figure 1](#pone-0017891-g001){ref-type="fig"}. The results of plasma and urine analyses are shown in [Table 4](#pone-0017891-t004){ref-type="table"}. ::: {#pone-0017891-g001 .fig} 10.1371/journal.pone.0017891.g001 Figure 1 ::: {.caption} ###### Representative immunoblots for hepatic and renal adenosine A~1~, A~2A~, A~2B~ and A~3~ receptors. Beta-actin was used for normalization before statistical analysis. Con, control; Cir, liver cirrhosis; Fur, furosemide; SLV, SLV329; kDa, kilodalton. :::  ::: ::: {#pone-0017891-t003 .table-wrap} 10.1371/journal.pone.0017891.t003 Table 3 ::: {.caption} ###### Effects of treatment with furosemide and SLV329 in rats with and without liver cirrhosis. ::: {#pone-0017891-t003-3} Group Con Con+Fur Con+SLV Cir Cir+Fur Cir+SLV ------------------------------------------------------------------- ---------- -------------------------------------- ------------------------------------------ ---------------------------------------------- ---------------------------------------------------------------------------- ------------------------------------------------------------------------------ Bodyweight week 0 (g) 358±6 344±10 356±11 351±6 354±6 352±5 Bodyweight week 8 (g) 443±11 433±13 439±15 333±8[\*\*\*](#nt116){ref-type="table-fn"} 345±6[\*\*\*](#nt116){ref-type="table-fn"} 335±5[\*\*\*](#nt116){ref-type="table-fn"} Bodyweight week 16 (g) 464±11 448±10 479±16 299±9[\*\*\*](#nt116){ref-type="table-fn"} 263±13[\*\*\*](#nt116){ref-type="table-fn"} 296±9[\*\*\*](#nt116){ref-type="table-fn"} [\#](#nt119){ref-type="table-fn"} Water intake week 0--8 (ml/kg[\*](#nt114){ref-type="table-fn"}d) 41±4 47±4 40±5 49±2[\*](#nt114){ref-type="table-fn"} 46±1 50±2 Water intake week 8--16 (ml/kg[\*](#nt114){ref-type="table-fn"}d) 65±2 69±2 70±4 44±2[\*\*\*](#nt116){ref-type="table-fn"} 52±2[\*\*\*](#nt116){ref-type="table-fn"} [§](#nt117){ref-type="table-fn"} 50±2[\*\*](#nt115){ref-type="table-fn"} [§](#nt117){ref-type="table-fn"} Water intake week 11 (ml/kg[\*](#nt114){ref-type="table-fn"}d) 63±4 76±5[§](#nt117){ref-type="table-fn"} 72±3[§§](#nt118){ref-type="table-fn"} 48±2[\*\*](#nt115){ref-type="table-fn"} 52±2[\*\*](#nt115){ref-type="table-fn"} 54±4[\*\*](#nt115){ref-type="table-fn"} [§](#nt117){ref-type="table-fn"} Systolic BP week 0 (mmHg) 124±2 126±6 130±3 126±1 125±2 127±2 Systolic BP week 12 (mmHg) 122±3 119±5 134±7[§](#nt117){ref-type="table-fn"} 110±5 118±5 118±3[\*](#nt114){ref-type="table-fn"} Liver weight (g) 12.5±0.5 11.9±0.5 13.2±0.7 13.9±0.5 13.4±0.5[\*](#nt114){ref-type="table-fn"} 14.1±0.3 Kidney weight (g) 3.1±0.1 3.0±0.1 3.2±0.1 3.1±0.1 3.0±0.1 2.9±0.1 Liver interstitial fibrosis (%) 0.3±0.1 0.4±0.1 0.9±0.3 8.6±1.1[\*\*\*](#nt116){ref-type="table-fn"} 5.2±1.1[\*\*\*](#nt116){ref-type="table-fn"} 8.6±1.4[\*\*\*](#nt116){ref-type="table-fn"} Kidney interstitial fibrosis (%) 3.0±0.3 3.4±0.6 3.5±0.6 2.9±0.3 3.1±0.6 3.2±0.4 Kidney media/lumen ratio 3.3±0.4 3.1±0.2 2.6±0.3 2.6±0.2 2.8±0.2 3.0±0.3 Glomerulosclerosis index (0--4) 1.8±0.04 1.8±0.1 1.6±0.1 1.7±0.05 1.7±0.1 1.8±0.1[\*](#nt114){ref-type="table-fn"} Liver A~1~ receptor 1.0±0.1 1.0±0.2 1.1±0.1 0.8±0.1 1.0±0.1 1.0±0.1 Liver A~2A~ receptor 1.0±0.2 1.2±0.1 0.9±0.1 0.7±0.1 0.8±0.1 0.8±0.1 Liver A~2B~ receptor 1.0±0.2 1.0±0.3 0.7±0.1 0.4±0.1[\*](#nt114){ref-type="table-fn"} 0.6±0.1 0.6±0.1 Liver A~3~ receptor 1.0±0.2 0.7±0.1 0.9±0.2 0.8±0.2 1.0±0.2 0.8±0.1 Kidney A~1~ receptor 1.0±0.03 1.0±0.2 0.6±0.1[§§](#nt118){ref-type="table-fn"} 0.8±0.1[\*](#nt114){ref-type="table-fn"} 0.8±0.1 0.7±0.1 Kidney A~2A~ receptor 1.0±0.1 0.9±0.2 1.0±0.1 0.9±0.2 1.2±0.2 1.2±0.1 Kidney A~2B~ receptor 1.0±0.1 0.9±0.2 1.4±0.2 1.3±0.2 1.3±0.1 1.0±0.2 Kidney A~3~ receptor 1.0±0.1 0.7±0.2 0.8±0.2 0.7±0.1 1.2±0.2 0.9±0.3 Con, control; Cir, liver cirrhosis; Fur, furosemide; SLV, SLV329, BP, blood pressure. Results of adenosine receptors from Western blots are expressed as fold expression vs. Con. Data are means ± standard error of the mean. n = 8--14 per group. \*p\<0.05, \*\*p\<0.01, \*\*\*p\>0.001 vs. Con group with same treatment, § p\<0.05, §§ p\<0.01 vs. same group without Fur or SLV treatment, \# p\<0.05 vs. same group with Fur treatment. ::: ::: {#pone-0017891-t004 .table-wrap} 10.1371/journal.pone.0017891.t004 Table 4 ::: {.caption} ###### Effects of treatment with furosemide and SLV329 on plasma and urine parameters in rats with and without liver cirrhosis. ::: {#pone-0017891-t004-4} Group Con Con+Fur Con+SLV Cir Cir+Fur Cir+SLV ---------------------------------- --------- --------- -------------------------------------------- ---------------------------------------------- ------------------------------------------------------------------------ ------------------------------------------------------------------------------- ALT week 8 (U/l) 26±3 36±5 29±4 41±6 27±3[\*](#nt123){ref-type="table-fn"} [§](#nt126){ref-type="table-fn"} 31±4 ALT week 16 (U/l) 55±4 48±2 47±4 73±9 66±11 53±7 Bilirubin week 8 (µmol/l) 1.3±0.1 1.5±0.3 1.4±0.2 7.3±0.7[\*\*\*](#nt125){ref-type="table-fn"} 6.0±0.8[\*\*\*](#nt125){ref-type="table-fn"} 5.7±0.9[\*\*\*](#nt125){ref-type="table-fn"} Bilirubin week 16 (µmol/l) 1.2±0.3 1.5±0.4 1.3±0.3 16±1.8[\*\*\*](#nt125){ref-type="table-fn"} 22±2.9[\*\*\*](#nt125){ref-type="table-fn"} 12±2.0[\*\*\*](#nt125){ref-type="table-fn"} [\#](#nt128){ref-type="table-fn"} Albumin week 8 (g/l) 31±0.4 32±0.4 31±0.2[\#](#nt128){ref-type="table-fn"} 31±0.5 32±0.5 31±0.4 Albumin week 16 (g/l) 29±0.5 29±0.7 28±0.3[\#](#nt128){ref-type="table-fn"} 26±0.6[\*\*](#nt124){ref-type="table-fn"} 25±1.0[\*](#nt123){ref-type="table-fn"} 26±0.5[\*\*\*](#nt125){ref-type="table-fn"} Creatinine week 8 (µmol/l) 52±2 54±3 54±3 55±2 52±2 49±1[§](#nt126){ref-type="table-fn"} Creatinine week 16 (µmol/l) 60±1 60±2 55±1[§§](#nt127){ref-type="table-fn"} ^\#^ 51±2[\*](#nt123){ref-type="table-fn"} 47±2[\*\*](#nt124){ref-type="table-fn"} 45±2[\*\*](#nt124){ref-type="table-fn"} [§](#nt126){ref-type="table-fn"} Urine volume week 8 (ml/d) 51±6 59±4 52±8 58±3 61±3 65±5 Urine volume week 16 (ml/d) 71±8 75±3 67±9 67±8 58±6 69±7 Na^+^ excretion week 8 (mmol/d) 3.2±0.5 3.1±0.2 2.9±0.5 3.6±0.5 3.7±0.4 2.9±0.3 Na^+^ excretion week 16 (mmol/d) 3.7±0.6 5.5±0.5 4.4±0.6 4.7±0.6 3.8±0.9 4.1±0.3 K^+^ excretion week 8 (mmol/d) 4.4±0.6 5.0±0.5 4.3±0.5 4.4±0.5 4.8±0.4 4.3±0.4 K^+^ excretion week 16 (mmol/d) 7.6±0.6 7.7±0.8 6.7±0.7 6.2±0.6 5.0±0.7 7.0±0.4 Con, control; Cir, liver cirrhosis; Fur, furosemide; SLV, SLV329. In week 0, there were no significant differences between the groups, except for plasma creatinine (Con+Fur 46±6 vs. Cir+Fur 53±5 µmol/l; p\<0.05) and urine volume (Con+Fur 70±16 vs. Cir+Fur 58±6 ml/d; p\<0.05). Data are means ± standard error of the mean. n = 8--14 per group. \*p\<0.05, \*\*p\<0.01, \*\*\*p\>0.001 vs. Con group with same treatment, § p\<0.05, §§ p\<0.01 vs. same group without Fur or SLV treatment, \# p\<0.05 vs. same group with Fur treatment. ::: Livers of the thioacetamide-treated rats showed macroscopic signs of nodular cirrhosis and tended to be heavier than livers from control groups. While liver fibrosis was seen in thioacetamide-treated rats (p\<0.001), kidney histology showed no relevant alterations in any of the groups. During the development of liver cirrhosis bilirubin rose markedly, whereas there was a significant decrease of plasma albumin. There was no relevant amount of ascites in any of the cirrhotic groups. Bodyweight (p\<0.001) and water intake (p\<0.01) was lower in rats treated with thioacetamide as compared to control animals. Furosemide and SLV329 caused an increase of mean water intake (probably reflecting increased diuresis), particularly at the beginning of the treatment (e.g. week 11, p\<0.05; [Table 3](#pone-0017891-t003){ref-type="table"}). Plasma creatinine was reduced in all cirrhotic animals, most probably due to reduced muscle mass because of wasting. Plasma creatinine was significantly lower in the SLV329-treated groups, in spite of higher mean body weight. The creatinine clearance was significantly reduced in cirrhotic animals compared to controls (−36,5%, p\<0.05; [Figure 2](#pone-0017891-g002){ref-type="fig"}). This was even slightly more pronounced in the group treated with furosemide (−41.9%, p\<0.01 compared to controls). In contrast, the creatinine clearance of the cirrhotic group treated with SLV329 did not differ significantly from any of the non-cirrhotic groups. ::: {#pone-0017891-g002 .fig} 10.1371/journal.pone.0017891.g002 Figure 2 ::: {.caption} ###### Creatinine clearance (week 16) in rats with and without liver cirrhosis. There were no significant differences between the groups in week 0 and week 8. In week 16 creatinine clearance of the cirrhotic SLV329 group was not significantly different from any of the non-cirrhotic groups. Data are means ± standard error of the mean. n = 8--10 per group (week 16). \*p = 0.05, \*\*p\<0.01 vs. non-cirrhotic control rats. ^§^p = 0.07 vs. cirrhotic animals with furosemide treatment. :::  ::: The reduction of mortality in cirrhotic animals treated with SLV329 compared to untreated cirrhotic animals did not reach statistical significance. However, mortality was significantly lower compared to cirrhotic animals treated with furosemide (17% vs. 54%, p\<0.05; [Figure 3](#pone-0017891-g003){ref-type="fig"}). ::: {#pone-0017891-g003 .fig} 10.1371/journal.pone.0017891.g003 Figure 3 ::: {.caption} ###### Kaplan-Meier mortality chart of rats with liver cirrhosis (Cir; mortality 5/14), cirrhotic rats with furosemide treatment (Cir+Fur; 7/13), cirrhotic rats with SLV329 treatment (Cir+SLV; 2/12) and all control groups without liver cirrhosis (Con; 0/24). p\<0.05 for Cir+Fur vs. Cir+SLV (log-rank test). :::  ::: In the Western blot analyses hepatic A~2~R seemed to be reduced in cirrhotic rats. This difference was statistically significant for A~2A~R and A~2B~R when testing all cirrhotic rats vs. all controls (p\<0.01). Treatment with SLV329 reduced A~1~R expression in renal tissue of non-cirrhotic animals (p\<0.01). Mean plasma concentrations of SLV329 were 49±8 µg/l on day 10 after the beginning of SLV329 treatment and 54±13 µg/l after 4 weeks in 6 random animals. Heart rate, lipase, creatine kinase and urinary protein excretion were not different between the groups at any time (data not shown). Discussion {#s3} ========== The present study was set out to investigate the receptor binding affinity of the novel A~1~R antagonist SLV329, to evaluate its in vivo effects on diuresis and natriuresis in healthy animals and to find out whether it exerts beneficial effects in an animal model of thioacetamide-induced liver cirrhosis. Receptor binding experiments showed that SLV329 behaves as a potent and selective A~1~R ligand *in vitro*. Intravenous treatment of healthy rats with SLV329 resulted in a strong dose-dependent diuretic and natriuretic effect, whereas the effect on kaliuresis was relatively small and the creatinine clearance remained unchanged. This is in line with effects that have been reported in a variety of rat strains including Wistar rats [@pone.0017891-Miracle1]--[@pone.0017891-Knight1]. The ability of diuretics to prevent sodium reabsorption results in an increased delivery of electrolytes to the distal tubule. This leads, in turn, to an augmented release of adenosine, which may activate A~1~R in afferent arterioles [@pone.0017891-Vallon2]. By blocking those receptors, A~1~R antagonists cause an uncoupling of the tubuloglomerular feedback which may at least partly explain the elevated concentrations of adenosine measured in the urine of animals treated with SLV329 [@pone.0017891-Wilcox3], [@pone.0017891-Vallon3]. However, urinary excretion of paracrine mediators do not necessarily reflect their local tissue concentrations. Alternatively, blockade of A~1~R might result in elevated intracellular cyclic adenosine monophosphate levels and release in the kidney, which will eventually lead to increased extracellular adenosine concentrations due to cyclic adenosine monophosphate degradation [@pone.0017891-Jackson2]--[@pone.0017891-Toya1]. However, in spite of the observed increase in urinary adenosine excretion, SLV329 did not decrease the rate of creatinine clearance, even at the highest dose. Previous studies demonstrated that a single application of an A~1~R antagonist causes an increase of renal sodium and water excretion in animals and patients with liver cirrhosis without affecting the glomerular filtration rate [@pone.0017891-Ming5], [@pone.0017891-Stanley2]. As a next step towards a possible clinical application, the present study investigated for the first time the effects of a chronic application of a selective A~1~R antagonist on kidney function and mortality starting at an early stage of liver cirrhosis. This proof-of-concept experiment included a low-dose furosemide-treated group because loop diuretics are often applied in cirrhotic patients and tend to deteriorate renal function. However, monotherapy with a loop diuretic is of course not the typical clinical situation at an early stage of liver cirrhosis without severe water retention. In the present study the creatinine clearance, used as a surrogate for the glomerular filtration rate, was significantly reduced in cirrhotic animals, especially in those receiving furosemide. In contrast, the A~1~R antagonist SLV329 was able to prevent this decline of creatinine clearance. The reduction of mortality in cirrhotic animals treated with SLV329 in comparison to vehicle treatment was not statistically significant. However, mortality was significantly lower in cirrhotic animals treated with SLV329 in comparison to animals treated with furosemide (17% vs. 54%). It is a limitation of this study that creatinine clearance was used instead of inulin clearance. Inulin clearance was not used because mortality would have increased further due to additional anesthesia. Creatinine clearance is influenced by muscle mass, liver function and tubular secretion of creatinine. However, there were no significant differences of neither body weight nor liver function between cirrhotic animals with and without SLV329 treatment. Tubular secretion of creatinine increases with declining glomerular filtration rate. Thus, the glomerular filtration rate of untreated cirrhotic animals and those treated with furosemide will be even lower than the creatinine clearance suggests, making the difference to the group treated with SLV329 even larger. Mean SLV329 plasma concentrations of 49--54 µg/l were actually reached by chronic application in the cirrhosis model. This plasma concentration was able to cause strong short-term effects on diuresis and saliuresis when administered intravenously as shown in [Table 2](#pone-0017891-t002){ref-type="table"}. A steady-state is usually reached in the course of long-term diuretic therapy by means of compensatory mechanisms of tubular reabsorption [@pone.0017891-Reyes1]. This explains why the 24-hour urine volume and electrolyte excretion in week 16 is not different between the groups of the cirrhosis model. However, when looking at mean water intake over 8 weeks, as a putative surrogate of diuresis, the effects of furosemide and SLV329 can be detected. It would have been interesting to evaluate fractional sodium excretion and free water clearance. This was not done due to limited plasma quantities. SLV329 does not affect the expression of hepatic or renal adenosine receptors in cirrhotic animals. The expression of A~2~R seems to be reduced in cirrhotic animals, independent of the treatment group. As yet, there are no reports on the expression of A~2~R in cirrhotic liver tissue. However, it is known, that hepatic A~2~R play an active role in the pathogenesis of hepatic fibrosis [@pone.0017891-Chan1]. It is a limitation of this study that adenosine receptor expression was evaluated only in liver and kidney homogenates by Western blot. Immunohistochemistry or radioligand binding experiments might give further detailed information. The beneficial effects of SLV329 in the cirrhosis model cannot be explained by morphological effects since SLV329 did not relevantly influence the degree of liver fibrosis, kidney histology or expression of hepatic or renal adenosine receptors. Animal studies suggest that liver cirrhosis activates the hepatorenal reflex via A~1~R, leading to renal water and sodium retention [@pone.0017891-Ming6], [@pone.0017891-Ming7]. Animal and human studies suggest that a resetting of the tubuloglomerular feedback contributes to the pathophysiology of kidney impairment in liver cirrhosis [@pone.0017891-Sansoe1]. Thus, inhibition of both the hepatorenal reflex and the tubuloglomerular feedback might explain the higher rate of creatinine clearance in the animals treated with SLV329. In addition to the effect on creatinine clearance, yet unknown effects of SLV329 might contribute to the reduction of mortality. As an addition to the present study it would be interesting to study liver cirrhosis in the established murine A~1~R knockout model. The ability of A~1~R antagonists to induce diuresis and natriuresis while not compromising glomerular filtration rate has become an attractive therapeutic option for the treatment of other fluid retention disorders, e.g. in kidney disease and heart failure, especially in conditions associated with diuretic resistance [@pone.0017891-Hocher1]. The preexisting experience with this class of drugs for other indications (including a large phase 3 trial) might facilitate the future translation of the results of this study to clinical application [@pone.0017891-Hocher2]. In conclusion, this study described some pharmacodynamic characteristics of the novel adenosine A~1~R-specific antagonist SLV329 and demonstrated its long-term safety and efficacy in an animal model of liver cirrhosis. Chronic SLV329 treatment starting at an early stage of liver cirrhosis prevented the decrease of creatinine clearance. Further studies will have to evaluate, whether SLV329 or other A~1~R antagonists are clinically beneficial at different stages of liver cirrhosis, either as an add-on to aldosterone antagonists or in combination with loop diuretics. Materials and Methods {#s4} ===================== Receptor binding and enzyme assays {#s4a} ---------------------------------- Receptor binding affinities as well as enzyme inhibitory properties of the new compound SLV329 were evaluated in a series of 94 receptors and 6 phosphodiesterases as described previously [@pone.0017891-Kalk1]. All cells and tissues needed for the assays were provided by Cerep (Celle l\'Evescault, France). Receptor binding assays were conducted as follows: after incubation of SLV329 with a receptor preparation and its radioactive ligand, the receptor preparations were rapidly filtered under vacuum through glass fiber filters, the filters were washed extensively with an ice-cold buffer using a harvester. Bound radioactivity was measured by scintillation counting using a liquid scintillation cocktail. Enzyme assays were carried out as follows: after incubation of SLV329 with an enzyme preparation and its radioactive substrate, radioactivity of the enzyme product was measured by scintillation counting using a liquid scintillation cocktail. Testing was done at a 3-log concentration range around a predetermined half-maximally inhibitory concentration (IC~50~) for the respective assay. The highest concentration tested for primes was 10 µM in receptor binding and 100 µM for enzyme assays. If no significant receptor binding or enzyme inhibition was detected at those concentrations SLV329 was considered to be inactive. Results were calculated as percentage of control values (enzyme assays) or for receptor binding assays as percentage of total ligand binding and that of nonspecific binding per concentration of SLV329. From the concentration-displacement curves IC~50~ values were determined by nonlinear regression analysis using Hill equation curve fitting. The inhibition constants (K~i~) were calculated from the Cheng-Prusoff equation K~i~ = IC~50~/(1+L/K~d~), where L is the concentration of radioligand in the assay and K~d~ the affinity of the radioligand for the receptor. Results were expressed as mean pKi values ± standard deviation (SD) of at least three separate experiments. Effects of SLV329 on renal function in healthy rats {#s4b} --------------------------------------------------- All animal experiments of this study were conducted in strict accordance with the European Convention for the Protection of Vertebrate Animals used for Experimental and Other Scientific Purposes (ETS123) and the German law on animal welfare and all efforts were made to minimize suffering. The study protocol was approved by the Niedersächsisches Landesamt für Verbraucherschutz und Lebensmittelsicherheit (Bezirksregierung Hannover, approval number 509.6.42502-04/875). Male Sprague-Dawley Crl:CD(SD)BR rats (150--170 g) were fasted overnight. The rats were anesthetized with 80 mg/kg thiobutabarbital intraperitoneally; additional doses of 40 mg/kg were given 2.5 h and 5 hours later. Catheters were placed in one jugular vein (for SLV329 or vehicle administration), one carotid artery (for blood sampling and blood pressure measurements), and the bladder. The rats were kept on a heated table to maintain body temperature at 37°C. After an equilibration period of 30 min, urine was sampled for a period of 3 h, then the animals received vehicle or SLV329 as follows: a slowly applied (30--60 s) loading bolus of 18, 54, 180, 540, and 945 µg/kg of SLV329 (in 0.45 ml/kg), followed by a continuous intravenous infusion at a rate of 0.37, 1.1, 3.7, 11, and 19.3 µg/kg SLV329 per minute (in 9.3 µl/kg per minute) for three hours. Urine and blood samples were collected 2 min before the SLV329 bolus and after three hours of SLV329 infusion. 5-sulfosalicylic acid was added to the urine aliquots for adenosine measurements at a final concentration of 8 g/l before freezing. Electrolytes and creatinine in plasma and urine were measured by standard automated analyzer by Medizinisches Labor Hannover (Hannover, Germany). Urine adenosine concentrations were quantified by high-pressure liquid chromatography using a MZ nucleosil C18 column (125×4 mm, 10 µm) with ultraviolet-detection (Immundiagnostik, Bensheim, Germany). Quantification of SLV329 plasma concentrations was performed after solid phase extraction using a validated reversed phase high-pressure liquid chromatography method with MS/MS-detection (Sciex Api 3000, Perkin Elmer, Waltham, MA, USA). The data describing the concentration-dependence of the diuretic and natriuretic effects of SLV329 were fitted to estimate the half-maximally effective concentration (EC~50~) using Prism 5.00 (GraphPad Software Inc., San Diego, CA, USA). Effects of SLV329 in thioacetamide-induced liver cirrhosis {#s4c} ---------------------------------------------------------- The study protocol was approved by the Landesamt für Gesundheit und Soziales, Berlin (approval number G0163/06). Male Wistar rats (250--300 g) were maintained under controlled conditions (20±2°C, 12 h light/dark cycle) and kept on a standard diet (0.2% sodium) with water *ad libitum*. Animals were divided into 6 groups: 1. Controls (Con; n = 8) 2. Controls with furosemide treatment (Con+Fur; n = 8) 3. Controls with SLV329 treatment (Con+SLV; n = 8) 4. Cirrhosis (Cir; n = 14) 5. Cirrhosis with furosemide treatment (Cir+Fur; n = 13) 6. Cirrhosis with SLV329 treatment (Cir+SLV; n = 12) Liver cirrhosis was induced by oral administration of thioacetamide via drinking water for 18 weeks. The initial concentration was 0.03%. This concentration was modified weekly according to weight changes in response to thioacetamide. The concentration was increased/reduced by 0.015% (absolute) in case of weight gain/loss \>25 g per week. Treatment with furosemide or SLV329 was started in week 8, since it is known that liver cirrhosis develops by then [@pone.0017891-Ikejima1]. Thioacetamide concentrations remained unchanged from week 8 until week 12. In week 12 the concentration was increased by 0.015% in all animals receiving thioacetamide. The resulting concentration was given until week 18. Furosemide was injected intraperitoneally (7.5 mg/kg) thrice weekly (always between 8 and 10 a.m.) in the respective groups starting from week 8 until study end. Also starting from week 8 the respective groups received a standard rat chow formulated with SLV329. Chow was obtained from Altromin (Lage, Germany) in several concentrations (0.0075%, 0.014%, and 0.05% SLV329) and fed to the rats according to food intake. The target concentration was 5 mg/kg per day. Subsequently calculated mean intakes of the groups were Con+SLV 5.4 mg/kg\*d and Cir+SLV 5.1 mg/kg\*d. Blood was taken from the retro-orbital vein plexus 10 days and 4 weeks after the beginning of SLV329 treatment in 6 random animals to measure the plasma concentration of SLV329. The samples were analyzed as described above. Body weight, water and food intake were measured weekly. Systolic blood pressure and heart rate were measured in week 0 and 12 by tail plethysmography as described previously.[@pone.0017891-Quaschning1] All animals were placed in metabolism cages in week 0, 8 and 16. Blood was taken from the retro-orbital vein plexus. The animals were sacrificed after week 20 and liver and kidneys were excised. Kidney samples were embedded in paraffin, cut into 3 µm sections and submitted to periodic acid-Schiff, elastica and sirius red staining. The extent of glomerulosclerosis, the renal media/lumen ratio and renal interstitial fibrosis was determined as described previously [@pone.0017891-Pfab1]. The extent of liver fibrosis was analyzed as in the kidneys. Alanine aminotransferase, bilirubin, albumin, creatinine, lipase and creatine kinase were measured in plasma using commercially available assays (ABX Pentra 400, Horiba Medical, Montpellier, France). Creatinine, sodium, potassium and protein were measured in urine also using ABX Pentra 400. Western blot {#s4d} ------------ Western blot was performed in liver and kidney tissue for A~1~, A~2A~, A~2B~ and A~3~ receptors using a method based on a previous publication [@pone.0017891-Jackson3]. In brief, snap frozen liver and kidney samples were pulverized in liquid nitrogen and dissolved in lysis buffer. After incubation at room temperature for 10 min, the suspension was centrifuged (16.200 g, 20°C, 45 min). Protein concentrations were determined in the supernatant using the Bradford method. Supernatant and Bradford solution (40 µl each) were mixed in a well of a microtiter plate and incubated for 10 min at room temperature on a shaker. Extinction was then measured photometrically at 595 nm. The protein concentration of each sample was calculated according to a standard dilution series. The samples were diluted with lysis buffer (5 g/l) to assure equal loading. Samples (23 µg protein per lane) were separated by sodium dodecylsulfate polyacrylamide gel electrophoresis (10%, 80 V for 30 min, then 110 V) and semi-dry-blotted (1,5 mA/cm^2^, 60 min) onto nitrocellulose membranes. Ponceau staining of membranes confirmed equal loading of proteins. Membranes were blocked with 5% skim milk and incubated overnight with primary rabbit antibodies directed against A~1~R (0.025%, Sigma-Aldrich, Munich, Germany), A~2A~R (0.1%, Millipore, Schwalbach, Germany), A~2B~R (0.1%, Millipore), A~3~R (0.2%, Millipore), and beta-actin for normalization (0.0025%, Sigma-Aldrich). The specificity of the antibodies has been documented by the manufacturers and was not tested again in this study. After extensive washing blots were incubated with a horseradish peroxidase-linked anti-rabbit IgG (60 min, 0.0001%, Santa Cruz Biotechnology, Santa Cruz, CA, USA). Immunoreactive bands were detected using an enhanced chemiluminescence system and were subsequently quantified with the AlphaEaseFC software (Alpha Innotech, San Leandro, CA, USA). The values thus obtained were corrected for different conditions between single runs, using a standard dilution series that was run on each gel. Results of adenosine receptors were then normalized to beta-actin. Statistical analysis {#s4e} -------------------- Data was analyzed with SPSS 17.0 (SSPS Inc., Chicago, IL, USA). The nonparametric Kruskal-Wallis and the Mann-Whitney-U test were used to detect significant differences between groups of interest. Mortality rates were estimated by the Kaplan-Meier method and compared by log-rank test. **Competing Interests:**The authors have read the journal\'s policy and have the following conflicts: Berthold Hocher was an employee of Solvay Pharmaceuticals at the time of the study. Dieter Ziegler and Yvan Fischer are employees of Abbott Products GmbH. Solvay Pharmaceuticals (now Abbott Products GmbH) financially supported the study. Abbott Products GmbH holds the patent of SLV329 and is developing this compound. This does not alter the authors\' adherence to all the PLoS ONE policies on sharing data and materials, as detailed online in the guide for authors. **Funding:**This study was partially supported by a grant of the Dr. Werner Jackstädt-Stiftung to Markus Alter. This funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. This study was also supported by Solvay Pharmaceuticals (now Abbott Products GmbH). Dieter Ziegler and Yvan Fischer are employees of Abbott Products GmbH and played a role in study design, data collection and analysis, decision to publish and preparation of the manuscript. The study design was set up together with Solvay Pharmaceuticals. Solvay Pharmaceuticals agreed to publish the manuscript. [^1]: Conceived and designed the experiments: BH SH MA PK DZ YF TP. Performed the experiments: SH KVW AMA JR TP. Analyzed the data: BH SH TP. Contributed reagents/materials/analysis tools: BH AMA DZ YV. Wrote the paper: TP.

PubMed Central

2024-06-05T04:04:19.901500

2011-3-10

PMC3053402

The authors would like to add additional funding. When Drs. Suhel Parvez and Julietta U. Frey were added to the author byline, their support from DFG and Alexander-von-Humboldt Foundation should have also been included in the financial disclosure for this article. The Funding section should read: \"This work was supported by NSFC (30870795, 30970920) and Scientific Innovation Projects from Fudan Univ. to TB and by the DFG (SFB779, TPB4 and TPB8 to JUF and MRK), the CBBS / Land Saxony-Anhalt / EU (C1-TP4 to MRK), DIP grant, the DZNE (to MRK), and the Schram Foundation (to MRK). SP was supported by the Alexander-von-Humboldt Foundation. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.\" **Competing Interests:**No competing interests declared.

PubMed Central

2024-06-05T04:04:19.906053

2011-3-01

PMC3053403

There is an update to the fifth author\'s affiliation. Stephen S.-T. Yau\'s affiliation is: \"Department of mathematical sciences, Tsinghua University, Beijing, P.R. China\" **Competing Interests:**No competing interests declared.

PubMed Central

2024-06-05T04:04:19.906431

2011-3-09

PMC3053404

Figures 2A and 2B were mislabeled; the terms \"Low\" and \"High\" should be reversed. **Competing Interests:**No competing interests declared.

PubMed Central

2024-06-05T04:04:19.907088

2011-3-04

PMC3053405

There were errors in the Author Contributions. The correct contributions are: Conceived and designed the study: ST. Performed the study and contributed reagents/materials/analysis tools: ST MB GB. Analyzed the data: ST. Wrote the paper: ST GB. **Competing Interests:**No competing interests declared.

PubMed Central

2024-06-05T04:04:19.907306

2011-3-09

PMC3053406

Figure 4 was replaced by Figure 3 in the published manuscript. Please view the correct Figure 4 here: ::: {#pone-ad8aa7d5-17c1-483d-8b69-610c8839bc3a-g001 .fig}  ::: **Competing Interests:**No competing interests declared.

PubMed Central

2024-06-05T04:04:19.907521

2011-3-10

PMC3053445

[^1]: Academic Editor: Sergi Ferre

PubMed Central

2024-06-05T04:04:19.907830

2009-1-18

PMC3053446

Classical repression by unliganded nuclear receptors {#sec0005} ==================================================== The first insights into how nuclear receptors repress transcription in the absence of their activating ligands were obtained through cloning of the corepressor proteins silencing mediator of retinoid and thyroid hormone receptor (SMRT) [@bib0005] and nuclear receptor corepressor (NCoR) [@bib0010] that interact with many unliganded nuclear receptors. It is now known that corepressor proteins exist as large multiprotein complexes containing enzymes such as histone deacetylases that repress transcription by regulating the nature of the chromatin local to the gene promoter. Corepressors are usually released on ligand binding through changes in the conformation and dynamics of the receptor ligand-binding domain, which favour recruitment of coactivator proteins ([Figure 1](#fig0005){ref-type="fig"}) [@bib0015]. An important feature of these interactions is that corepressors and coactivators bind to overlapping surfaces on the nuclear receptors such that their binding is mutually exclusive. This is important because competition between coregulator proteins can play a key role in the tissue-specific regulation of some target genes [@bib0020]. Although the classic activity of most ligand-bound nuclear receptors activates transcription, ligand-bound nuclear receptors repress the transcription of certain genes, a process termed negative regulation. The classic example of this is repression of thyroid-stimulating hormone (TSHα) by the thyroid receptor (TR) in the presence of T3. It is now known that such negative regulation occurs with many other nuclear receptors and genes. Over the years, various mechanisms have been suggested to explain how ligand-dependent repression occurs ([Figure 1](#fig0005){ref-type="fig"}). Here we discuss these different mechanisms and examine how they can be reconciled with recent genome-wide studies. Negative response elements {#sec0010} ========================== A relatively long-standing idea in the nuclear receptor field is the concept of negative response elements. This concept emerged when it became clear that ligand-bound glucocorticoid (GR) and TR receptors downregulate specific target genes, which raised the question as to what mechanism might explain how ligands could have opposite transcriptional effects on certain genes. Studies of GR downregulation of the prolactin gene revealed that ligand-bound GR, associated with a negative glucocorticoid response element (nGRE), acts through the 'reversal of a constitutive enhancer activity' [@bib0025]. The authors speculated that when ligand-bound GR binds to nGREs, its conformation does 'not support' transcriptional activation. However, the molecular mechanism underlying this phenomenon remained uncertain. Subsequent studies investigating downregulation of the genes encoding TSHα [@bib0030] and TSHβ [@bib0035] identified a negative thyroid hormone response element (nTRE) in the proximal promoter, between the TATA box and the transcriptional start site of the gene. It was proposed that downregulation was the result of steric interference with other components of the transcriptional machinery. Intriguingly, examination of the DNA sequence of both nGRE and nTRE suggested that the sequences are significantly divergent from the classical positive response elements that mediate transcriptional activation by these receptors, which brought into doubt whether or not receptors actually bind to negative response elements. This issue was resolved using a TRβ mutant defective in DNA binding, which was unable to suppress expression of *TSH* genes in response to T3 in both cell-based assays [@bib0040] and knock-in mice [@bib0045]. Thus, these findings demonstrated that DNA-binding activity is necessary to downregulate *TSH* genes. Support for the idea that negative response elements function through promoter interference mechanisms came from further studies of the prolactin and β-amyloid precursor genes. It was shown that a GR binding site (nGRE) in the prolactin promoter overlaps with the binding sites for other transcription factors, including Oct-1 and Pbx. Addition of an isolated GR DNA-binding domain to the nuclear extract precluded DNA binding of both Oct-1 and Pbx [@bib0050], which implies that direct competition exists for DNA-binding on this promoter. Analogously, the binding sites for TR and the transcription factor SP1 overlap each other in the β-amyloid precursor protein (APP) promoter [@bib0055]. T3 binding to TR increases the DNA-binding affinity of TR, preventing SP1--DNA complex formation and consequently downregulating SP1-dependent expression of APP. Together, these findings suggest an interplay between competing transcriptional regulators on certain promoters. Consequently, ligand-bound nuclear receptors can block transcriptional activation by other factors, which leads to downregulation of these genes. Role reversal of coregulators {#sec0015} ============================= Since the early studies of negative response elements, we have learnt that nuclear receptors (like most transcription factors) regulate gene expression through the recruitment of large complexes containing a variety of coregulators and effector enzymes that target chromatin and other factors. As discussed, in the classical activity of nuclear receptors, ligand-bound receptors recruit coactivator complexes and unliganded receptors bind corepressor complexes. This of course raises the question as to what type of coregulator complex is recruited by liganded receptors on a negatively regulated promoter. This is a difficult problem because multiple studies have explained how ligand binding to nuclear receptors promotes interaction with coactivator complexes containing histone acetylases (and methyl transferases) and displaces corepressor complexes associated with histone deacetylases (and demethylases). So what explains downregulation by ligand-bound receptors? Do they somehow recruit corepressors on negatively regulated genes? Or do histone acetylases somehow repress these genes? This conundrum is illustrated by the cartoon in [Figure 2](#fig0010){ref-type="fig"}. An early study of negative regulation of the gene encoding TSHα provided evidence that the role of coregulators might be reversed on negatively regulated genes, because recruitment of corepressors to the gene encoding TSHα was associated with activation [@bib0060]. The reversal of action seems to lie in the finding that corepressor recruitment to this gene results in histone acetylation. Similarly, the corepressor SMRT mediates an increase in transcription at the nTRE within the Rous sarcoma virus long terminal repeat of TSHα [@bib0065]. In this case, protease digestion and mobility shift assays suggested that the TR--SMRT complexes had different conformations depending on whether they were bound to a negative or positive hormone response element. Further studies lend support to the concept of coregulator role reversal depending on the particular response element. TR mutants defective in corepressor recruitment no longer activate an nTRE present in the *SOD1* promoter. Conversely, a receptor defective in coactivator recruitment, but still able to interact with corepressors, shows impaired downregulation in response to thyroid hormone [@bib0070 bib0075]. The role of coactivators in mediating repression is supported by several studies. Mice lacking the steroid receptor coactivator-1 (SRC1) revealed a role for this coactivator in activating some liganded TR-responsive genes and also repressing transcription from liganded and unliganded TR-responsive genes [@bib0080 bib0085]. Similarly, the coactivator SRC3 functions as a repressor in lymphocytes [@bib0090]. Taken together, these findings suggest that both promoter and cellular context can determine whether a particular coregulator acts as an activator or a repressor. Two aspects of these role reversals remain unclear. First, what is it about a particular promoter element or cellular environment that results in coregulator role reversal? Second, how are such role reversals implemented by the coregulator complexes? Evidence that the sequence of DNA response elements can influence transcriptional outcome came from studies of various GR binding sites that seem to require or exploit different activation domains within the receptor [@bib0095]. Indeed, a single base-pair change in the DNA of the response element influences GR conformation [@bib0100]. Thus, it seems that DNA serves as a sequence-specific allosteric ligand that modulates the regulatory activity of the GR. The histone demethylase LSD1 illustrates a well-established example of the mechanism through which the role of a coregulator can be reversed. LSD1 normally acts as a corepressor when recruited to chromatin as part of the CoREST complex. However, LSD1 can act as a coactivator of the androgen receptor [@bib0105]. The mechanism for this switch in activity has recently been established. When acting as a corepressor, LSD1 demethylates lysine 4 on histone 3 (H3K4). When histone 3 is phosphorylated (H3T6) by PKCβ kinase, which is recruited by the androgen receptor, then LSD1 demethylates H3K9 and leaves methyl groups on H3K4, which leads to activation of transcription [@bib0110]. This illustrates how a simple post-translational modification can lead to reversal of transcriptional activity. It seems likely that covalent modifications, as well as differential splicing of coregulators, will explain many examples of coregulator role reversal. In addition to a large repertoire of splice variants, coregulator proteins are extensively modified by acetylation, phosphorylation, ubiquitination, methylation and SUMOylation [@bib0115 bib0120 bib0125 bib0130]. Together, these variations give enormous scope for fine-tuning of the transcriptional outcome. Inverse recruitment of corepressors {#sec0020} =================================== Role reversal of coregulators (e.g. a corepressor acting as a coactivator) is now well established. However, it was recently shown that another mechanism plays a role in negative regulation. In this case, it seems that a ligand-bound TR can repress genes by recruiting the corepressor NCoR. This can be thought of as inverse recruitment of coregulators. The evidence for this arises from a recent study in mice containing a mutant corepressor, NCoR, harbouring mutations in the deacetylase activation domain (DAD), which abrogate interaction with the deacetylase HDAC3. In these mice, several TH-responsive genes are modestly activated in the absence of TH, which suggests that failure to recruit HDAC3 leads to a failure of normal gene repression. However, more surprisingly, several genes that are normally repressed by ligand-bound TH are activated in the mutant mice [@bib0135]. This suggests that on positively regulated genes, NCoR is displaced on ligand binding to the TR, which allows recruitment of coactivators, but on the negatively regulated TSHα promoter, NCoR is recruited to the ligand-bound TR, which leads to transcriptional repression. It remains to be understood through what mechanism this inverse recruitment of corepressors on negatively regulated genes might be achieved. Inverse coregulators recruited by normally activating ligands {#sec0025} ============================================================= The concept of coregulator role reversal derives from the finding that coregulators that normally bring about one outcome, can in certain circumstances, bring about the opposite outcome. At least one coregulator has been identified that seems to be the extreme example of role reversal. Receptor interacting protein of 140 kDa (RIP140) is a coregulator with multiple interaction motifs that allow it to be recruited to ligand-bound receptor [@bib0140]. However, RIP140 acts as a corepressor protein that recruits histone deacetylase enzymes through several complexes [@bib0145]. Thus, RIP140 can be considered an inverse coregulator. Its biological role seems to be regulation of metabolism by balancing the activities of the conventional coactivator PGC1α [@bib0150]. Similarly, LCoR is a corepressor that is recruited to ligand-bound oestrogen receptor α (ERα). Like RIP140, LCoR recruits other repressor molecules such as the C-terminal binding protein and histone deacetylases [@bib0155]. Inverse agonists promote corepressor recruitment {#sec0030} ================================================ Nuclear receptors are important targets for pharmaceutical intervention and there has been much effort to develop agonists to activate the receptor and antagonists to compete with and block the activation activity of the natural ligand. Importantly, a third type of synthetic ligand for nuclear receptors has emerged, inverse agonists. These pharmaceutical ligands bind to nuclear receptors and promote the recruitment of corepressor complexes, which leads to active repression of target gene transcription in response to ligands. One example of an effective pharmacological inverse agonist is tamoxifen, which binds to the oestrogen receptor, promotes recruitment of corepressors such as NCoR, and represses oestrogen receptor target genes. It has only been relatively recently recognized that naturally occurring ligands can act as inverse agonists. REV-ERBα and β were originally described as orphan receptors that seemed to function as constitutive transcriptional repressors, directly binding to DNA and recruiting the NCoR corepressor [@bib0160]. The lack of activation was explained by sequence analyses that suggested that they lack the carboxy-terminal helix of the ligand-binding domain, which is essential for coactivator recruitment by other nuclear receptors. Recent studies have revealed that haem serves as a regulatory ligand for REV-ERB. However, haem binding does not activate the receptor, but instead increases the affinity for the NCoR corepressor, which in turn enhances the repression activity [@bib0165]. Like REV-ERB, no regulatory ligand for RORβ had been identified. However, the crystal structure of RORβ revealed a fatty acid ligand (stearate) in the ligand-binding pocket. Surprisingly, the steric acid ligand does not seem to activate or antagonize RORβ transcriptional activity [@bib0170]. More recently, however, it was found that RORβ binds all-*trans* retinoid acid, which downregulates the transcriptional activity of RORβ [@bib0175]. Thus, it seems that retinoic acid might act as an inverse agonist of RORβ. Another receptor that seems to be regulated by a natural inverse agonist is the constitutive androstane receptor (CAR), a xenobiotic-responsive transcription factor. This receptor, as suggested by its name, has strong transcriptional activity in the absence of ligand. This constitutive activity is explained by some structural differences that lock the receptor in an active conformation [@bib0180]. Some years ago, it was shown that the androstane metabolite (androstanol) acts as an inverse agonist and reverses the constitutive activity of CAR [@bib0185]. The crystal structure of CAR with androstenol shows that this inverse agonist binds within the ligand-binding pocket and locks the receptor in an inactive conformation that does not support coactivator binding, but instead allows the recruitment of corepressors [@bib0190 bib0195]. Indirect repression by ligand-bound nuclear receptors {#sec0035} ===================================================== Nuclear receptors often engage in crosstalk with other transcriptional regulators, and in many cases this results in repression of the activity of other factors. This is termed *trans*-repression and is commonly dependent on ligand binding to the nuclear receptor. One of the earliest examples of this involved an interaction between the ER and the tissue-specific transcription factor Pit-1, which leads to downregulation of the prolactin gene [@bib0200]. Later, using different *in vivo* strategies, a model was proposed for such *trans*-repression by GR that involves tethering of the receptor by direct binding to other transcription factors such as NF-κB [@bib0205] and AP1 [@bib0210 bib0215]. More recently, it was demonstrated for AP1 transcription factors that only dimers containing FOS are *trans*-repressed by GR. Thus, the dimer composition of AP1 can regulate the positive and negative transcriptional activity [@bib0220]. In this type of *trans*-repression mechanism, it seems that the nuclear receptor itself does not bind directly to DNA and indeed in some cases the DBD is not needed for *trans*-repression to occur [@bib0200 bib0210]. However, in other cases the nuclear receptor DBD does seem to be required, which suggests that the DBD plays a role in the interaction with other transcription factors [@bib0225 bib0230 bib0235]. Recent experiments on negative regulation of the gene encoding TSHβ suggest that crosstalk occurs between ligand-bound TR and the transcription factor GATA2 [@bib0240]. The Zn-finger region of GATA2 interacts with the TR DBD, and this complex is required for negative regulation by thyroid hormone. In this case, the effect of the ligand perhaps controls the differential affinity of TR for the GATA2-RE and the negative regulatory element [@bib0245]. Another important example of *trans*-repression is mediated by PPARγ. Ligand-dependent SUMOylation of the PPARγ ligand-binding domain results in PPARγ recruitment of the corepressor (NCoR)--histone deacetylase-3 (HDAC3) complex to inflammatory gene promoters [@bib0250]. Insights from genome-wide studies {#sec0040} ================================= Genome-wide studies of transcription have dramatically changed our view of transcriptional repression by nuclear receptors. Although we used to believe that downregulation of genes in response to nuclear receptor ligands was a relatively minor affair, it turns out that as many as half the genes regulated by nuclear receptor ligands are in fact downregulated [@bib0255]. Clearly in some cases this will be a secondary effect of upregulation of another factor such as a repressive transcription factor or even a corepressor protein. However, it seems that much of the downregulation by nuclear receptor ligands occurs on the same time scale as upregulation, which argues against secondary effects being responsible for this downregulation. How can this large-scale downregulation be explained? Microarray-based gene-expression profiling experiments have identified genes that are either up- or downregulated by nuclear receptors. Analysis of these experiments together with ChIP-on-chip and ChIP-Seq experiments revealed the location of the binding sites for nuclear receptors. What has emerged from these studies is that the majority of upregulated genes, both for the ER [@bib0260] and GR [@bib0265], are associated with binding sites for the receptor. In stark contrast, few of the downregulated genes seem to be located in realistic proximity to binding sites for the receptors. Thus, it seems that much of the downregulation observed occurs without interaction between the nuclear receptor and a promoter or enhancer sequence close to the regulated gene. To explain this, we must give some thought to the inherent transcriptional potential of the cell at any one time. If we assume that there is a maximum amount of transcription a cell can perform, perhaps through limited concentrations of certain components of the transcriptional machinery, then a sudden upregulation of one set of genes in response to an activating ligand will necessarily result in downregulation of other genes. This phenomenon is well known in the transcription field as squelching and is commonly observed when a transcriptional regulator is overexpressed [@bib0270]. The excess non-DNA-bound material titrates out other components of the transcriptional machinery and hence causes downregulation of other genes because the squelched components become limiting. If squelching is really the explanation for this large-scale downregulation of gene expression in response to activating nuclear receptor ligands, we must ask whether this downregulation is physiologically important. A recent puzzle has emerged through mapping of the genome-wide locations of histone acetyltransferases (HATs) and HDACs. These genome-wide studies revealed that not only are HATs associated with actively transcribed genes, but HDACs (well-known components of repression complexes) are also found almost exclusively at active genes [@bib0275]. This suggests that we need to completely reconsider the roles of the coregulators. One explanation, of course, is that active genes are associated with acetylated histones. Acetylated histones are substrates for HDACs, so perhaps it is natural that HDACs should be at active genes. Indeed, a study examining ERα-induced expression of the gene encoding PS2 revealed that transcriptional activation is a cyclical process in which recruitment of ERα and proteins involved in activation and repression cycle approximately every 40 min [@bib0280]. Acetylation, deacetylation, methylation and demethylation of histones H3 and H4 also followed a cyclical pattern. These findings suggest that corepressor recruitment might be an essential part of transcriptional activation by serving to reset the transcriptional machinery. Perspectives {#sec0045} ============ Over the last 25 years of studies of transcription regulation by nuclear receptors, many overarching general principles have emerged that have stood the test of time. These include the concepts of: (i) DNA response element recruitment of receptors to promoters or enhancers of target genes; (ii) binding of ligands to nuclear receptors to control the recruitment of coregulator complexes; and (iii) implementation by coregulator complexes containing histone-modifying enzymes of a histone code that directs transcriptional activity of target genes by modifying the structure of chromatin. However, as the details have been explored, it has emerged that each of these broad concepts encompasses an enormous range of diversity with multifunctional complexes and outcomes. At one time, downregulation of genes in response to nuclear receptor ligands seemed to conflict directly with the established principles. Now, set in the context of the widely diverse details of gene regulation, this downregulation no longer seems so surprising, nor should it be unexpected that many different mechanisms contribute to this ligand-dependent downregulation. Downregulation : reduction in the rate of transcription to a level below that which would otherwise be observed. The opposite is termed upregulation. Impairment of activation : prevention of or reduction in the activation of transcription by other factors. Inverse regulation : activation of transcription by nuclear receptors in the absence of ligand, and downregulation in the presence of ligand. Negatively regulated genes : genes that are downregulated under conditions that would normally be expected to increase their rate of transcription. Repression : reduction in the rate of transcription to a level lower than basal through the recruitment of repressive factors or complexes. Squelching : indirect interference in the transcription of a gene through the sequestering of factors normally required for its transcription. This can result in down- or upregulation of the gene in question. We would like to thank Jo Westmoreland ([Figure 1](#fig0005){ref-type="fig"}) and Bira Dantas ([Figure 2](#fig0010){ref-type="fig"}) for artwork. LF and JWRS are supported by the Wellcome Trust (085408). ::: {#fig0005 .fig} Figure 1 ::: {.caption} ###### Diverse modes of signalling by nuclear receptors. (**a**) Classical nuclear receptor (NR) signalling. **(i)** Transcriptional repression through recruitment of corepressor complexes to unliganded nuclear receptors. **(ii)** Transcriptional activation through recruitment of coactivators to ligand-bound nuclear receptors. (**b**) Ligand-dependent repression of transcription. **(i)** Negative response element: direct recruitment to DNA. (1) Nuclear receptor interference with the activation of transcription by other factors. For example, ligand-bound TR prevents the general transcriptional activator SP1 from binding to the β-amyloid precursor gene. (2) Coactivator role reversal leading to transcriptional repression. For example, SRC1 contributes to repression by ligand-bound TR. (3) Inverse recruitment of corepressors to ligand-bound receptors. For example, the corepressor NCoR has been implicated in association with ligand-bound TR on the gene encoding TSHα. (4) Factors such as RIP140 act as inverse regulators because they serve as corepressors yet are recruited to ligand-bound receptors. (5) Synthetic and natural inverse agonists serve as negative ligands because they promote recruitment with corepressor complexes. For example, haem-binding promotes repression by REV-ERB. **(ii)***Trans*-repression by ligand-bound receptors. (6) Ligand-bound GR interacts with and prevents activation of AP1-mediated transcription. (7) Ligand-bound TR contributes to repression of the gene encoding TSHβ through interaction with the GATA2 transcription factor. **(iii)** Downregulation through off-DNA mechanisms. (8) Genome-wide studies suggest that many downregulated genes do not directly recruit nuclear receptors. Hence, the downregulation observed is likely to be due to squelching effects. :::  ::: ::: {#fig0010 .fig} Figure 2 ::: {.caption} ###### The conundrum of downregulation by ligand-bound nuclear receptors. When ligand-bound nuclear receptors negatively regulate target genes, many of the classical principles of nuclear receptor signalling are reversed. :::  :::

PubMed Central

2024-06-05T04:04:19.908075

2011-3-1

PMC3053447

Cell-to-cell spread of retroviruses {#s0005} =================================== Retroviruses are a diverse family of enveloped RNA viruses that encompass a number of medically important human pathogens including the Human Immunodeficiency Virus (HIV), which alone has accounted for approximately 25 million deaths worldwide. Over the past decade huge scientific and medical endeavour has been focussed towards understanding the biology of viral pathogenesis and transmission between and within hosts. Like a number of other mammalian viruses, retroviruses can disseminate between susceptible cells either by cell-free infection or by direct cell-to-cell spread (reviewed in ([@bb0460])). Retroviruses spread directly between cells by taking advantage of their immunotropic properties to infect CD4^+^ T cells, macrophages and dendritic cells that inherently form intimate, dynamic and transient contacts ([@bb0245]). In this way, retroviruses can co-opt specialized properties of immune cells that normally operate during intercellular communication such as antigen presentation and T cell activation to promote their dissemination between cells. Direct cell-to-cell spread of the human retroviruses HIV type-1 (HIV-1) and HTLV-1 (Human T-lymphotropic Virus Type-1) predominantly takes place at specialized contact-induced structures known as virological synapses (VS) that act as "hot-spots" for virus transmission ([@bb0205 bb0230 bb0260 bb0355]). VS were so named because of their resemblance to immunological synapses (IS) and the term VS was coined to describe a specific membrane receptor architecture that evolves following intimate contact between a HIV-infected T cell and an uninfected target T cell ([@bb0260]). Cell-to-cell spread of HIV-1 at synapses is a generalized feature of viral dissemination and VS have been described between infected and uninfected CD4^+^ T cells ([@bb0230]), between macrophages and CD4^+^ T cells ([@bb0140 bb0155]) and between virus-exposed dendritic cells and CD4^+^ T cells ([@bb0355]). This phenomenon is not restricted to HIV-1, and one of the first VS described was that of HTLV-1 ([@bb0205]). Longer-range intercellular transmission of HIV-1 between T cells has also been observed along cellular projections known as membrane nanotubes ([@bb0500]), while the related retrovirus murine leukeamia virus (MLV) utilizes virus-induced filopodia for efficient dissemination ([@bb0485]). The relative contribution of cell-to-cell spread at VS, membrane nanotubes and via cell-free infection is difficult to quantify, but *in vitro* culture systems have demonstrated that cell-cell spread is the predominant mode of HIV-1 dissemination and this is mostly via VS ([@bb0495]). At present there is considerable effort towards delineating retroviral protein trafficking in infected cells during cell-to-cell spread and understanding the molecular regulators of transmission in both donor and target cells ([@bb0015 bb0060 bb0140 bb0155 bb0195 bb0445 bb0550 bb0205 bb0355 bb0020 bb0230 bb0235 bb0240 bb0250 bb0255 bb0310 bb0400 bb0405]) and readers are directed to a recent series of comprehensive reviews that consider this in detail ([@bb0105 bb0220 bb0350 bb0375 bb0395 bb0465 bb0575]). In the context of viral pathogenesis, direct cell-to-cell transmission is likely to confer a number of advantages for retrovirus compared to classical cell-free infection. Firstly, cell-to-cell spread increases infection kinetics by directing virus assembly and budding to sites of cell-to-cell contact and may be one or more orders of magnitude more efficient than equivalent cell-free infection ([@bb0085 bb0345 bb0060 bb0335 bb0455 bb0495]). This is achieved by obviating the rate-limiting step of extracellular diffusion that is required of cell-free virus to find a susceptible target cell. Furthermore, polarizing virus budding towards sites of cell-to-cell contact at which viral entry receptors are clustered increases the number of potentially productive transmission events and increases the likelihood of productive infection. Secondly, it has been hypothesised that cell-to-cell spread of retroviruses could provide a replicative advantage to the virus by limiting exposure of particles to neutralizing antibodies ([@bb0330]). It has generally been assumed that cell-to-cell spread of retroviruses at VS might allow escape from neutralizing antibodies either by limiting the window of opportunity for antibody to engage viral antigens, or by providing a relatively protected domain at cell-to-cell interfaces that could physically exclude the relative bulk of antibodies from gaining access to virions before they attach and enter to target cells. Whether VS protect retroviruses from humoral immunity is still unclear and there are conflicting reports on this in the literature ([@bb0060 bb0115 bb0335 bb0340]). Possible explanations for disparate results have been considered elsewhere ([@bb0465]) and so will not be elaborated in detail here. Humoral immunity to human retroviruses such as HIV-1, the causative agent of Acquired Immune Deficiency Syndrome is of particular interest within the context of cell-to-cell spread because of the implications of immune evasion for vaccine design and viral pathogenesis. The innate immune response is intimately linked to the generation of an effective adaptive immune response. Thus retroviral-induced innate immune responses may have a direct impact on cell-to-cell transmission but may also modulate adaptive immunity and thereby control of viral infection. The role of innate immunity during cell-to-cell spread of retroviruses has only recently been explored; however, it is increasingly apparent that harnessing innate immunity might provide a crucial opportunity to tackle HIV-1 at some of the earliest steps of infection, and that the interplay between HIV-1 and innate immunity has important implications for disease pathogenesis ([@bb0040]). In the context of HIV-1 cell-to-cell spread the balance between viral suppression and enhancement by innate immune responses is intriguing, although relatively little studied. Here I will discuss some recent insights into cell-to-cell spread and innate immunity and consider how the interplay between HIV-1 and innate effectors may modulate cell-to-cell dissemination. I will focus predominantly on HIV-1, but it is likely that some aspects of innate immunity and HIV-1 will be applicable to other retroviruses. Recognition of HIV-1 by innate immune receptors during cell-to-cell spread {#s0010} ========================================================================== At the earliest time points after infection, before adaptive immunity has been activated, the innate immune system provides the first line of antiviral defences and alerts the wider immune system of challenge. An important feature of innate immunity that facilitates such rapid response is the recognition of generalized pathogen-associated molecular patterns (PAMPS). This is mediated by via a range of pattern recognition receptors (PRR) including C-type lectins (CLR), Toll-like receptor (TLRs) and cytosolic sensors such as NOD-like receptors and the retinoid acid-inducible gene (RIG) like receptors RIG-I and MDA5. Recognition of ubiquitous microbial patterns leads to signal transduction, activation of the transcription factors such as NF-κB, mitogen-activated protein kinase (MAPK) and interferon regulatory factor (IRFs), and culminates in secretion of proinflammatory and immunomodulatory cytokines such as type-1 interferons (interferon-α and interferon-β). TLRs are located at the cell surface or in endocytic compartments and collectively recognize a range of viral and bacterial ligands including hydrophobic molecules, glycoproteins, bacterial cell wall components and nucleic acid, the latter being a particularly potent activator. To date, 10 different TLRs have been identified in humans. In addition, other receptors such as C-type lectins and scavenger receptors on cell surfaces can act as TLR coreceptors and bind to microbes via PAMPs which culminates in a signaling cascade that alerts the wider immune system of danger. In the context of HIV-1, a number of steps in the viral life cycle have been shown to activate immunity via PRR recognition including attachment of the HIV-1 envelope glycoprotein (Env) subunit gp120 by the C-type lectin DC-SIGN (Dendritic Cell-Specific Intercellular adhesion molecule-3-Grabbing Non-integrin) ([@bb0150]); TLR7/8-mediated detection of HIV-1 RNA ([@bb0025 bb0175 bb0360]) and more recently the identification of an intrinsic dendritic cell sensor that detects the interaction between newly synthesized HIV-1 capsid and cylophilin A and activates the transcription factor IRF3 ([@bb0315]). Interestingly, cell-free HIV-1 infection may escape innate immune detection in some situations and it has been proposed that that macrophages lack a functional PRR for HIV-1 therefore attenuating NF-κB and IRF3 activation and type-1 interferon induction ([@bb0410 bb0540]). So far, most studies examining innate immune recognition of HIV-1 have utilized cell-free virus or viral constituents, and characterized their effects on dendritic cells and macrophages. Therefore, it is unclear whether infection during cell-to-cell contact will trigger the same innate immune sensors as cell-free infection or whether it bypasses or activates different checkpoints for innate immune activation. It is reasonable to assume that infection may have diverse consequences for the cell depending upon whether viral transmission was mediated by cell-free or cell-to-cell spread. For example, it has been suggested that HIV-1 entry during cell-to-cell contact may involve endocytosis ([@bb0035 bb0060 bb0195]) rather than fusion at the plasma membrane. Although the concept of productive infection via endocytosis remains controversial, if it is correct then the use of different modes of virus entry (e.g., fusion at the plasma membrane vs. endocytosis) may mean that viral constituents could be differentially presented or protected from innate immune receptors. Furthermore, polarization of viral egress towards target cells during cell-to-cell spread increases the amount of viral protein and nucleic acid that enters the target cell that may in turn increase viral antigen above a critical threshold to trigger an innate response. Notably, the Greene lab have very recently reported that the accumulation of abortive reverse transcription intermediates in resting CD4^+^ T cells following contact with infected cells activates proinflammatory and apoptotic host defences by the persistent exposure to cytoplasmic DNA, resulting in death of these non-productively infected cells ([@bb0090]). Intriguingly, it was noted that indirect killing was dependent on close cell-to-cell contact and was mediated by transmitting HIV-1 virions but not cell-associated Env ([@bb0090]). This effect could be recapitulated by spinoculation of cell-free virus onto cells, suggesting that the quantity of virus transferred to resting T cells during cell-to-cell contact results in an accumulation of defective cytoplasmic viral DNA triggering an IRF-3-dependent innate immune response by a TLR-independent mechanism ([@bb0510]). It will be informative to determine to what extent cell-to-cell spread of HIV-1 results in cell death rather than productive infection in activated and resting T cells when virus is transmitted across the different types of VS and how this is regulated. For example mRNA encoding TLR 1, 2, 3, 4 5, 7 and 9 have been detected in human CD4^+^ T cells ([@bb0190 bb0590]). TLR activation can functionally result in cytokine secretion from T cells but we do not yet know if cell-to-cell spread triggers TLR recognition in CD4^+^ T cells. It is also unclear if dendritic cells and macrophages can serve as target cells rather than donor cells during cell-to-cell spread. DCs and macrophages efficiently transmit infectious HIV-1 *to* susceptible CD4^+^ T cells, but whether infected T cells can transmit virus *back* to DCs and macrophages, is unknown. This is an important question since it is this lineage of cell, rather than T lymphocytes, that dominate innate immune type-1 interferon secretion *in vivo*. It is clear that more work is needed in this area to address these and other questions. Modulating innate immunity to promote cell-to-cell spread of HIV-1 {#s0015} ================================================================== As the first line of the cellular innate response to infection, patrolling sentinel cells such as plasmocytoid and myeloid dendritic cells (pDC and mDC, respectively), Langherhan cells and macrophages are all poised to detect and engage foreign invaders. Paradoxically, some of these cells are also among the earliest targets for HIV-1 *in vivo.* Whether cell-to-cell spread triggers or suppresses antiviral responses, either virus-induced or coincidental, that the virus might use to its own advantage is an interesting proposition and number of recent studies have started to explore how HIV-1 may modulate innate immunity specifically during cell-to-cell spread. Mature myeloid DCs in particular are thought to play a key role in HIV-1 transmission between hosts by capturing incoming virions at mucosal surfaces and disseminating virus directly to CD4^+^ T cells by cell-to-cell transmission, thereby allowing HIV-1 to take advantage of the inherent response of activated mDCs to migrate to secondary lymphoid organs and interact intimately with T cells. DCs do not usually become productively infected with HIV-1 even *in vitro*, although they do express the appropriate HIV-1 entry receptors (CD4 and a chemokine co-receptor) and it generally held that productive infection by HIV-1 is a property of immature mDCs ([@bb0550]). As a consequence, dissemination of HIV-1 by mature mDCs to T cells is considered to mostly occur *in trans* - a process by which HIV-1 virions are captured by receptors expressed on the surface of mDCs and transferred directly to T cells during intimate cell-to-cell contact. This occurs by capture of HIV-1 particles by DC-SIGN present on the surface of DCs ([@bb0120]) that recognise moieties on the Env subunit gp120, although other receptors can also mediate *trans*-infection ([@bb0070 bb0295 bb0545]). Once the HIV-1-DC-SIGN complex is formed it may remain at the plasma membrane or become internalized into a partially protective endocytic compartment (thus avoiding complete degradation of infectious virus) and subsequently trafficked to the cell-to-cell junctions during VS formation ([@bb0120 bb0350 bb0355]). By contrast, HIV-1 capture by Langerhans cells expressing the receptor Langerin results in degradation of virus ([@bb0075]). To date, most work has focussed on conventional mDCs and it is not clear whether pDC can also capture HIV-1 and transmit virus to T cells and this is no doubt complicated by the fact that pDC are more difficult to work with and comprise only 1% of peripheral blood mononuclear cells (PBMCs); however, there are reports that the related retrovirus HTLV-1 can infect pDCs ([@bb0065 bb0180]) and so parallels to HIV-1 may exist. DC-SIGN is a PRR and a C-type lectin that normally binds carbohydrate-containing ligands and initiates a response to foreign antigen via activating Toll-like receptors ([@bb0440]). Insight into how HIV-1 may modulate signaling via DC-SIGN to promote cell-to-cell spread has come from recent studies investigating the DC-SIGN signalosome. Using gene expression profiling and phosphoproteomics, Hodges et al., observed that HIV-1 interaction with DC-SIGN triggers a signaling pathway leading to activation of LARG and increased RhoA-GTPAse activity and that this is necessary for efficient DC-T cell VS formation and cell-to-cell spread, possibly by RhoA modulation of exocytosis from DCs or regulation of actin dynamics ([@bb0185]). HIV-1 induced DC-SIGN activation also synergises with TLR8 activation by HIV-1 ssRNA for recruitment of transcription factors required for full-length viral transcript synthesis under conditions where DCs do become productively infected and transmit virus to T cells *in cis* ([@bb0150]). In these ways, HIV-1 can take advantage of binding to a PRR to direct downstream signaling events that favour cell-to-cell spread. How does HIV-1 avoid degradation when internalized into DCs and transmit efficiently to T cells? Recent evidence suggests that a contributing factor maybe the ability of HIV-1 to down-regulate the autophagy pathway in cells. Autophagy is a specialized lysosomal degradation pathway of self-digestion that is necessary for correct antigen processing and presentation by MHC class II and for the delivery of TLR ligands to endosomes for innate immune activation ([@bb0570]) and there is evidence that this pathway can be hijacked by viruses to promote pathogenesis ([@bb0285]). HIV-1 gp120 binding to CD4 on DCs has been shown to down-regulate autophagy in DCs by mTor activation and regulate cell-cell spread *in trans* ([@bb0030]). Inducing autophagy with rapamycin also inhibits DC-T cell transmission by virus-pulsed DCs ([@bb0030]) suggesting that regulating the autophagy pathway may play a key role in early HIV-1 spread by diverting virus from antigen processing pathways and allowing infectious virus to remain within the DC for subsequent delivery to the VS. Cell-free HIV-1 infection of macrophages also inhibits rapamycin-induced autophagy ([@bb0560]) and increases HIV-1 cell-free yield ([@bb0100 bb0290]); however whether this affects cell-to-cell spread specifically was not investigated. Moreover, HIV-1 also down-regulates autophagy in CD4^+^ T cells during productive infection. By contrast to DCs however, CD4-Env binding on T cells was found to activate autophagy, but this was overridden by virus infection although the effect on cell-to-cell spread was not elucidated ([@bb0100]). In addition to potentially directly affecting infectious virus trafficking in cells, another consequence of loss of autophagosome is seen in HIV-1 exposed DCs that display altered TLR4 and TLR8 responses ([@bb0030]). The direct association of HIV-1 with DC-SIGN also has consequences for TLR4 signaling leading to increased expression of IL-6, IL-10 and IL-12 with presumed downstream effects on Th differentiation ([@bb0145]) and possible consequences for viral dissemination. There is also evidence that HIV-1 limits DC maturation with consequences for CD4^+^ T cell proliferation, cytokine secretion and adaptive immunity ([@bb0270]). Thus it appears that the ability of HIV-1 to modify the cellular autophagy pathway in immune cells and thus avert innate and adaptive immunity may be at the heart of efficient cell-to-cell spread and dissemination of HIV-1 from mucosal surfaces, thereby allowing the virus to establish a foothold during early transmission and contributing to subsequent spread between target cells and cellular reservoirs. Cell-to-cell spread and interferon-induced, antiviral restriction factors {#s0020} ========================================================================= The induction of type-1 interferon upregulates expression of a large number of interferon inducible genes (ISGs), some of which encode proteins with direct antiviral properties. One of the most important of these within the context of retroviral dissemination is a group of proteins with potent antiviral properties known collectively as "restriction factors". Restriction factors are cellular proteins that are constitutively expressed or induced by type-1 interferon and are able to limit viral replication by targeting specific steps of the retroviral viral life cycle (reviewed in ([@bb0580])) rendering cells less permissive or non-permissive to infection. In the context of human retrovirus infection three major restriction factors have now been described -- APOBEC3G/F (apolipoprotein B mRNA editing complex catalytic subunit) ([@bb0475]); TRIM5 (Tripartite motif-containing protein 5) ([@bb0525]) and tetherin (also known as BST-2, CD317 and HM1.24) ([@bb0390]). The importance of innate restriction factors to infection is highlighted by the fact that lentiviruses contain genes encoding for accessory proteins specifically to antagonize restriction - Vif that inhibits APOBEC3-mediated cytidine deamination of viral transcripts, and in the case of HIV-1 Vpu that overcomes tetherin-mediated inhibition of nascent particle release from the plasma membrane of virus-producing cells ([Fig. 1](#f0005){ref-type="fig"}). Restriction factors are of particular interest to retroviral pathogenesis because they exert such potent inhibition of viral replication and are upregulated in some cell types by interferon, suggesting that they form part of the innate antiviral defence against viral challenge. This has raised the possibility of harnessing innate immunity and type-1 interferon induction to upregulate endogenous restriction factor expression *in vivo*, or possibly using gene-therapy to introduce restriction factors from another species that may be resistant to antagonism by viral proteins or that may have restrictive properties that do not exist in equivalent proteins from the host species. Restriction factors and inhibition of cell-free virus by retroviruses and other enveloped viruses has been a very active area of research for a number of years (reviewed in ([@bb0380 bb0580])). By contrast, less is known about whether restriction factors are inhibitory during cell-to-cell spread, although some studies are beginning to address these questions. TRIM5 and APOBEC {#s0025} ---------------- When considering the site of action of restriction factors in the context of cell-to-cell spread it is helpful to consider restriction factors in two groups, divided on the basis of which steps they target in the cell-free retroviral life cycle: those that block HIV-1 exit from the virus transmitting donor cell (e.g., tetherin and ISG15) and those that may act to restrict the early steps of virus infection (e.g., APOBEC3 and TRIM). Regarding restriction factors that target early steps (pre-integration) in the retroviral life cycle once it enters a target cell, the pertinent question is whether the restriction factor is saturable and therefore whether cell-to-cell spread may overwhelm the existing cytoplasmic pool of protein laying in wait for viral invaders. Despite the uncertainty about the mechanism of productive entry during cell-to-cell spread (endocystosis or fusion at the plasma membrane fusion), restriction factors target steps of the retroviral life cycle (uncoating, reverse transcription and integration) that are essential for successful proviral integration and must be achieved for productive infection to ensue. Rhesus TRIM5α is a potent inhibitor of HIV-1 infection that restricts at different stages of the viral life cycle probably by promoting capsid disassembly ([@bb0055 bb0530]), inhibiting reverse transcription ([@bb0525]) and preventing integration ([@bb0010 bb0080 bb0585]). Notably, there is clear species specificity in TRIM5 activity and HIV-1 is largely resistant to restriction by human TRIM5 but is sensitive to restriction by TRIM5 from Old World Monkeys ([@bb0525]). This has raised the possibility of engineering human cells to express rhesus TRIM5 (rhTRIM) as a therapeutic intervention to target HIV-1. Since human TRIM5 cannot restrict HIV-1 the issue of cell-to-cell versus cell-free restriction was addressed by engineering primary human CD4^+^ T cells to express rhTRIM5 and investigating whether transmission of virus between T cells was efficiently restricted ([@bb0430]). Taking this approach Richardson et al., reported that rhTRIM5 efficiently blocked cell-free infection, but not infection mediated by cell-to-cell spread. Moreover they observed that inhibition of HIV-1 spreading infection in *in vitro* culture required rhTRIM5 to be expressed in both the virus-producing cell and the target cell and rhTRIM5 was unable to inhibit cell-to-cell spread unless expressed by the majority of cells in culture. It has previously been suggested, albeit controversially, that TRIM5 can restrict the production infectious HIV-1 when expressed in the producer cell ([@bb0450]) but this is unlikely to be the mechanism of restriction of cell-to-cell transmission. It is most likely that cytoplasmic TRIM5 in target cells can be saturated by the increase in capsid that is transmitted during cell-to-cell spread and so expression of rhTRIM5 in the donor cell allows rhTRIM to also associate with capsid during virus production, thus tipping the scales in favour of restriction upon entry into target cells. Members of the APOBEC family of restriction factors are incorporated into nascent virions in the virus-producing cell that inhibit retroviral infection in target cells by deaminating dC to dU in nascent minus-strand DNA, resulting in G-to-A hypermutation ([@bb0165 bb0305 bb0325]) thereby inhibiting reverse transcription and integration (reviewed in ([@bb0005]). This is overcome by HIV-1 Vif that prevents packaging of APOPBEC into particles during virus assembly in infected cells, in part by proteasome-dependent degradation and/or possibly by direct inhibition of encapsidation (reviewed in ([@bb0005])). APOBEC is induced by type-1 interferon in macrophages but not activated primary T cells or T cell lines ([@bb0280 bb0425 bb0515]) and it has been suggested that within the target cell APOBECs are unlikely to be strong inhibitors ([@bb0135]). To date no studies have addressed whether APOBECs restrict cell-to-cell spread as efficiently as cell-free infection, but it seems most likely that cell-to-cell spread would be similarly sensitive to APOBEC-mediated DNA editing since there is no reason why Vif should not exclude APOBEC encapsidation during *de novo* virus assembly at the VS ([@bb0195]), unless the rapid and polarized assembly of virions temporally or spatially precludes efficient Vif activity for some reason. Further work is needed to fully understand how APOBECs get incorporated into nascent virus. For example, at what stage they encounter viral RNA, whether this may differ between cell-free virus production or assembly at the VS, and where Vif interacts with APOBEC in order to speculate about whether cell-to-cell spread may be similarly susceptible to APOBEC-mediated restriction in the absence of Vif, or whether some APOBEC may slip through. In addition to TRIM5 and APOBECs it is very likely there are other as yet undiscovered restriction factors that target post-entry steps of HIV-1 infection in target cells ([@bb0135]) and further investigation of restriction of cell-to-cell spread is certainly warranted. Tetherin and ISG15 {#s0030} ------------------ Evidence to date suggests that the general mechanism of assembly and budding of retroviruses from productively infected CD4^+^ T cells and macrophages at the VS is the same as that of cell-free virus release ([@bb0140 bb0155 bb0195 bb0230 bb0335 bb0345]). Live cell imaging has previously revealed *de novo* assembly of HIV-1 and MLV with preferential viral assembly at the contact site ([@bb0195 bb0215]). Retroviral cell-to-cell spread may also occur without significant *de novo* assembly via the transfer of infectious virions that have budded through the plasma membrane but remain associated with the cell surface by interactions with cellular proteins ([@bb0420 bb0480]). It can postulated then that restriction factors that target late steps in virus production in donor cells might be similarly active at inhibiting viral egress during either cell-free transmission or direct cell-to-cell spread. Conversely cell-to-cell spread may saturate a restriction factor if it was not associated with the VS at the right time and in sufficient quantity. Moreover, different mechanisms of cell-to-cell spread (e.g., *de novo* virus production at cell-to-cell contacts versus lateral movement of budding virus from distal membrane domains towards the VS) may be susceptible to restriction while others may be less so. Two studies have recently addressed this question and considered whether cell-to-cell spread of HIV-1 between T cells is sensitive or resistant to restriction by tetherin, with apparently conflicting results. Casartelli et al. reported that tetherin inhibited cell-to-cell spread of HIV-1 and that tetherin blocked productive transmission by reducing viral fusigenicity and thus infectivity. Unusually large aggregates of virus were transferred from tetherin-expressing HeLa donor cells to target cells and this virus was less able to initiate productive infection ([@bb0045]). By contrast, Jolly et al. found that productive cell-to-cell spread of HIV-1 between T cells was not restricted by endogenous tetherin expressed on donor T cells and infectious virus was transmitted across the T cell VS resulting in productive infection ([@bb0225]). The observation that Vpu-defective virus is transmitted as efficiently, if not more so than WT virus is in agreement with a number of previous studies ([@bb0160 bb0275 bb0470 bb0495 bb0520 bb0535 bb0595]). Possible reasons for the differences observed between our study and Casartelli et al. may not be immediately obvious since both groups agree that tetherin is present at the T cell VS and that synapses appear to be form normally in the presence of tetherin on the virus-producing cell ([@bb0045 bb0225]). In both studies, cell-to-cell spread of HIV-1 was interrogated in the presence or absence of Vpu using Vpu-expressing or non-expressing virus: Vpu-defective virus does not antagonize tetherin and results in the well-characterized budding defect where nascent, proteolytically matured virions remain attached to the surface of the virus-producing cell by membrane tethers ([@bb0390 bb0520 bb0555]). In the presence of Vpu, tetherin activity is abrogated and the protein is degraded via the proteasome ([@bb0125 bb0320 bb0555]) and/or lysosome ([@bb0095 bb0210 bb0365]), but whether this results in global down-regulation of tetherin from the plasma membrane, or simple exclusion of tetherin from membrane regions that prevent association with viral proteins is unclear. The findings of Casartelli et al., and Jolly et al., may be somewhat reconciled by considering the type of the donor cells used to examine cell-to-cell spread and the chronicity of the infection, since the experimental assays were broadly similar. It may be that in the presence of lower levels of tetherin, such as is endogenously expressed on T cells (used by ([@bb0225])), productive cell-to-cell spread can take place without effective restriction ([Fig. 2](#f0010){ref-type="fig"}). Tetherin expressed on the surface of T cells would be sufficient to retain virions at the cell surface and in this way mature infectious virus attached to the plasma membrane by tetherin would be poised to engage CD4 on target cells facilitating VS and polysynapse formation and more rapid cell-to-cell spread as we observed with Vpu-defective HIV-1 ([@bb0225]). Notably, we did not detect any loss of viral infectivity of virus produced from T cells when Vpu was absent and tetherin was unantagonized ([@bb0225]). By contrast HeLa cells or 293T cells transfected with plasmid-encoding tetherin express higher levels of tetherin at the cell surface ([@bb0370 bb0385 bb0390 bb0555]). When these cells were used as donor cells productive cell-to-cell spread was inhibited due to the formation of unusually large viral aggregates. Viral aggregates were transferred to target cells but remained stuck at the cell surface and virus did not appear to fuse appropriately at the plasma membrane, leading to reduced infectivity ([@bb0045]) ([Fig. 2](#f0010){ref-type="fig"}). This effect may be explained by large amounts of tetherin being incorporated into virions. A recent report similarly observed an apparent reduction HIV-1 infectivity in the presence of tetherin ([@bb0600]); however in this study virus was also produced from epithelial cells co-transfected with tetherin-encoding plasmid therefore resulting in higher tetherin expression that would be expected on T cells. It should be noted that variations in tetherin expression have also been reported between different T cell lines ([@bb0370]) and this too may account for some of the discordant results in cell-to-cell spread when T cells were used. It will be interesting to see if future studies probe the infectivity of HIV-1 produced from infected T cells and if so, whether they observe reduced infectivity or not in the presence of tetherin. Considering the chronicity of the infected cells, we generally used cells later after initial infection in order to maximise the percentage of infected cells used in our assays to obviate any effect of virus spreading within the donor population in co-culture assays. It is possible that allowing the infection to proceed for longer in T cells could also increase the amount of proteolytically mature, infectious virus tethered at the cell surface thus enhancing transmission of Vpu-defective virus. Moreover, it is possible that longer incubations may provide opportunity for cell-to-cell contacts to occur that could induce budding but keep nascent virus trapped at the cell surface, there by forming structures similar to tetherin-containing viral biofilms that have been shown to facilitate cell-to-cell spread of HTLV-1 at VS ([@bb0420]). Using cells at earlier times post-infection may result in less tethered virus accumulating at the plasma membrane of T cells infected with Vpu-defective virus resulting in reduced virus transfer. It may also be speculated that *de novo* virus assembly at the VS may result in a shift in the balance of immature vs. mature virions being transferred by cell-to-cell spread in favour of non-infectious, immature virus. It is interesting to note that neither HTLV-1 nor MLV encode known tetherin antagonist but disseminate predominantly via cell-to-cell spread. Recent evidence suggests that tetherin can partly reduce MLV transmission in mouse cells by an as yet unknown mechanism ([@bb0130]) but whether this is due to unidentified antagonism by an MLV protein or another mechanism is unclear. Thus it seems possible that HIV-1 may use different mechanisms under certain conditions and that retroviruses may also be able to use tetherin to its advantage in some situations. Moreover, it is also possible that different cell-to-cell interactions (e.g., DC-T cell or macrophage-T cell) may shift the balance towards transmission or restriction by tetherin during cell-to-cell spread. Notably, T cell-to- T cell spread is less sensitive to interferon-mediated inhibition than cell-free infection suggesting that cell-to-cell transmission can partially overcome interferon-induced blocks to transmission ([@bb0565]). It is tempting to speculate that differences in cell-to-cell versus cell-free spread in the presence or absence of interferon may account for the variable results seen using interferon-based therapy ([@bb0110 bb0170 bb0200 bb0265 bb0300 bb0435 bb0490 bb0505]). Clearly more work is needed to clarify this area and to more completely delineate the different possible mechanism of HIV-1 cell-to-cell spread and how tetherin may fit into this. Other interferon inducible cellular proteins can limit HIV-1 release such as ISG15, a ubiquitin-like protein that inhibits ubiquitination of Gag and Tsg101 which arrests HIV-1 budding at a late stage ([@bb0415]) but whether ISG15 is similarly active against cell-to-cell spread in unclear. Late-budding defects involving ESCRT are characterized by the accumulation of immature virus at the cells surface and immature HIV-1 is not infectious, thus is it likely that ISG15 would interfere with cell-to-cell spread. Concluding remarks {#s0035} ================== Innate immune activation and interferon secretion following exposure of cells to retroviruses such as HIV-1 has important consequences for immune regulation and is a double-edged sword during HIV-1 infection ([@bb0050]). In the short term, innate immune activation is necessary for recruitment of effector cells and initiating adaptive immunity; however the recruitment of target T cells and macrophages, for example to sites of virus infection, and subsequent T cell activation increases the local pool of susceptible target cells able to support robust viral replication. Moreover, chronic immune activation will result in the loss of CD4^+^ T cells, homeostatic inbalance and T cell exhaustion, increasing the risk of disease progression and opportunistic infections. How cell-to-cell spread is modulated by innate immunity and how viral dissemination maybe counteracted by innate recognition is an emerging area of research and further work is clearly needed to address the molecular mechanisms of cell-to-cell spread in the context of innate immunity and the role of interferon-induced restriction factors in this. Early insights into what is a budding area of research, are turning out to be intriguing and probably have consequences for therapeutic intervention and future efforts to tackle HIV-1/AIDS. C.J. is supported by a Medical Research Council Career Development Award (G0800312). I wish to thank Mark Turmaine (University College London) for assistance with the scanning electron microscopy. ::: {#f0005 .fig} Fig. 1 ::: {.caption} ###### Scanning electron micrographs (SEM) showing the characteristic tetherin-mediated restriction of HIV-1 release from T cells. A) Jurkat T cells infected with Vpu-defective HIV-1 (ΔVpu- HIV-1) and B) Jurkat T cells infected with Vpu-expressing WT HIV-1 (WT HIV-1) were imaged by SEM at 7 days post-infection. Left hand panel scale bar = 500nm. Right hand panels show higher magnification of the T cell plasma membrane arrows, scale bar = 200nm. Some example virions are indicated with arrows. Note that more virions are tethered at the plasma membrane of cells infected with Vpu-defective virus (upper panels). :::  ::: ::: {#f0010 .fig} Fig. 2 ::: {.caption} ###### A proposed model for cell-to-cell spread of Vpu-defective HIV-1 in the presence of different amounts of unantagonized tetherin on the surface of the virus-producing cell. Under conditions where the donor cell expresses lower levels of tetherin (e.g., endogenous expression on T cells) cell-to-cell spread can occur and target T cells become productively infected. When the donor cell expresses very high levels of tetherin (e.g., epithelial cells such as HeLa cells or cells over-expressing plasmid-encoded tetherin) then virus could be transferred as unusually large, tethered aggregates that cannot fuse appropriately at the plasma membrane and productive infection would be blocked. :::  :::

PubMed Central

2024-06-05T04:04:19.910183

2011-3-15

PMC3053476

Introduction ============ Quinolones are molecules structurally derived from the heterobicyclic aromatic compound quinoline, the name of which originated from the oily substance obtained after the alkaline distillation of quinine ([@b67]). Since the isolation of quinine from *Cinchona* bark in 1811, many other quinoline derivatives have been isolated from natural sources ([Fig. 1](#fig01){ref-type="fig"}). In particular, 2-hydroxyquinoline and 4-hydroxyquinoline, which predominantly exist as 2(1*H*)-quinolone and 4(1*H*)-quinolone, respectively, and form the core structure of many alkaloids, were isolated from plant sources. Several different animal and bacterial species also produce compounds of the quinolone class. These differ not only in the varied substitutions in the carbocyclic and heteroaromatic rings but also have other rings fused to the quinolone nucleus. These have been reviewed on a yearly basis by J.P. Michael in Natural Product Reports ([@b128]). Some of these naturally occurring quinolones have profound medicinal properties while others have served as lead structures and provided inspiration for the design of synthetic quinolones as useful drugs. For example ([Fig. 1](#fig01){ref-type="fig"}), among 2-quinolones, rebamipide is an antiulcer agent and repirinast has antihistamine properties useful in the treatment of allergic asthma ([@b187]). While screening compounds for potential cancer chemopreventive properties, casimiroine, isolated from the seeds of *Casimiroa edulis*, was found to have antimutagenic activity ([@b83]). Several 4-quinolone alkaloids, mainly isolated from plant and microbial sources, have antimicrobial activity. For example, 2-alkyl-4(1*H*)-quinolones (AQs) ([Fig. 1](#fig01){ref-type="fig"}, compounds 1--4) and 1-methyl-2-\[(4*Z*)-tridecenyl\]-4(1*H*)-quinolone, evocarpine, its structural isomers and unsaturated homologues ([Fig. 1](#fig01){ref-type="fig"}, compounds 5--9) isolated from the extracts of *Evodia rutaecarpa* show antibacterial activity against *Helicobacter pylori*, which is implicated in the pathogenesis of chronic gastritis, peptic ulcers and gastric cancers ([@b156]; [@b72];). The alkaloid 1 shown in [Fig. 1](#fig01){ref-type="fig"} is rather rare as it bears *n*-decyl, an even number of carbons in the 2-position. Also, no fewer than eight further 4-quinolones ([Fig. 1](#fig01){ref-type="fig"}, compounds 10--17) isolated from the fermentation broth of the actinomycete *Pseudonocardia* spp. CL38489 are active in inhibiting the growth of *H. pylori*. The most potent compound is the epoxide ([Fig. 1](#fig01){ref-type="fig"}, compound 16), which has a potent bactericidal \[minimal inhibitory concentration (MIC) 10 ng mL^−1^\] and an even more pronounced bacteriostatic effect (MIC 0.1 ng mL^−1^) ([@b41]). These quinolones are characterized by the presence of a geranyl or oxidized geranyl side chain at C-2 in place of the usual fatty acid-derived alkyl or alkenyl chain normally found in microbial quinolones. The screening of synthetic analogues of quinine for novel antiplasmodial drugs led to the serendipitous discovery of a precursor used in the synthesis of chloroquine, 7-chloroquinoline, which exhibited antimicrobial activities *in vitro*. Further investigation of this and similar compounds such as the structurally related 1,8-naphthyridones (which are quinolones with a nitrogen atom substituting C-8) resulted in the discovery of nalidixic acid (1-ethyl-7-methyl-4-oxo-1,8-naphthyridine-3-carboxylic acid), which was to become the first practical synthetic quinolone antibiotic ([@b104]). This rapidly led to the development of several other 4-quinolone-based antibiotics such as oxolinic acid, cinoxacin and flumequine ([Fig. 1](#fig01){ref-type="fig"}), used clinically to treat Gram-negative bacterial infections, and later on to second-generation drugs such as norfloxacin and ciprofloxacin ([Fig. 1](#fig01){ref-type="fig"}), also effective against some Gram-positive bacteria. All of the quinolone antibiotics are characterized by the presence of a carboxylic acid function at C-3 ([@b160]; [@b117]; [@b9]; [@b63]; [@b1];). ::: {#fig01 .fig} Fig. 1 ::: {.caption} ###### Natural and synthetic quinolones of medicinal interest, quinolone antibiotics. Several plant, animal and microbial species produce quinolone compounds of medicinal interest such as the antimalarial quinine extracted from *Cinchona* spp., or the 2-quinolone casimiroine, an antimutagen extracted from *Casimiroa edulis*. Among the synthetic 2-quinolones are the antiulcer agent rebamipide and the antihistamine, repirinast. Naturally occurring quinolones having antimicrobial activities such as evocarpine and related compounds (nos 1--17) produced by *Evodia rutaecarpa* are active against *Helicobacter pylori*, a causative agent of peptic ulcers and gastric cancer. The quest for synthetic analogues of quinine led to the discovery of nalidixic acid, oxolinic acid and cinoxacin, and then to the development of an extensive family of fluoroquinolone antibiotics such as flumequine, norfloxacin and ciprofloxacin. The heteroaromatic ring atom numbering common to all quinolones is indicated for quinine. :::  ::: Interest in the antipathogenic properties of common bacteria started with the pioneering work of Louis Pasteur. Notably, in 1877, Pasteur reported that the coinoculation of *Bacillus anthracis* with other common living bacteria in animals prevented the development of anthrax, when septicaemia could be avoided. This was interpreted, following the idea that 'life can prevent life', as being the result of a competition for oxygen ([@b142]). After Emmerich and Pawlowsky were able to prevent the development of anthrax in preinfected rabbits and guinea-pigs by the inoculation of *Streptococcus* spp., Charles Bouchard reproduced this effect in 1889 with pure cultures of *Bacillus pyocyaneus* (*Pseudomonas aeruginosa*) ([@b12]). Ten years later, in 1899, Rudolf Emmerich and Oscar Löw concluded that *P. aeruginosa* released an active antibacterial substance into the medium after cell-free preparations from this organism were found to be sufficient to prevent the development of anthrax, and as it was thought to be the result of an enzymatic process, they called it *pyocyanase* ([@b52]). In 1945, this preparation was determined to consist of a mixture of heat-stable compounds that were separated, partially characterized and named the Pyo compounds ([@b75]). Pyo I--IV, which later were found to be AQs ([Fig. 2](#fig02){ref-type="fig"}), presented strong antibacterial activities against Gram-positive organisms, although much less against Gram-negative bacteria, with Pyo II being 10 times more potent compared with the others. However, Pyo II was toxic and ineffective at protecting mice at subtoxic doses against *Streptococcus pneumoniae* or *Mycobacterium tuberculosis* infections ([@b200]). ::: {#fig02 .fig} Fig. 2 ::: {.caption} ###### Structure, IUPAC names and abbreviations of AQ molecules synthesized by *Pseudomonas aeruginosa* and a synthetic analogue. Both the tautomeric lactam and the phenolic forms of each molecule are shown. Arrows indicate the equilibrium of these molecules as would exist under physiological conditions. Where more than one name exists for a molecule, the IUPAC designation is indicated, although this may not be the nomenclature used most frequently. The compound C1-PQS is a synthetic analogue that is not produced by *P. aeruginosa*. :::  ::: Natural antimicrobial quinolones ================================ In addition to having antimicrobial activity *in vitro*, the Pyo compounds produced by *P. aeruginosa* were found to antagonize, under certain conditions, the action of streptomycin and dihydrostreptomycin against Gram-positive bacteria ([@b109], [@b110]). This inhibitory activity was rapidly attributed to Pyo II, which is a mixture of 2-alkyl-4-hydroxyquinoline *N*-oxides (AQNOs) ([@b33], [@b34]; [@b111]). The inhibitory activity of these *N*-oxides (at concentration ratios with respect to streptomycin of the order of 1 : 100) was found to correlate with the potent inhibition of electron transport in both heart-muscle and bacterial cells through the cytochrome *bc*~1~ segment (ubiquinol:cytochrome *c* oxidoreductase) of the respiratory chain ([@b111], [@b112]). This is in line with the need for respiration and the transmembrane potential required for bacteria to take up aminoglycoside antibiotics ([@b73]; [@b37]; [@b7]; [@b180];). 2-Heptyl-4-hydroxyquinoline *N*-oxide (HQNO) acts as a ubiquinone and menaquinone analogue on quinone-reactive cytochrome *b* enzymes in various organisms ([@b188]; [@b174]; [@b162];). The antimicrobial effect of the *N*-oxides appears to be limited to Gram-positive bacteria and offers an explanation as to how *P. aeruginosa* becomes the dominant species over *Staphylococcus aureus* in cystic fibrosis (CF) lung infections ([@b115]), although additional factors such as pyocyanin, cyanide and *N*-(3-oxododecanoyl)-[l]{.smallcaps}-homoserine lactone have been shown to play a similar role ([@b152]; [@b193];). Furthermore, long-term exposure of *S. aureus* to physiological concentrations of HQNO selects for aminoglycoside-resistant, small-colony variants that are typically found in chronic lung infections ([@b80]; [@b11];). Interestingly, the formation of *S. aureus* small-colony variants is the first step towards the development of dual target resistance against fluoroquinolones ([@b140]). The low *in vivo* efficacy, combined with the strong toxicity on mitochondrial respiration, prevented the development of AQNOs as therapeutic antibiotics. However, due to its interference with quinone-dependent cytochromes, HQNO became an invaluable reagent for the study of electron transport chains. A number of quinolones with interesting antimicrobial properties are also produced by various pseudomonads and other microorganisms. For example, under iron limitation, *Pseudomonas fluorescens* ATCC 17400 produces quinolobactin (8-hydroxy-4-methoxyquinaldic acid, [Fig. 3](#fig03){ref-type="fig"}), which acts as a siderophore ([@b132]). Quinolobactin results from the rapid hydrolysis of the precursor molecule 8-hydroxy-4-methoxy-2-quinolinethiocarboxylic acid (thioquinolobactin), which, as opposed to quinolobactin, has strong antifungal activity against the plant pathogen *Pythium debaryanum* ([@b124]). Thioquinolobactin is synthesized via a unique pathway from [l]{.smallcaps}-tryptophan via xanthurenic acid and sulphurylation by QbsE, a small sulphur carrier protein ([@b123]; [@b69];). *Pseudomonas fluorescens* G308, a potential biocontrol strain, produces *N*-mercapto-4-formylcarbostyril \[Cbs, 4-formyl-1-sulphanyl-2(1*H*)-quinolone, [Fig. 3](#fig03){ref-type="fig"}\], a quinolone that contains an unusual *N*-mercaptoamide functional group and that has strong antifungal properties against plant pathogens such as *Fusarium* spp., *Cladosporium cucumerinum* and *Colletotrichum lagenarium* ([@b57]). Although the biosynthetic pathway for Cbs formation has not yet been elucidated, it was suggested that this compound may be derived from AQs produced via a similar biochemical pathway as that in *P. aeruginosa*. The marine bacteria *Pseudomonas bromoutilis* ([@b210]) and *Alteromonas* strain SWAT5 ([@b113]) both synthesize 2-pentyl-4-quinolone (PHQ, also called 2-*n*-pentyl-4-quinolinol) and 2-heptyl-4-hydroxyquinoline (HHQ, also called 2-*n*-heptyl-4-quinolinol), which were identified as a consequence of their antibacterial activities. PHQ inhibits the growth of cyanobacteria (*Synechococcus*), algae (*Chaetoceros simplex*, *Cylindotheca fusiformis* and *Thalassiosira weissflogii*) and impacts on particle-associated marine bacterial communities ([@b113]). HHQ and PHQ have antibacterial activity against *Vibrio anguillarum*, *S. aureus*, *Candida albicans* and *Vibrio harveyi* ([@b210]). From sponge-associated marine pseudomonads, several other AQs with various substitutions have been identified and appear to have antibacterial, antiplasmodial, antiviral or cytotoxic properties ([@b39]; [@b15];). The obligate aerobic yeast *Yarrowia lipolytica* produces 1-hydroxy-2-dodecyl-4(1*H*)-quinolone, a potent inhibitor of the alternative NADH:ubiquinone oxidoreductases, which acts as a ubiquinone analogue ([@b53]), an activity reminiscent of the quinolone *N*-oxides produced by *P. aeruginosa*, which act on the cytochrome *bc*~1~ complex. In addition, a large number of additional quinoline alkaloids produced by a variety of other microorganisms, plants and animals have been discovered every year (annually reviewed by J.P. Michael), of which the majority still have to be studied with respect to their biological properties. However, as all the natural quinolones described so far lack the 3-carboxy group, which is essential for the binding and blocking of DNA-type IIA topoisomerase complexes, the antibacterial mechanism of action of these compounds remains to be elucidated. ::: {#fig03 .fig} Fig. 3 ::: {.caption} ###### Sulphur-containing quinolones produced by some pseudomonads. Thioquinolobactin, a compound exhibiting strong antifungal properties, is produced by *P. fluorescens* ATCC 17400. Upon spontaneous hydrolysis, thioquinolobactin is rapidly converted into quinolobactin, which then acts as a siderophore. *Pseudomonas fluorescens* G308 produces Cbs, which also exhibits potent fungicidal properties. :::  ::: Synthetic quinolone antibiotics =============================== The practical applications of nalidixic acid ([Fig. 1](#fig01){ref-type="fig"}) as an antimicrobial of therapeutic interest became evident soon after it was discovered ([@b104]; [@b198];). Because it is a polar molecule that avidly conjugates to serum proteins, and therefore presents a large volume of distribution, it is inadequate for the systemic treatment of infections. However, both nalidixic acid and its principal 7-hydroxymethyl metabolite that remains active undergo rapid renal excretion and readily accumulate in the urinary tract ([@b161]; [@b190];). As nalidixic acid is notably efficient at arresting the growth of common enterobacteria, its principal indication was in the treatment of uncomplicated urinary tract infections ([@b198]). It is, however, of little use against infections occurring outside of the urinary tract or those that are caused by organisms such as *P. aeruginosa* and Gram-positive pathogens that are intrinsically resistant to the practical therapeutic concentrations of the antibiotic. From the 1980s onwards, there appeared successive generations of antibiotics related to nalidixic acid such as the fluoroquinolones, which, due to substitutions in the molecule, more specifically the addition of a 6-fluoro group, have extended therapeutic spectra and enhanced pharmacokinetic properties. The development of fluoroquinolones ([Fig. 1](#fig01){ref-type="fig"}) such as flumequine, norfloxacin and ciprofloxacin (one of the most consumed antibiotic worldwide; [@b165]) extended the spectrum of activity of quinolone antibiotics against infections caused by a variety of otherwise resistant organisms such as *P. aeruginosa* and both aerobic and anaerobic Gram-positive pathogens, and enabled the treatment or the prevention of more severe conditions such as renal, respiratory, abdominal and sexually transmitted bacterial infections (for an extensive review on quinolone antibiotics, see [@b190]). Quinolone antibiotics act by inhibiting the two type IIA bacterial topoisomerases: DNA gyrase and topoisomerase IV (bacterial type IIA topoisomerases have been reviewed recently by [@b172]). DNA gyrase is a heterotetramer formed by two subunits encoded by *gyrA* (*nalA*) and *gyrB* (*nalC*). GyrA together with GyrB acts by creating DNA gates or double-stranded gaps in the DNA through which the strands are passed, introducing negative supercoils into DNA and relaxing the positive supercoiling resulting from replication as the strands unwind ([@b25]; [@b196]; [@b32]; [@b99]; [@b50];). Nalidixic and oxolinic acid, a more potent, but structurally similar quinolone antibiotic ([@b176]), were initially found to inhibit *in vitro* the supercoiling activity of purified DNA gyrase ([@b66]; [@b179];). Whereas only gyrase is able to introduce negative supercoiling, the function of DNA topoisomerase IV, a heterotetramer formed by two ParC-ParE subunits similar to the GyrA-GyrB subunits of DNA gyrase, is essential for the relaxation of supercoiled DNA and the resolution of catenated DNA molecules after replication ([@b88], [@b89]; [@b2]; [@b146]; [@b82]; [@b26]). The precise mode of action of the quinolone antibiotics on type IIA topoisomerases has long been debated. However, crystallographic studies strongly suggest that these molecules essentially act by blocking the DNA--topoisomerase complexes when the nucleic acid is cleaved ([@b26]; [@b95];). Thus, in addition to sharing structural and functional similarities, both DNA gyrase and topoisomerase IV can be inhibited by quinolone antibiotics, leading to bacterial cell death due to chromosome fragmentation. Whereas the target of quinolone antibiotics in *Escherichia coli* and other Gram-negative bacteria is mainly DNA gyrase, in Gram-positive species such as *S. aureus* or *S. pneumoniae*, their principal mode of action lies mainly in the inhibition of topoisomerase IV, with exceptions depending on the particular fluoroquinolone compound and bacterial species ([@b79]; [@b51]; [@b95];). Resistance to quinolone antibiotics (reviewed by [@b85] and by [@b118]) can be achieved by three nonexclusive mechanisms: (1) by the acquisition of point mutations in the genes encoding either of the two type IIA topoisomerases targeted, DNA gyrase and DNA topoisomerase IV, (2) by reducing the effective concentrations of the drugs in the cytoplasm, either passively by alterations in the membrane permeability or actively by overexpressing efflux systems, and (3) by acquisition of mobile quinolone resistance determinants ([@b165]; [@b85];). While only target modification confers high-level resistance to quinolone antibiotics, the low-level resistance (less than a 10-fold increase in the minimum inhibitory concentration) conferred by the other mechanisms augments the probability of developing such mutations. Mutants that exhibit more than a 100-fold decrease in quinolone sensitivity have been found to carry single point mutations in the chromosomally encoded DNA gyrase subunit gene *gyrA*, with additional enhanced resistance when certain point mutations are simultaneously present in the distantly located *gyrB* gene ([@b213]; [@b215]; [@b36]; [@b216], [@b217]; [@b137]; [@b20]; [@b165]). Molecular details of these mutations and the binding of quinolones to DNA--topoisomerase complexes have been described elsewhere extensively ([@b134]; [@b16]; [@b78]; [@b62]; [@b85];). In Gram-negative bacteria, a reduction in antibiotic permeability can be achieved by altering the cell envelope. For example, *E. coli* or *P. aeruginosa* clinical isolates that produce altered lipopolysaccharides differ in the accumulation of quinolones compared with wild-type strains, a nonspecific mechanism thought to involve changes in surface hydrophobicity and consequently affecting passive drug diffusion ([@b29]; [@b106]; [@b56]; [@b27];). However, a reduction in permeability to quinolones is most often achieved by disrupting or downregulating a number of outer-membrane proteins that form channels through which quinolones enter the bacterial cell or by the activation of tripartite multidrug efflux systems belonging to the AcrAB-TolC resistance-nodulation-division (RND) family that can prevent quinolone antibiotics reaching effective concentrations in the cytoplasm (reviewed by [@b119]). For instance, *P. aeruginosa* encodes at least 12 RND systems ([@b178]; [@b169];), eight of which (MexAB-OprM, MexCD-OprJ, MexEF-OprN, MexXY-OprM, MexJK-OprM, MexHI-OpmD, MexVW-OprM and MexPQ-OpmE) have been reported to export fluoroquinolones and other antibiotics ([@b28]; [@b107]; [@b171]; [@b189]; [@b130];), although antimicrobial resistance does not appear to be their primary biological function ([@b3]; [@b189]; [@b149], [@b150]). Resistance to quinolone antibiotics can also result from the acquisition of plasmid-borne determinants. The MFS-type efflux pump QepA and the Aac(6′)-Ib-cr enzyme that confers decreased susceptibility to piperazinyl fluoroquinolones such as ciprofloxacin and norfloxacin by acetylation are examples of recently discovered resistance genes carried by plasmids ([@b159]; [@b147]; [@b214];). More frequent, however, are the plasmids carrying Qnr quinolone resistance loci, of which at least five families have been identified, mostly in *Enterobacteriaceae* ([@b84]; [@b24]; [@b197];). These genes encode proteins of the pentapeptide repeat family that interact with DNA gyrase and topoisomerase IV, preventing quinolone inhibition by mimicking DNA, which probably reduces the availability of holoenzyme--DNA targets for quinolone inhibition ([@b184]; [@b77]; [@b185],[@b186];). As with the other quinolone resistance determinants, this function may be considered biologically fortuitous because natural quinolones inhibiting type IIA topoisomerases have not been discovered. Quinolones produced by *P. aeruginosa* ====================================== Besides Pyo II, which is a 2 : 1 mixture of HQNO and 2-nonyl-4-hydroxyquinoline *N*-oxides (NQNO) ([@b109], [@b110]), with small quantities of 2-undecyl-4-hydroxyquinoline *N*-oxide (UQNO) ([@b34]), *P. aeruginosa* also releases a large number of related molecules. Using ozonolysis and UV absorption spectra in comparison with synthetic standards, [@b201] identified Pyo Ib as 2-heptyl-4(1*H*)-quinolone (HHQ), Pyo Ic as 2-nonyl-4(1*H*)-quinolone (NHQ) and Pyo III as a monounsaturated alkyl side chain variant of NHQ ([@b199]; [@b201];). The AQ biosynthetic enzymes of *P. aeruginosa* enable this organism to generate a diverse range of related AQ molecules ([Fig. 2](#fig02){ref-type="fig"} and Box 1). An early study using GC and electron capture MS identified over 20 different AQs ([@b182]), with HHQ being the most prevalent, followed by NHQ. Variations of these compounds containing saturated and monounsaturated alkyl side chains varying from one to 13 carbons in length, and the two major *N*-oxides, HQNO and NQNO, were also found. Two subsequent studies used electrospray ionization and LCMS to obtain the mass spectra of over 50 different AQs. These mainly consisted of 2-heptyl-3-hydroxy-4(1*H*)-quinolone \[termed the *Pseudomonas* quinolone signal (PQS)\], HHQ, HQNO and NHQ, with several other saturated and monounsaturated alkyl side chains of various lengths ([@b101], [@b102]). Additional AQs that have been found in significant amounts are 2-nonyl-3-hydroxy-4(1*H*)-quinolone (C9-PQS), 2-undecyl-4-hydroxyquinoline (UHQ), NQNO and UQNO ([@b182]; [@b43]; [@b102];). Several variations of these compounds are produced ([Fig. 2](#fig02){ref-type="fig"} and Box 1), but many at seemingly biologically insignificant levels, perhaps as a consequence of a lack of specificity of the AQ biosynthetic enzymes for β-keto fatty acids of different chain lengths rather than for any particular biological function. In addition, a metabolite identified as 2,4-dihydroxyquinoline (DHQ) was found in cultures of both *P. aeruginosa* and *Burkholderia thailandensis* ([@b103]; [@b221];). DHQ, although structurally related, is technically not an AQ as it lacks a 2-alkyl chain. It is neither a degradation product nor a precursor of AQs and the precise function of this molecule remains unknown. However, DHQ inhibits the growth and cell viability of mouse lung epithelial MLE-12 cells ([@b221]) and therefore this molecule may play a role in pathogenicity in respiratory tract infections. ::: {.caption} ###### Box 1. Nomenclature and abbreviations of AQs used in this review ::: The structures, IUPAC-based nomenclature and abbreviations of all the major AQs produced by *Pseudomonas aeruginosa* are summarized in [Fig. 2](#fig02){ref-type="fig"}. Some non-IUPAC names that have been used to describe some of these same molecules in the scientific literature have been included for clarity. AQs, 2-alkyl-4(1*H*)-quinolones (lactam form) are tautomeric with 2-alkyl-4-hydroxyquinolines (phenolic form), of which the predominance of one form over the other is determined by the pH ([@b90]; [@b91]; [@b96]). For example, it has been demonstrated using p*K*~a~ values for 2-methyl-3-hydroxy-4(1*H*)-quinolone (C1-PQS) that over physiological pH ranges, the neutral 4-quinolone form is the predominant tautomer ([@b49]). These tautomeric forms are shown in [Fig. 2](#fig02){ref-type="fig"}, with their relative ratios indicated by the arrows. Ideally, for consistency and structural accuracy, nomenclature and abbreviations based on only one tautomeric form should have been uniformly adopted. However, this causes some difficulties because in the available scientific literature individual research groups have subjectively referred to these molecules in either one form or the other. For example, even the names used for the two main AQ molecules involved in signalling, PQS and HHQ, are inconsistent with each other with regard to tautomerism: *Pseudomonas* quinol[one]{.underline} signal and 2-heptyl-4-hydroxyquinol[ine]{.underline}. The nomenclature and abbreviations used in this review therefore amount to a compromise between what is technically correct, taking into account IUPAC designations and structural predominance due to physiological pH, and also what has been the prevalent terminology used in the scientific literature for each molecule. Hence, the abbreviation PQS to designate 2-heptyl-3-hydroxy-4(1*H*)-quinolone has been maintained and the alkyl side chain variants of this molecule abbreviated by the number of carbon atoms in the side chain, for example C1-PQS, C9-PQS. The designation HHQ for 2-heptyl-4-hydroxyquinoline has also been maintained, and the other alkyl side-chain derivatives have been abbreviated accordingly (e.g. PHQ for 2-pentyl-4-hydroxyquinoline, etc.). It should be noted that the *N-*oxide series of compounds (AQNOs, e.g. HQNO, 2-heptyl-4-hydroxyquinoline *N-*oxide; NQNO, 2-nonyl-4-hydroxyquinoline *N*-oxide) can adopt the 2-alkyl-1-hydroxy-4(1*H*)-quinolone form ([Fig. 2](#fig02){ref-type="fig"}) but not at physiological pH. DHQ or 2,4-dihydroxyquinoline can exist in both, 4-hydroxy-2(1*H*)-quinolone (predominant at physiological pH) and 2-hydroxy-4(1*H*)-quinolone tautomeric forms ([Fig. 2](#fig02){ref-type="fig"}), but to avoid confusion and to conform to the literature citations, DHQ is used to denote this molecule in this review. To further help the reader, below is a list of the proposed nomenclature of the AQs that are mentioned along with their abbreviations. Included in this table are the associated synonyms that have been used to describe these same molecules elsewhere in the scientific literature. ::: {#d32e1370 .table-wrap} Suggested nomenclature Synonyms ----------------------------------------------- ---------------------------------------- 2-alkyl-4(1*H*)-quinolone (AQ) 2-alkyl-4-hydroxyquinoline (AHQ) 4-hydroxy-2-alkylquinoline (HAQ) 2-alkyl-4-hydroxyquinoline *N*-oxide (AQNO) 4-hydroxy-2-alkylquinoline *N*-oxide 2-alkyl-1-hydroxy-4(1*H*)-quinolone 2-heptyl-3-hydroxy-4(1*H*)-quinolone 2-heptyl-3,4-dihydroxyquinoline *Pseudomonas* quinolone signal (PQS) 2-heptyl-3,4-quinolinediol 3-hydroxy-2-nonyl-4(1*H*)-quinolone (C9-PQS) 3,4-dihydroxy-2-nonylquinoline 2-nonyl-3,4-quinolinediol 2-pentyl-4-hydroxyquinoline (PHQ) 2-pentyl-4(1*H*)-quinolone 4-hydroxy-2-pentylquinoline 2-pentyl-4-quinolinol 2-heptyl-4-hydroxyquinoline (HHQ) 2-heptyl-4(1*H*)-quinolone 4-hydroxy-2-heptylquinoline 2-heptyl-4-quinolinol 2-nonyl-4-hydroxyquinoline (NHQ) 2-nonyl-4(1*H*)-quinolone 4-hydroxy-2-nonylquinoline 2-nonyl-4-quinolinol 2-undecyl-4-hydroxyquinoline (UHQ) 2-undecyl-4(1*H*)-quinolone 4-hydroxy-2-undecylquinoline 2-undecyl-4-quinolinol 2-heptyl-4-hydroxyquinoline *N*-oxide (HQNO) 4-hydroxy-2-heptylquinoline *N*-oxide 2-heptyl-1-hydroxy-4(1*H*)-quinolone 2-nonyl-4-hydroxyquinoline *N*-oxide (NQNO) 4-hydroxy-2-nonylquinoline *N*-oxide 2-nonyl-1-hydroxy-4(1*H*)-quinolone 2-undecyl-4-hydroxyquinoline *N*-oxide (UQNO) 4-hydroxy-2-undecylquinoline *N*-oxide 2-undecyl-1-hydroxy-4(1*H*)-quinolone 2,4-dihydroxyquinoline (DHQ) 4-hydroxy-2(1*H*)-quinolone 2-hydroxy-4(1*H*)-quinolone 2,4-quinolinediol ::: Properties of AQs ================= Generally, AQs have a low aqueous solubility. For example, the solubility of PQS is around 1 mg L^−1^ (∼5 μM) in water ([@b101]). Because of this hydrophobic nature, a high proportion of the AQs are associated with the bacterial outer membrane and with membrane vesicles (MVs) ([@b121]). Of the total amount of PQS produced by *P. aeruginosa* PA14, around 80% appears to be contained within vesicles, in contrast with \<1% of either of the *P. aeruginosa N*-acyl-homoserine lactone (AHL)-based signal molecules *N*-(3-oxododecanoyl)-[l]{.smallcaps}-homoserine lactone (3-oxo-C12-HSL) and *N*-butanoyl-[l]{.smallcaps}-homoserine lactone (C4-HSL) ([@b120]). The PQS contained within these MVs is seemingly both bioactive and bioavailable because the addition of MVs containing PQS restored the production of pyocyanin in a PQS-negative mutant (PQS being indispensable for the production of pyocyanin in *P. aeruginosa*). The MVs themselves do not seem to have any direct effect on the production of pyocyanin and PQS does not need to be packaged into MVs to exert its effects. MV formation in *P. aeruginosa* PA14 would not seem to be an active process as it occurs independent of growth or of protein synthesis ([@b120]). Instead, PQS appears to initiate the formation of MVs, into which it is then packaged due to its lipophilic nature ([@b120]). A mechanism for MV formation has been proposed via the interaction of PQS with the 4′-phosphate and acyl chain of bacterial lipopolysaccharide ([@b121]). Because HHQ is much less efficient in inducing vesicle formation, this activity seems to be dependent on the 3-hydroxy group of PQS and its analogues ([@b121], [@b122]). A *pqsH* mutant, deficient in the conversion of HHQ to PQS, is defective in vesicle formation. Of PQS and its analogues, MV formation appears to be optimal when a C7 2-alkyl side chain moiety is present, although C5 and C3 alkyl side chain variants also exhibit some activity and MVs can still be induced to some extent by PQS analogues lacking a 2-alkyl side chain, indicating that this group is dispensable. Additionally, compounds that can inhibit PQS production such as indole and its derivatives reduce MV formation, presumably as there is less PQS available to induce vesicle formation ([@b181]). It has been suggested that packaging into MVs could protect PQS from degradation by surrounding cells or competing microbial communities. *Arthrobacter nitroguajacolicus* Rü61 produces the cytoplasmic enzyme Hod \[3-hydroxy-2-methyl-4(1*H*)-quinolone 2,4-dioxygenase\], which catalyses the 2,4-dioxygenolytic ring cleavage of PQS with the concomitant formation of carbon monoxide and *N*-octanoyl-anthranilic acid ([@b151]). As purified Hod is capable of inhibiting AQ signalling when added to cultures of *P. aeruginosa*, at present, the extent of protection conferred by MVs to AQs against enzymatic degradation is unclear. Rhamnolipids are produced by *P. aeruginosa* and act as biosurfactants, facilitating swarming motility ([@b17]). However, rhamnolipids also enhance the aqueous solubility and activity of AQs *in vitro*. The addition of increasing amounts of rhamnolipids enhanced the ability of PQS at a range of concentrations to induce the expression of a *lasB*′-′*lacZ* translational reporter, suggesting that in *P. aeruginosa* the induction of elastase production by PQS is enhanced in the presence of rhamnolipids ([@b19]). Whether rhamnolipids are indeed effectively utilized to solubilize PQS *in vivo* is, however, not known at present, and with respect to the above reporter system, an excess of rhamnolipids even seems to be detrimental to its expression. A reason for this may be that above a certain threshold concentration, PQS is sequestered into rhamnolipid micelles and therefore becomes less available to the cells ([@b19]). AQNOs such as HQNO are potent inhibitors of the cytochrome *bc*~1~ complex and an interesting, but as yet undescribed facet is the mechanism by which *P. aeruginosa* avoids self-poisoning as a consequence of the endogenous production of these molecules. Gram-negative bacteria, as opposed to Gram-positive species such as *Bacillus subtilis* or *S. aureus*, are normally resistant to these compounds, and possible explanations for this have been (1) a reduced cell wall permeability, (2) enzymatic inactivation or (3) an active efflux system to transport the molecule out of the cells ([@b115]). However, none of these mechanisms can account for the resistance of *P. aeruginosa* towards endogenously produced HQNO. Aerobic respiration in *P. aeruginosa* is achieved by a branched electron transport chain ending in five different terminal oxidases, three of which (cytochrome oxidases *cbb*~3~-1, *cbb*~3~-2 and *aa*~3~) receive electrons from ubiquinone via the cytochrome *bc*~1~ complex and cytochromes *c*, whereas the remaining two are the cytochrome *bo*~3~ and the cyanide-insensitive cytochrome *bd* quinol oxidases, which bypass the cytochromes *bc*~1~--*c* electron transfer pathway and get their electrons directly from ubiquinone ([@b205]). Thus, if the cytochrome *bc*~1~ complexes of *P. aeruginosa* were sensitive to HQNO, this compound alone would have the potential to inhibit three out of five electron transport chains, and in combination with the production of cyanide, 80% of aerobic respiration would be inhibited. *Pseudomonas aeruginosa* is also able to perform anaerobic respiration using nitrogen oxides as terminal electron acceptors. Nitrite, nitric oxide (NOR) and nitrous oxide terminal reductases receive electrons from the cytochrome *bc*~1~ complex, while nitrate reductase (NAR) obtains electrons directly from ubiquinone and also in part via a dedicated membrane-bound formate dehydrogenase ([@b205]). In this case, HQNO could potentially block all the nitrogen oxide anaerobic respiration, except for the NAR respiratory chain involving formate oxidation. A recent study reported that abolishing AQ production in *P. aeruginosa* enhanced anaerobic growth on nitrate, and that addition of PQS appeared to repress the growth of the wild type and inhibited denitrifying enzymes ([@b183]). Although this effect was attributed to the iron-chelating properties of PQS, the involvement of increased HQNO production cannot be excluded. Whether *P. aeruginosa* prevents self-poisoning with HQNO under aerobic conditions by favouring the cytochrome *bo*~3~ and/or the cyanide-insensitive cytochrome *bd* oxidase pathways or whether there are mechanisms to prevent competitive inhibition of ubiquinone-dependent enzymes under aerobic and anaerobic growth on nitrate remains to be determined. Biosynthesis of AQs in *P. aeruginosa* ====================================== AQ biosynthesis requires multiple genes, which were initially identified by screening a *P. aeruginosa* transposon mutant library for clones displaying reduced pyocyanin production and were termed *pqsABCDE*, *pqsR* (*mvfR*), *pqsH* and *pqsL* ([@b22]; [@b38]; [@b64]; [@b100];). The *pqsABCDE* (PA0996-PA1000) genes are arranged in an operon, and adjacent to these are the anthranilate synthase genes *phnAB* (PA1001-PA1002) and *pqsR* (*mvfR*, PA1003). Two other genes are also involved in AQ biosynthesis, *pqsH* (PA2587) and *pqsL* (PA4190), but both of these are located separately elsewhere on the chromosome. The *pqsR* gene encodes a LysR-type transcriptional regulator that has a helix-turn-helix motif at the *N*-terminus with the first 280 amino acids sharing high similarity (62--71%) with other LysR-type regulators ([@b22]; [@b116];). PqsR is the transcriptional regulator of both the *pqsABCDE* and the *phnAB* operons and is of crucial importance for AQ production. A mutation in the gene coding for this regulator in *P. aeruginosa* strain PA14 resulted in the abolition of *phnAB* and *pqsABCDE* transcription along with PQS and AQ biosynthesis and had corresponding effects on other virulence determinants including pyocyanin, elastase, exoprotein and 3-oxo-C12-HSL production and consequently the reduced ability to cause disease in plants and animals ([@b22]; [@b43];). The *pqsABCD* genes are involved in the biosynthesis of all AQs ([@b43]). The first step in this biosynthesis involves the activation of anthranilate by PqsA, an anthranilate coenzyme A ligase ([@b30]). PqsB and PqsC, which are similar to β-keto-acyl-ACP (acyl carrier protein) synthases involved in fatty acid metabolism, are predicted to elongate acyl side chains of AQ precursors. However, little is currently known about the enzymatic functions of these proteins. PqsD shares some sequence similarity with the Cys-His-Asn active site of the *E. coli* initiation condensing enzyme FabH ([@b114]; [@b158]; [@b13]; [@b221];). The crystal structure of PqsD with and without a potential covalently bound anthranilate-AQ intermediate product has been resolved recently ([@b10]). Nonpolar mutations in either the *pqsA*, *pqsB* or *pqsD* genes completely abolish the production of AQs ([@b46]; [@b221];). In addition to these four AQ biosynthesis genes, the *pqsABCDE* operon also encodes PqsE, which has sequence similarities to proteins of the metallo-β-hydrolase superfamily. This extensive family of hydrolytic enzymes mediates a wide range of functions, such as β-lactamases, glyoxalases, AHL-lactonases and arylsulphatases. These enzymes are usually characterized by a conserved metal ion-binding HXHXDH amino-acid motif, which is also found in PqsE and that forms an active site able to bind two iron atoms ([@b218]). However, although the PqsE the crystal structure is available, little knowledge has been gained about its exact function, its natural substrate remaining unknown. The deletion of *pqsE* reduces the production of several virulence factors including pyocyanin, lectin and hydrogen cyanide (HCN), while overexpression of *pqsE* has the opposite effect. A transcriptomic analysis has recently revealed that the abundances in the mRNAs of ≈400 genes depend on the level of expression of *pqsE*, and that virulence in plant and animal infection models in the absence of AQ depends on this gene ([@b154]). The activity of PqsE has also been reported to be dependent on RhlR, which acts downstream, but in synergy with PqsE ([@b76]). Interestingly, PqsE is not involved in AQ biosynthesis, and appears to be a crucial element mediating the cellular response to PQS to achieve full virulence ([@b64]; [@b46]; [@b44]; [@b59];). The *pqsH* gene encodes a predicted FAD-dependent monooxygenase that hydroxylates the 3′ carbon atom of HHQ in the final step of PQS biosynthesis. As such, *pqsH* mutants do not produce 3-hydroxylated quinolones, but continue to produce other AQs ([@b43]). Because *pqsH* is regulated by LasR, but not by PqsR, it is therefore conceivable that under certain circumstances, a differential regulation of these two elements may lead to the overproduction of HHQ with respect to PQS, due to a lack of PqsH ([@b43]). The *pqsL* gene encodes a second, distinct monooxygenase that is required for the synthesis of HQNO and related *N*-oxides via oxidation of the quinolone ring nitrogen atom. The detailed mechanism of AQNO biosynthesis is still unknown, but interestingly, HHQ does not appear to be a precursor of HQNO, as the addition of deuterated HHQ to a culture of *P. aeruginosa* resulted in the biosynthesis of deuterated PQS, but not of deuterated HQNO ([@b43]). Additionally, a *pqsL* mutant overproduces PQS compared with its isogenic wild-type parent, suggesting that PqsL interacts with and diverts a fraction of the HHQ precursor products towards AQNO biosynthesis and away from HHQ and PQS biosynthesis ([@b38]). A simplified scheme for the biosynthesis of AQs is detailed in [Fig. 4](#fig04){ref-type="fig"}. Before the role of these molecules in signalling was discovered, an AQ biosynthesis pathway had already been proposed. This was based on radiolabelled precursor feeding experiments, which indicated condensation of anthranilic acid with β-keto fatty acids, releasing CO~2~ and H~2~O ([@b34]; [@b114]; [@b158];). This was confirmed more recently by MS and nuclear magnetic resonance (NMR) analysis of the AQs produced after feeding ^13^C and ^15^N isotope-labelled precursors to *P. aeruginosa* ([@b13]). This study also ruled out a second possible pathway of AQ biosynthesis that involved the formation of a kynurenic acid precursor resulting from a reaction between orotic acid and anthranilate. The heterologous expression of AQ biosynthesis genes in *E. coli* revealed that the production of DHQ ([Fig. 2](#fig02){ref-type="fig"}) only requires PqsA and PqsD ([@b221]). Activated anthraniloyl-CoA, generated by PqsA, is transferred to the cysteine residue in the active site of PqsD (unactivated anthranilate does not transfer). Here, it reacts with either malonyl-CoA or malonyl-ACP to form 3-(2-aminophenyl)-3-oxopropanoyl-CoA, a short-lived intermediate that undergoes an internal rearrangement to form DHQ. Some variation of this biosynthesis, utilizing longer chain β-keto fatty acids in place of malonyl-CoA or malonyl-ACP, is possibly the mechanism by which AQ molecules such as HHQ are produced. However, because the above process only utilizes PqsA and PqsD, AQ biosynthesis is likely to be more complex as the functions of PqsB and PqsC are still unclear ([@b221]; [@b10];). ::: {#fig04 .fig} Fig. 4 ::: {.caption} ###### Proposed biosynthetic pathway of PQS, HHQ, HQNO and DHQ in *Pseudomonas aeruginosa*. AQs are derived from a condensation reaction between anthranilate and β-keto fatty acids. Anthranilate is derived from either the PhnAB/TrpEG or the KynABU metabolic pathways using either chorismate or tryptophan as precursors, respectively. Anthranilate is first activated with coenzyme A (CoA) by PqsA. Anthranilate-CoA and an activated β-ketodecanoate are condensed, possibly via the PqsBCD enzymes to HHQ, releasing CO~2~ and H~2~O. The monooxygenase PqsH converts HHQ to PQS. HQNO is derived from the same starting products as HHQ, but utilizes the additional monooxygenase PqsL. HHQ is not a precursor for HQNO. DHQ, which technically is not an AQ, is produced by PqsD independent of PqsB and PqsC. :::  ::: The substrates required for AQ biosynthesis have also been investigated. Two pairs of genes responsible for anthranilate biosynthesis had been described previously: *phnAB* ([@b54]), located adjacent to the *pqs* operon ([@b64]), and *trpEG*, which encode enzymes involved in tryptophan biosynthesis ([@b55]). The *phnAB* genes code for proteins resembling the *E. coli* anthranilate synthase subunits TrpE and TrpG ([@b54]) and are cotranscribed by PqsR ([@b22]). It was initially thought that these would provide anthranilate as a precursor for phenazine biosynthesis as inactivation of these genes reduced pyocyanin production. However, it was subsequently shown that PhnA and PhnB do not appear to be involved in this function ([@b125]) and therefore the reduction in pyocyanin production in a *phnAB* mutant is instead most likely to be the consequence of the reduced availability of AQs. In addition to PhnA and PhnB, TrpE and TrpG also direct the synthesis of anthranilate from chorismate, which can then be utilized either for tryptophan or for AQ biosynthesis. A third source of anthranilate comes from the homologues of the tryptophan 2,3-dioxygenase KynA, the kynurenine formamidase KynB and the kynureninase KynU to produce anthranilate from tryptophan ([@b93]; [@b58];). Detailed analyses indicate that in rich growth media containing the aromatic amino acid tryptophan, the kynurenine pathway is the main source of anthranilate for AQ production, whereas the *phnAB* genes supply anthranilate in minimal media in the absence of exogenous tryptophan ([@b58]). Interestingly, increased PQS production by *P. aeruginosa* strains isolated from infected CF lungs has been correlated with the presence of aromatic amino acids in the growth medium ([@b139]). In this case, the transcription of *pqsA* was found to be induced by tryptophan, phenylalanine and tyrosine, while the nonaromatic amino acid serine had little effect. The kynurenine pathway may therefore be the principal source of anthranilate in a lung infection context. The requirement of anthranilate for PQS production has been demonstrated. When *P. aeruginosa* PAO1 parent and isogenic *las* QS mutants unable to produce PQS were grown in the presence of anthranilate labelled with ^14^C in the heteroaromatic ring, most of the radioactivity was found in the AQ extracts for those strains able to generate PQS, whereas very little was found in the supernatant extracts of the QS mutants ([@b18]). This suggested that the strains not producing PQS would incorporate anthranilate, but not convert it into AQs. Additionally, when *P. aeruginosa* was grown with increasing amounts of methyl-anthranilate, PQS biosynthesis levels were reduced as this compound acted as a competitor of anthranilate ([@b18]). The production of elastase, which is dependent on PQS signalling, was also inhibited by methyl-anthranilate in a concentration-dependent manner. At 1.5 mM, methyl-anthranilate practically abolished elastase production, the suggested consequence of a much reduced level of PQS production. Feeding experiments with isotope-labelled AQ precursors such as ^15^N-anthranilate coupled with GC--MS analysis resulted in the production of AQs having incorporated around 66% of ^15^N, further demonstrating that anthranilate serves as a common precursor for AQs and that the heteroaromatic nitrogen in the quinolone ring originates from this molecule ([@b13]). Similarly, feeding labelled ^13^C-acetate to *P. aeruginosa* PAO1 demonstrated that the heteroaromatic ring of the quinolone moiety was formed from acetate. The resulting GC--MS fragmentation pattern, together with confirmation by NMR spectroscopy, indicated that the mechanism of this reaction was via a direct head-to-head reaction involving anthranilate and β-keto fatty acids derived from acetate ([@b13]). β-Keto fatty acids are therefore essential precursors in the biosynthesis of AQs. Some studies had suggested that there is a link between rhamnolipid biosynthesis and AQ production, which was interesting because rhamnolipids are composed of a rhamnose moiety and fatty acids of the same chain lengths as those involved in AQ biosynthesis. Rhamnolipids have also been shown to increase PQS solubility and may mediate this function *in vivo* ([@b19]). It was initially thought that *rhlG* coded a potential β-ketoacyl-ACP reductase that could participate in the provision of fatty acids utilized as a substrate for AQ biosynthesis ([@b13]) as RhlG was assumed to direct the incorporation of these fatty acids into rhamnolipids ([@b21]; [@b42]; [@b175];). However, recent studies have contradicted this, as an *rhlG* mutant was unaltered in rhamnolipid production compared with the corresponding wild type ([@b222]). Furthermore, the crystal structure of RhlG revealed that its function was inconsistent with the proposed fatty acid biosynthetic pathway ([@b129]). Therefore, it appears that RhlG is not involved in rhamnolipid or AQ biosynthesis. In *P. aeruginosa*, PQS is likely to be the end product of the AQ synthetic pathway or is not substantially converted into other molecules, as when labelled PQS was added to wild-type cultures, no additional compounds could be identified ([@b43]). Quorum sensing (QS) and AQ production in *P. aeruginosa* ======================================================== When favourable nutritional conditions are encountered, bacteria will proliferate to form established multicellular communities that have the potential to adapt to and modify their environment. This allows further exploitation of nutrient resources that would otherwise be restricted for individual cells. The mechanism by which a bacterium adapts from the lifestyle of an individual cell to a community capable of modifying their environment has been termed QS and it is defined as a mechanism by which bacteria regulate specific target genes in response to a critical concentration of endogenously produced signal molecules dedicated to the probing of the cell population density ([@b191]; [@b207];). This process is mediated by the production and sensing of autoinducers, small signalling molecules, whose concentration in the extracellular medium reflects cell population density. *Pseudomonas aeruginosa* produces two AHLs as QS signal molecules, each acting as the autoinducer of a specific sensing and responding system: 3-oxo-C12-HSL acts on the *las* system and C4-HSL acts on the *rhl* system. The core of each system is composed of a synthase producing an AHL for the activation of a specific transcriptional regulator: LasI produces 3-oxo-C12-HSL for the activation of LasR ([@b65]; [@b141]; [@b143];) and RhlI produces C4-HSL for the activation of RhlR ([@b98]; [@b144]; [@b208];). Initially, each synthase gene is expressed at basal levels and the AHLs produced diffuse into the surrounding medium. Autoinduction is achieved when the accumulation of an AHL reaches a threshold concentration and the activated transcriptional regulators LasR and RhlR further enhance the expression of the synthase genes *lasI* and *rhlI*, respectively, generating positive feedback loops ([@b170]). When the transcriptional regulators are activated they will induce the transcription of overlapping subsets of genes. For example, LasR will induce the production of virulence factors such as elastase ([@b141]) and pyoverdin ([@b177]), while RhlR will increase the production of rhamnolipid biosurfactants ([@b135]), cytotoxic lectins, pyocyanin and elastase, among other virulence factors ([@b145]). In addition to some overlap between the genes targeted by both AHL QS systems due to the similarities of the palindromic *las*/*rhl* boxes recognized by LasR and RhlR ([@b168]; [@b166], [@b167]), activated LasR will also induce the *rhl* system ([@b97]), creating a hierarchical regulatory network, which in turn is further modulated by additional regulatory elements (reviewed in [@b194] and in [@b206]). Besides the AHL-based QS systems, *P. aeruginosa* utilizes an autoinducer regulatory system based on the AQs. This system relies on the PQS and its precursor molecule HHQ to control global gene expression ([@b148]; [@b43];). The transcriptional regulator PqsR controls the expression of the *pqsABCDE* and *phnAB* biosynthetic operons and therefore *pqsR* is essential for the production of AQs ([@b64]; [@b43]; [@b126];). The *pqsR* gene is convergently transcribed with respect to the *pqsABCDE*-*phnAB* operons and two transcriptional start sites have been mapped 190 and 278 bp upstream of its start codon ([@b195]). The distant promoter appears to have a typical σ^70^-binding site signature, indicative of basal transcription, and a putative *las*/*rhl* box operator sequence is found centred 239--258 bp upstream of this transcriptional start site (517--536 bp upstream of the start codon). *In vitro*, PqsR binds at two different locations upstream of *pqsA*, and the strength and position of the binding depend on the presence of PQS ([@b195]). The *pqsA* transcriptional starting point has been mapped 71 bp upstream of the start codon ([@b126]). Alterations of a LysR-type box located at −45 in the *pqsA* promoter can result in the loss of PqsR-binding capacity and in the reduction of transcription initiation, suggesting that this element plays a central role in the regulation of the *pqsABCDE* operon by PqsR and PQS ([@b212]). Overexpression of *pqsR* strongly repressed the transcription of *antA*, which encodes an anthranilate 1,2-dioxygenase. This is thought to ensure an adequate supply of anthranilate for the biosynthesis of AQs by reducing its metabolic degradation ([@b136]). When the *pqsABCDE* operon and *pqsR* were cloned in *E. coli* and expressed from their native promoters, HHQ and NHQ were produced, but not PQS because *E. coli* lacks a *pqsH* homologue. Similarly, compared with the wild type, the activity of the *pqsA* promoter and AQ production levels (except for PQS) remained comparable when *pqsH* was disrupted. This indicates that in addition to PQS, other AQs can also act as autoinducers ([@b211]). It has been suggested that HHQ induces a conformational change in PqsR, as binding of PqsR to the *pqsA* promoter *in vitro* is enhanced by HHQ, although not as much as with PQS. In an AQ-negative double *pqsA pqsH* mutant derived from strains PAO1 or PA14, PQS was found to be 100 times more potent at inducing the *pqsA* promoter than HHQ ([@b211]; [@b49];). In strain PA14, the deletion of *pqsH* reduced the overall expression of the *pqsR* regulome by less than twofold, and the addition of exogenous PQS to this mutant did not revert the expression levels of this regulome substantially above wild-type levels, further implying a role for HHQ in inducing many of the genes. An exception to this was *phzA1*, as PQS appears to be essential for the transcription of this gene and for the production of pyocyanin ([@b211]). Altogether, these studies indicate that HHQ acts as an autoinducer independent of PQS. Other AQs such as NHQ can also activate PqsR and as such could potentially be considered as autoinducers, although not as potent as PQS ([@b211]; [@b60];). The *las* and *rhl* QS systems are linked to AQ production and regulation, forming an incoherent feed-forward loop likely to produce accelerated pulse-like responses ([@b5]): the *las* system positively controls AQ production by inducing the *pqsR* and *pqsA* promoters and the *rhl* system downregulates its effects ([@b148]; [@b127]; [@b126]; [@b195]; [@b212];) ([Fig. 5](#fig05){ref-type="fig"}). In a *lasR* mutant, transcription of *pqsR* is reduced about fourfold compared with the wild type ([@b195]) and LasR appears to induce *pqsR* transcription by binding to a conserved *las*/*rhl* box situated 517--536 bp upstream of its translational start site ([@b126]; [@b212]; [@b68];). In line with this, a transcriptional *pqsR*-*lacZ* fusion can be significantly induced in *E. coli* expressing *lasR* by the addition of 3-oxo-C12-HSL, indicating that the LasR/3-oxo-C12-HSL system acts as an inducer of *pqsR* ([@b195]). A *lasR* mutant accumulates the HHQ series of AQs, but produces very little PQS early in growth, a consequence of LasR also positively controlling the expression of *pqsH*, which encodes the monooxygenase required for the conversion of HHQ to PQS ([@b203]; [@b64]; [@b43];). The transcription of *pqsA* is considerably reduced in a *lasI* mutant ([@b126]). However, a functional *las* QS system is not required for AQ biosynthesis, as a *lasR* mutant still produces PQS in the late stationary phase and expressions of *pqsR* and *pqsH* in a *lasR* mutant are delayed, but not abolished during growth ([@b46]; [@b212];). As *rhlR* overexpressed from a plasmid partially overcomes the delay in PQS production caused by a *lasR* mutation in strain PA14 ([@b40]), it appears that RhlR could replace some of the functions of LasR with respect to the *pqsA* and *pqsH* promoters to induce the production of PQS, although this is somewhat paradoxical because RhlR is generally considered to be a repressor of AQ production and indicates that the current LasR-RhlR-AQ QS hierarchy model in *P. aeruginosa* may be somewhat more sophisticated than currently thought. ::: {#fig05 .fig} Fig. 5 ::: {.caption} ###### Regulation of AQ production in*Pseudomonas aeruginosa*. The *las* QS system positively regulates the transcription of *pqsR*, *pqsABCDE* and *pqsH*. The PqsABCD proteins synthesize HHQ, which is converted to PQS by PqsH. Autoinduction occurs when either HHQ or PQS binds to PqsR and enhances the expression of the *pqs* operon. The *rhl* QS system, also positively controlled by the *las* system, exerts a negative effect on the AQ system, although it is itself positively regulated by AQs. The terminal output of this regulatory network is the PqsE protein of still unknown enzymatic function. In addition, PQS, via an unknown mechanism, positively controls the transcription of the small RNA *RsmZ*, which in turns has a negative effect on the RNA-binding protein RsmA involved in post-transcriptional regulation. Biosynthetic enzymes are represented by globular shapes, while transcriptional regulators are shown as cubes. Filled arrows and blunted lines represent positive and negative regulation, respectively. :::  ::: While the *las* QS system positively regulates AQ and PQS production, the *rhl* system acts as a negative modulator of their regulatory effects ([Fig. 5](#fig05){ref-type="fig"}). A 50% increase in *pqsR* transcription has been observed in an *rhlR* mutant, suggesting in this case that RhlR has a repressive effect ([@b195]). Similarly, transcription from the *pqsA* promoter is enhanced in an *rhlI* mutant and addition of C4-HSL to antagonize the induction of *pqsA* by 3-oxo-C12-HSL, with the consequence of reducing the production of PQS ([@b126]). Two *las*/*rhl* boxes are found at 311 and 151 bp upstream of the *pqsA* transcriptional start site ([@b212]). Deletion of the distal *las*/*rhl* box in this promoter increases transcription, while additional deletion of the proximal box does not further increase *pqsA* promoter activity. The deletion of *rhlR* causes an increase in the transcription of *pqsA* independent of the presence of the −311 box, suggesting that RhlR binds to this box and causes a downregulation of the *pqsA* promoter, whose mechanism is still unclear. *In vitro* electrophoretic mobility shift assays carried out on a 253-bp DNA fragment containing part of the *pqsA* promoter using lysates of *E. coli* producing RhlR in the presence or absence of C4-HSL-RhlR did not indicate binding to this region; however, the fragment used did not include the *las*/*rhl* box situated 311 bp upstream of the transcriptional starting point ([@b195]). Identification of LasR targets *in vivo* using chromatin immunoprecipitation coupled to DNA microarray hybridization (ChIP-chip) identified this distal *las*/*rhl* box as a LasR-binding site ([@b68]). As the *rhl* system is itself driven by the production of PQS, a negative autoregulatory feedback loop is formed ([@b46]). The simultaneous provision of exogenous C4-HSL and PQS restores *rhlI* transcription levels in a *lasR* mutant comparable with the wild type. However, under the same conditions, the addition of these molecules separately did not cause increased *rhlI* transcription, suggesting a synergistic mechanism involving the two signalling molecules ([@b127]). Thus, in *P. aeruginosa*, the autoinducible AQ system is upregulated by the *las* and downregulated by the *rhl* QS systems. AQ production is furthermore indirectly self-limited by the positive regulatory effects it exerts on the *rhl* QS system ([Fig. 5](#fig05){ref-type="fig"}). Regulation of virulence factor expression by AQs ================================================ The first demonstration that AQs regulate virulence factor production in *P. aeruginosa* was that PQS positively controlled the expression of the *lasB* (elastase) gene ([@b148]). It was later shown that this effect was considerably enhanced when PQS and C4-HSL acted synergistically to upregulate *lasB* expression ([@b127]). The regulation of virulence factor production by AQs is not restricted to elastase. Addition of PQS upregulates the expression of *lecA* and pyocyanin production in a concentration-dependent manner ([@b46]). However, PAO1 cultures growing in the presence of PQS above a concentration of 100 μM had an extended lag phase and reached reduced ODs at the stationary phase. Despite this, the expression of *lecA* occurred at lower population densities and therefore the maximal expression was still observed during the early stationary phase, although the addition of PQS still resulted in an advancement of *lecA* expression and elastase and pyocyanin production into the logarithmic phase ([@b46]). It worth noting that these effects were not seen when either HHQ or 3-formyl-HHQ instead of PQS was added to the cultures. Previous studies found that both RhlR and RpoS are essential for *lecA* expression ([@b209]) and addition of PQS failed to restore *lecA* transcription in *rhlR* or *rpoS* mutants, confirming the importance of these two regulators for *lecA* promoter activity. However, PQS was able to overcome the repression of *lecA* by the H-NS-type protein MvaT and the post-transcriptional regulator RsmA ([@b46]). We have recently found that PQS, but not HHQ, induces the transcription of the small regulatory RNA RsmZ ([Fig. 5](#fig05){ref-type="fig"}), a mechanism that explains how post-transcriptional regulation by RsmA can be overcome by PQS and reveals that this molecule can act on the expression of virulence genes at both the transcriptional and the post-transcriptional levels (S. Heeb *et al*., unpublished data). Virulence factor production is also affected when AQ production is inhibited. Addition of the anthranilate analogue, methyl-anthranilate, to *P. aeruginosa* caused a decrease in the production of PQS and a subsequent reduction in elastase produced ([@b18]). The effects observed with methyl-anthranilate are not restricted to elastase, with concentrations of 500 μM completely inhibiting the expression of *lecA* and pyocyanin production, but with no adverse effect on growth. This effect could partially be restored by the provision of exogenous PQS ([@b46]). The subset of genes regulated by AQs has now been examined in greater detail using transcriptomic analysis. It has been found that PqsR, through the induction of the *pqsABCDE* operon and the action of PqsE, positively regulates a subset of LasR- and RhlR-dependent genes. A *pqsR* mutant of strain PA14 displayed the upregulation of 121 and the repression of 22 mRNAs when compared with the corresponding wild type ([@b44]). In this *pqsR* mutant, the transcription of the *pqsABCDE* and *phnAB* operons was abolished and that of *pqsR* itself was reduced. The transcription of the *phz1* operon, *hcnABC*, *chiC* (chitinase), *mexGHI*-*opmD*, *lecA* and *lecB* was also found to be reduced in the absence of PqsR. However, the role of AQs in the regulation of virulence gene expression is now the subject of some debate. It had been demonstrated previously that in both *pqsR* and *pqsE* mutants, pyocyanin production, *phzA1* expression, LecA, elastase and rhamnolipid production levels were considerably reduced compared with the wild type ([@b22]; [@b64]; [@b46]; [@b44];) and that the addition of PQS, HHQ or HQNO to these mutants could not restore these phenotypes ([@b64]; [@b46]; [@b44];). Altogether, these studies suggested that AQ production may not be indispensable for the regulation of these phenotypes. A subsequent study shed new light on the mechanisms by which AQs induce gene transcription by revealing that PqsE alone can drive the expression of the target genes, through the *rhl* QS system ([@b59]). By expressing PqsE in AQ-negative *pqsA* or *pqsR* mutants, it was demonstrated that pyocyanin, rhamnolipid and elastase production could be restored in the absence of AQs. This restoration of exoproducts was not observed in an *rhlR* mutant, which suggests that PqsE may exert its effects through the *rhl* system ([@b59]). These findings raise a number of intriguing questions as to the function of AQs in *P. aeruginosa*. For example, is the primary function of both, PQS and HHQ, to bind PqsR and to upregulate the *pqsABCDE* operon, thereby forming, on the one hand, an autoinduction loop and ultimately, on the other, producing as the major output, an increase in the levels of PqsE? Another intriguing question raised by the data is about the function of PQS itself. There are conflicting reports on the necessity of PQS for virulence in different *P. aeruginosa* wild-type strains, although different hosts have been used: PQS has been shown to be necessary for the virulence of strain PAO1 in nematodes ([@b64]), but unnecessary for PA14 in a burned mouse model ([@b211]). There are also conflicting reports as to the efficacy of PQS at inducing the *pqsA* promoter via PqsR. One study found that in PA14, PQS was more effective than HHQ at upregulating *pqsA* ([@b211]), but conversely, another study demonstrated that in strain PAO1, PQS was the less effective molecule ([@b60]). This contradiction may be due to differences in strain-specific mechanisms, but taken together with new research on the role of PqsE, PQS may not be as important to the direct regulation of virulence factors in *P. aeruginosa* as first envisioned and may have evolved as a fortuitous byproduct with other functions ([@b14]; [@b49];). Furthermore, the primary role of PqsR requires some further clarification. It is probable that the loss of virulence noted in *pqsR* mutants ([@b22]; [@b44];) is primarily due to the corresponding loss of PqsE production, seen in the fact that mice mortality in strain PA14 was much decreased from the wild type and was equivalent in both *pqsA* and *pqsE* mutants ([@b44]). Therefore, the primary role of PqsR may be that it is responsible for the expression of *pqsE* via the production of AQs and the corresponding autoinduction of the *pqsABCDE* operon, at least as far as the production of pyocyanin and expression of *lecA* are concerned. The induction of pyocyanin production by the AQ QS system further leads to the regulation of the PYO stimulon, a set of around 50 genes whose expression is affected, primarily via the transcriptional regulator SoxR, by this phenazine ([@b45]). Role of AQs in iron metabolism ============================== In addition to its role as a cell-to-cell signalling molecule, PQS is also able to chelate ferric iron (Fe^3+^). The presence of the 3′-hydroxy group on the molecule mediates this and allows two or three PQS molecules to bind Fe^3+^ at physiological pH ranges of 6--8. Compounds similar to PQS (such as C9-PQS) also possess iron-binding capabilities, but molecules lacking the 3-hydroxy group such as HHQ are unable to do so ([@b13]; [@b49];). Addition of PQS to *P. aeruginosa* cultures upregulates the genes involved in the production of the siderophores pyoverdine and pyochelin, which are produced in response to iron starvation, as indicated by the upregulation of siderophore-mediated iron transport systems such as the pyochelin biosynthetic clusters (*pchDCBA* and *pchEGF*), the iron pyochelin outer-membrane receptor *fptA* and the pyoverdine genes *pvdE* and *pvdS* ([@b13]; [@b49];). The *pch* genes were upregulated at 5, 11 and 20 h after inoculation between 3- and 25-fold. Also, the genes *pvdJAD* encoding pyoverdine synthetases were upregulated between 2- and 10-fold at 11 and 20 h ([@b13]). Quantitative real-time PCR showed that *pvdA* and *pchE* are upregulated by 6- and 17-fold, respectively, upon addition of 20 μM PQS ([@b49]). Also, in strain PAO1 wild type as well as in *pqsA*, *pqsE* or *pqsR* mutants, the addition of PQS, but not of HHQ, strongly induced pyoverdine production ([@b49]). PQS, with its effect on free iron levels, also affects the transcription of other genes. During growth in iron-replete media, both *lecA* and *pqsA* were strongly induced by the addition of 50 μM PQS in a PAO1 *pqsA* mutant. However, the induction of *pqsA* was not due to the iron-chelating properties of PQS because when grown in an iron-deficient casamino acid (CAA) medium, PQS, PQS--Fe^3+^ (3 : 1) and HHQ all induced the *pqsA* promoter, but methyl-PQS did not ([@b49]). Around 60% of the PQS produced by *P. aeruginosa* is associated with the cell envelope ([@b101]; [@b49];), and the membranes of cells grown in iron-rich media are visibly pink due to complexed Fe^3+^, possibly stored in AQ-containing inclusion bodies ([@b163], [@b164]). Therefore, there is the possibility that PQS could act as an iron trap and storage molecule in the cell membrane and that it may be able to deliver iron directly to the cells. However, experiments carried out with a *P. aeruginosa pvdD*/*pchEF* double mutant, which lacks any iron acquisition systems, revealed that it was unable to grow in an iron-deficient CAA medium in the presence of added PQS. In contrast, this mutant had a similar growth compared with the parental PAO1 strain when exogenous PQS was not added to the medium. These data suggest that although PQS may trap iron in the cell membrane, it is unlikely that it can act as a siderophore *per se* ([@b49]). Iron-dependent regulation of AQ production appears to be controlled by the availability of one of their precursors, anthranilate. Under iron-limiting conditions, the ferric uptake regulator Fur does not repress the transcription of two genes *prrF1* and *prrF2*, encoding small regulatory RNAs ([@b204]), which post-transcriptionally repress the expression of the *antABC* and *catBCA* operons specifying enzymes for the degradation of anthranilate. Hence, in a *prrF1 prrF2* double mutant, PQS production is abolished under iron-limiting conditions, probably as a consequence of anthranilate depletion ([@b136]). Therefore, under iron-limiting conditions, the supply of anthranilate for the biosynthesis of AQs is controlled by Fur and the PrrF sRNAs, an effect that was further reinforced by the iron starvation response resulting from the iron-chelating property of PQS ([@b14]; [@b49];). Because HHQ performs functions similar to those of PQS, such as the induction of the *pqsA* promoter ([@b211]; [@b49];), many of the specific effects observed upon addition of PQS may be due to its iron-chelating properties. It is also probable that the production of PQS and its chelating effects could confer a survival advantage when *P. aeruginosa* is growing with other competing microorganisms in iron-limited environments. The red-coloured PQS--Fe^3+^ complex can also be toxic to other organisms. For example, its production has been found to confer the 'red death' lethal phenotype to *P. aeruginosa* in a *Caenorhabditis elegans* infection model ([@b219]). Iron availability also influences the levels at which AQs induce the activity of PqsR as a transcriptional activator, and therefore, iron also acts directly as a modulator of the AQ signalling system in *P. aeruginosa* ([@b76]). Interestingly, iron has also been found bound to PqsE, although without knowledge of the function of this enzyme, the biological significance of this remains unclear ([@b218]). Additional regulators of AQ production in *P. aeruginosa* ========================================================= Besides autoinduction by PQS and its precursor HHQ, modulation by the *las* and *rhl* QS systems, and metabolic and regulatory adjustments following iron availability, AQ production is regulated by additional factors ([Table 1](#tbl1){ref-type="table"}). For example, AQ production is enhanced under phosphate-limiting conditions. In *P. aeruginosa*, the transcriptional regulator PhoB mediates responses to phosphate limitation ([@b6]). As a PHO box has been found overlapping the distal transcriptional starting point of *pqsR* and as AQ production is no longer enhanced in a *phoB* mutant, these elements may mediate the increased AQ production observed following phosphate limitation ([@b87]). ::: {#tbl1 .table-wrap} Table 1 ::: {.caption} ###### Factors influencing AQ production ::: Factors Mechanisms References ------------------------ ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ------------------------------------------------------------------------------------------------------- PqsR and AQs Some AQs positively regulate AQ biosynthesis by autoinduction. PqsR binding to the *pqsA* promoter is enhanced and the transcription of the *pqsABCDE* operon is induced by PQS, and to different levels by other AQs such as HHQ and NHQ [@b148], [@b64], [@b43], [@b126], [@b195], [@b211],[@b212];, [@b49], [@b60], [@b136] *las* QS system LasR and 3-oxo-C12-HSL positively regulate AQ production by inducing the transcription of *pqsR* and *pqsABCDE*, and further enhances the production of PQS by inducing *pqsH* [@b148], [@b203], [@b127], [@b64], [@b46], [@b43], [@b126], [@b195], [@b211],[@b212];, [@b40], [@b68] *rhl* QS system RhlR negatively regulates AQ production by repressing *pqsR* transcription. The expression of *rhlI* and the biosynthesis of C4-HSL also negatively affect the activity of the *pqsA* promoter [@b127], [@b46], [@b126], [@b195], [@b211],[@b212]; Fur and Fe^3+^ Under low iron conditions, the metabolism of anthranilate is adjusted by Fur and the PrrF sRNAs, maintaining AQ production. Iron saturation increases AQ production, probably by inducing the kynurenine pathway leading to anthranilate. Iron levels also affect the activities of AQs as inducers of PqsR [@b136], [@b76] PhoB and PO~4~^3−^ AQ production is enhanced by phosphate limitation. PhoB could be mediating this by binding to a PHO box present in the *pqsR* promoter [@b87] PtxR Reduces the expression of the *pqsABCDE* operon independently of *pqsR*. PtxR also acts positively on the *las* and negatively on the *rhl* QS systems [@b23] PmpR PmpR negatively affects the transcription of *pqsR* by binding to its promoter [@b108] PpyR PpyR appears to be essential for the transcription of *pqsABCDE* and *psqH* [@b8] Dynorphin κ-opioid receptor agonists dynorphin and U-50,488 enhance AQ production by inducing the *pqsA* promoter [@b220] Farnesol Reduces *pqsA* transcription and AQ production by interfering with PqsR [@b35] Indole and derivatives Indole, its oxidation products and other bicyclic compounds, including some naphthalene analogues and 8-quinolinol, inhibit MV formation and PQS synthesis by unknown mechanisms [@b181] Sputum Growth in sputum, rich in aromatic amino acids such as tryptophan, induces the *pqsA* promoter and increases AQ production [@b139], [@b58] ::: PtxR is a transcriptional regulator that positively affects the production of exotoxin A and negatively affects the production of pyocyanin. PtxR reduced the expression of the *pqsABCDE* operon, probably indirectly and not via the repression of *pqsR* ([@b23]). However, PtxR also regulated the *las* positively and the *rhl* QS systems negatively, which paradoxically should have resulted in an induction of the *pqsA* promoter. Therefore, it appears that PtxR could be part of an intricate network of feed-forward loops ([@b5]) that connect the *las*, *rhl* and *pqs* QS systems. The gene *pmpR* (*pqsR*-mediated PQS regulator, PA0964) was found by screening transposon mutants for clones in which the *phzA1* promoter had altered expression profiles. PmpR, a protein of the YebC-like superfamily, binds to the *pqsR* promoter and affects its transcription negatively. A *pmpR* mutant was therefore found to have increased mRNA levels of *pqsR*, *pqsA* and *pqsH*, which was suggested to result in the observed induction of the *phzA1* promoter and of pyocyanin production, and enhanced swarming motility and biofilm formation ([@b108]). The gene *ppyR* (*psl* and pyoverdine operon regulator, PA2663) appears to encode a membrane sensor that positively regulates exopolysaccharide and pyoverdine production, perhaps in response to the presence of NOR ([@b8]). The deletion of *ppyR* caused the downregulation of several genes including the *pqsABCDE* operon and the *psqH* gene for AQ and PQS biosynthesis, as well as the *antABC* operon for anthranilate degradation. As a consequence, a *ppyR* mutant produced no detectable PQS. However, the mechanism by which PpyR exerts its effects or the signals that it senses are unknown. Impact of AQs on microbial interactions ======================================= The prominence of *P. aeruginosa* as a major opportunistic pathogen in nosocomial infections and in lung infections in CF patients ([@b70]) led to the investigation of the role of AQs in the regulation of virulence factor production and the establishment and severity of infection. The initial studies using clinical isolates from sputum in CF patients showed that they all produced PQS ([@b31]) and also HHQ, HQNO, NQNO and UQNO ([@b115]). In addition, there was a correlation between the levels of PQS and the bacterial sample load. Furthermore, PQS was also found in isolates from paediatric CF patients and from patients at early stages of *P. aeruginosa* infection ([@b71]). The regulation of PQS production in some of these isolates was irregular as this molecule was detected early in growth, during the log phase. The use of a simulated CF sputum medium has been shown to support the growth of *P. aeruginosa* to high population densities ([@b139]) and also the differential regulation of PQS production. In particular, the expression of the *phnAB* genes is induced 14--22-fold, in line with the upregulation of the *pqsABCDE* operon (17--19-fold) compared with the expression of these genes in a morpholinepropanesulphonic acid-buffered glucose medium. This upregulation results in a fivefold increase in PQS production and presumably the other AQs, and is not triggered by changes in AHL levels. It is possibly linked to the presence of aromatic amino acids in the sputum medium such as tryptophan, which is used by *P. aeruginosa* for anthranilate biosynthesis ([@b58]). A recent study revealed that although PhhR is an aromatic amino acid-responsive transcriptional regulator that controls genes involved in phenylalanine and tyrosine catabolism in *P. aeruginosa*, the biosynthetic genes for AQs are not differentially expressed according to the presence of this regulator ([@b138]). *Pseudomonas aeruginosa* forms biofilms to protect itself from the harsh environmental conditions generated by the host immune system and antimicrobials. AQs play an important role in the establishment and maintenance of the biofilm lifestyle by a number of different mechanisms. The exogenous addition of 60 μM PQS to growing cultures of *P. aeruginosa* PAO1 resulted in a significant enhancement in biofilm formation partly due to the induction of expression of the lectin gene *lecA* ([@b46]) as this gene plays a role in maintaining biofilm architecture in this organism ([@b48]). *Pseudomonas aeruginosa* can also release, possibly through lysis of cell subpopulations, extracellular DNA, which acts as an interconnecting matrix in bacterial biofilms ([@b202]). DNA has cation-chelating and antimicrobial properties and can cause the disruption of the bacterial outer membrane by chelating Mg^2+^, which is essential for membrane stability. This in turn could result in more DNA release ([@b133]). In addition, Mg^2+^ chelation induces the expression of the PhoPQ two-component system, increasing the resistance of *P. aeruginosa* towards aminoglycosides such as gentamicin and cationic antimicrobial peptides. These broad-spectrum antimicrobial peptides are released from host immune cells and can disrupt the bacterial outer membrane, causing cell death. Maximum DNA release takes place in the late log phase when PQS production is at its highest ([@b46]; [@b101];). Similarly, a *pqsA* mutant releases low levels of extracellular DNA and forms flat, thin unstructured biofilms with increased sensitivity to detergents. The detergent sensitivity may be due to the loss of this extracellular DNA as a wild-type biofilm treated with DNase retains this sensitivity ([@b4]; [@b74];). A correlation between bacterial cell lysis and PQS levels has been established, which may explain the release of the extracellular DNA observed in biofilms ([@b38]). A mutation in the *pqsL* gene (which results in PQS overproduction) resulted in pronounced lysis in bacterial colonies, whereas those from *pqsA* and *pqsR* mutants displayed no lysis, but this could be restored upon addition of exogenous PQS. It has been proposed that PQS induces prophage-mediated lysis and that this is responsible for the DNA release ([@b38]). The chromosome of *P. aeruginosa* harbours the filamentous Pf4 prophage, whose deletion results in the loss of bacterial autolysis and aberrant biofilm formation ([@b157]). PQS also acts as a pro-oxidant, which can increase the sensitivity of *P. aeruginosa* to peroxide and ciprofloxacin ([@b74]), possibly resulting in cell lysis and DNA release. AQs inhibit the growth of *S. aureus* and the yeast *C. albicans*, suggesting that they may be used as antibiotics by *P. aeruginosa*, during the early stages of infection, enabling it to eradicate any competing organisms ([@b115]). This idea is further supported by the fact that AQs packaged in MVs inhibited the growth of *S. epidermidis* ([@b120]), whereas mutants in *kynAU* were unable to kill *S. aureus* and a *kynB* mutant displayed reduced killing, presumably due to the lack of AQ production ([@b58]). As mentioned earlier (Natural antimicrobial quinolones), both HHQ and PHQ have antibacterial activities, while PHQ additionally presents antialgal properties ([@b210]; [@b113];). HHQ and HQNO produced by a clinical isolate of *P. aeruginosa* inhibited the growth of metronidazole-resistant *H. pylori* in a cross-streak assay ([@b94]). These findings may explain why early colonizers of the CF lung such as *S. aureus* are sometimes absent upon *P. aeruginosa* colonization, which outcompetes other organisms sharing the same niche ([@b115]). Consequently, the combined iron-chelating properties and the impact on virulence factor production of AQs help *P. aeruginosa* to generate a highly favourable environment in which to thrive. Interestingly, farnesol, a sesquiterpene signal molecule produced by *C. albicans*, reduces the transcription of *pqsA* by interacting with PqsR and probably interfering with the normal binding of this transcriptional regulator to the *pqsA* promoter ([@b35]). This results in a decrease in both PQS and pyocyanin production and suggests that this type of interspecies competition can be reciprocal. In addition, HQNO has been shown to induce the formation of persistent small-colony variants of *S. aureus* that may resist *P. aeruginosa* niche colonization and possibly explain the coexistence of these two organisms in some infections ([@b80]). Roles of AQs in infection ========================= The role of AQs in virulence and the severity of infection has been demonstrated using several disease models. A mutation in *phnAB* resulted in a fourfold decrease in virulence compared with the wild type in a wax moth (*Galleria mellonella*) larvae model ([@b86]). In addition, mutations in *pqsC*, *pqsD*, *pqsE*, *pqsR*, *pqsH* and *phnA* resulted in severely reduced killing of the nematode (*C. elegans*) by *P. aeruginosa* to between 37 and 39% of the wild-type levels ([@b64]). Using a burned mouse model, mutants in *pqsA* and *pqsE* also exhibit reduced virulence ([@b44]; [@b154];). In the same disease model, a *pqsR* mutant showed an ∼35% reduced mortality rate compared with the wild type. This mutant also showed reduced PQS, 3-oxo-C12-HSL, pyocyanin, elastase and exoprotein production ([@b22]). Most interestingly, a *pqsH* mutation in *P. aeruginosa* PA14 was not attenuated, suggesting that PQS may not be essential for virulence and that virulence may be regulated via the biosynthetic precursor, HHQ ([@b211]). Virulence factor and AQ production are upregulated in response to host stress responses to *P. aeruginosa*. The synthetic opioid U-50,488 and the endogenous κ-opioid receptor agonist dynorphin, which is released into the human small intestine during inflammation and appears to bind and enter bacterial cells, have been tested in *P. aeruginosa* PAO1 and were found to enhance virulence factor production ([@b220]). Furthermore, the addition of U-50,488 or dynorphin to a PAO1 culture induced a dose-dependent increase in pyocyanin production and enhanced *pqsA* and *lecA* (but not *pqsR*) expression, with a corresponding increase in PQS, HHQ and HQNO production. These opioid agonists also enhanced *P. aeruginosa* virulence against *Lactobacillus* and *C*. *elegans*, probably as a result of the above increases in virulence determinant production. AQs may also interfere with host responses by acting as immune modulators. PQS suppresses T-cell proliferation and interleukin-2 release in concanavalin A-activated human peripheral blood mononuclear cells (hPBMCs). PQS also induces tumour necrosis factor-α release from lipopolysaccharide-activated hPBMCs, at concentrations around 10 μM ([@b81]). *In vitro*, PQS reduces the release of interleukin-12 from lipopolysaccharide-stimulated bone marrow-derived dendritic cells, preventing the development of naïve T cells into T-helper type 1 cells, which promote cell-mediated immunity. The concentration of PQS required to lower the cytokine release to 50% in this case was below 20 μM ([@b173]). Additionally, both HHQ and PQS appear to suppress host innate immune systems by interfering with the nuclear transcription factor-κB signalling pathway. This effect can be achieved with cell-free extracts from cultures of wild-type *P. aeruginosa*, but not from the cultures of a corresponding *pqsA* mutant ([@b92]). It therefore seems possible that AQs play a role in the dysregulation of the host immune response during infection. Production of AQs by other bacteria =================================== DNA database analysis has revealed the presence of *pqs* gene homologues in \>40 species and strains that are more or less related to *P. aeruginosa*. In particular, *Burkholderia pseudomallei* and *B. thailandensis* appear to have the complete putative *pqsABCDE* operons in their chromosomes, sharing 31--53% identity to that of *P. aeruginosa* ([@b47]). These were named *hhqABCDE* as no PQS had been detected in these organisms. The *hhqA* and *hhqE* genes are functionally conserved with their *P. aeruginosa* homologues as they were able to complement PAO1 *pqsA* and *pqsE* mutants, respectively, and restore PQS, HHQ, lectin and pyocyanin production in the *pqsA* mutant and pyocyanin and lectin production in the *pqsE* mutant. Using a combination of a novel AQ bioreporter and LCMS/MS, HHQ was detected in culture supernatants of *Pseudomonas putida* and *Burkholderia cenocepacia* and HHQ, NHQ, UHQ and HQNO in *B. pseudomallei* ([@b46]). Although a mutant unable to generate AQs in *B. pseudomallei* presented altered colony morphology and increased elastase production, the actual role of these molecules in the biology of this organism remains to be unravelled. AQs have also been identified in a number of species of *Burkholderia* such as *Burkholderia ambifaria*, *B. thailandensis*, *B. pseudomallei* and *Pseudomonas cepacia* (probably an unclassified *Burkholderia*). The main AQs produced by these organisms are 3-methyl derivatives of PHQ, HHQ and NHQ termed 4-hydroxy-3-methyl-2-alkylquinolines ([@b131]; [@b192];). Consequently, the operon responsible for their synthesis has been renamed *hmqABCDEFG* (formerly *hhqABCDE*), with the predicted methyltransferase *hmqG* being involved in the biosynthesis of these AQs. None of the above bacteria has *pqsH* orthologues nor produces PQS, and previous efforts to detect PQS in other pseudomonads such as *P. fluorescens*, *Pseudomonas syringae* and *Pseudomonas fragi* have been unsuccessful ([@b102]). Furthermore, in *B. thailandensis*, *B. ambifaria* and *B. pseudomallei*, the −3′ position is largely methylated ([@b192]), which would presumably preclude any *pqsH* analogue hydroxylating in these molecules and hence the production of PQS. These findings suggest that these organisms may lack the complexity of PQS signalling found in *P. aeruginosa*. Concluding remarks ================== The discovery of quinine and related natural antiplasmodial alkaloids, combined with the advances of synthetic chemistry, spurred significant research in the field and resulted in the development and evaluation of thousands of novel synthetic compounds. Among these were nalidixic acid and the extensive family of synthetic quinolone antibiotics. In parallel, the quest for antimicrobials of natural origin lead to the discovery of the AQNOs and of the extensive family of AQ compounds mostly produced by *P. aeruginosa* and related bacteria. Synthetic and natural quinolone antimicrobials, however, appear to share little in common with respect to their mode of action. The biological roles of the natural quinolones of bacterial origin are diverse and include intercellular signalling. The discovery of a non-AHL-based QS system in *P. aeruginosa* mediated via AQs and linked to the *las* and *rhl* QS systems provides a major insight into a complex regulatory network that plays key roles in infection via the regulation of virulence and biofilm maturation. Some AQs, such as PQS, are also able to sequester iron and have multiple functionalities. AQ biosynthesis requires several proteins and occurs via a condensation reaction between anthranilate and β-keto fatty acids. Their production is upregulated by both the *las* QS system and by AQs themselves and downregulated by the *rhl* QS system. AQs are present in bacteria other than *P. aeruginosa*, mainly other pseudomonads and *Burkholderiaceae*, although the role in these organisms is not, at present, very well defined. It is possible that AQs could be produced by many other bacterial genera, as the number of studies where these compounds have been specifically screened has been small and yet several AQs-producing species have been discovered ([@b102]; [@b47]; [@b192];). In most of the species that produce AQs, the role of these compounds or their biosynthesis remains unclear, although for many where genomic sequences are available, orthologues of the *pqsABCDE* operon extensively studied in *P. aeruginosa* can often be identified. Quorum quenching is the process by which the signalling mediating QS is interfered with, leading to the disruption of the normal means by which bacteria coordinate their behaviour according to their population density and preventing colonization ([@b153]). Quorum quenching can be exerted naturally by microorganisms to prevent the establishment of competing species and offers a strategy for the development of novel antimicrobial drugs ([@b155]). Hence, quinolone quenching offers the possibility to interfere with AQ signalling in pathogens such as *P. aeruginosa*, which rely on it to control virulence. As a primary target to interfere with AQ-mediated signalling, binding of PQS and HHQ to PqsR could be blocked. This would not only interfere with the production of AQs and their associated properties beneficial for the bacterial cell by preventing the expression of the *pqsABCDE* operon, but would also downregulate all the other AQ-regulated genes, including those essential for virulence. Compounds such as farnesol inhibit the induction capacity of PqsR on the *pqsA* promoter ([@b35]). However, as these compounds were found to act only at relatively high concentrations (in the mM range), other, more potent inhibitors are needed to stimulate clinical interest. Another possibility would be to inhibit the action of PqsE, whose function is still unclear, but that is required for the production of several virulence factors. Analogues of AQ precursors such as methyl-anthranilate or halogenated derivatives of anthranilate have been found to inhibit AQ synthesis, thus interfering with the signalling system probably by acting as competitive inhibitors of PqsA ([@b18]; [@b105]; [@b30];). This approach has recently shown promising results in limiting the systemic proliferation of *P. aeruginosa* infection in mice ([@b105]). To complete our understanding of the AQ signalling system in *P. aeruginosa*, the roles of some components still remain to be fully unravelled. PqsB, PqsC and PqsL are likely to be involved in the biosynthesis of AQs as revealed by the structural domains they share with other known enzymes and by mutational analysis ([@b64]; [@b46]; [@b13]; [@b59];). However, the exact role of PqsB and PqsC in the biosynthesis of HHQ and PQS still remains unclear. Similarly, little is known about the interaction of PqsL with precursor and other biosynthetic proteins to generate the *N*-oxide AQNOs, because the otherwise common precursor molecule HHQ does not appear to be required for their production ([@b43]). Even more critically, the role played by PqsE, which appears to be an enzyme of the metallo-β-hydrolase superfamily mediating the signal transduction that upregulates swarming motility, the production of pyocyanin, lectin, HCN, the transcription of many genes and ultimately virulence ([@b154]) remain to be elucidated. The biology of AQ biosynthesis, its regulation and the signalling functions of the AQs control the behaviour and virulence of *P. aeruginosa*. AQ signalling is proving to be complex, leading to many open questions that still remain unanswered. For example, the biological roles of AQs such as DHQ or the *N*-oxide derivatives in *P. aeruginosa* remain to be elucidated, as are the functions of AQs in other pathogenic and beneficial microorganisms producing them. From their inconspicuous discovery as potentially useful compounds having weak antimicrobial properties, quinolones in general and AQs in particular are highly versatile molecules that play central roles in the biology of the producer organisms. We are very grateful to the Biological and Biotechnological Sciences Research Council UK (BBF0143921) and the European Union (FP6 Marie Curie EST ANTIBIOTARGET MST-CT-2005-020278 and FP7 NABATIVI HEALTH-F3-2009-2009-223670) for kindly supporting the research performed by the authors in the topic of this review. Box 1, [Figs 2](#fig02){ref-type="fig"} and [4](#fig04){ref-type="fig"} have been adapted from [@b61], with special permission from Springer Science+Business Media, Dordrecht, the Netherlands, to whom we are also grateful. Statement ========= Re-use of this article is permitted in accordance with the Terms and Conditions set out at <http://wileyonlinelibrary.com/onlineopen#OnlineOpen_Terms>

PubMed Central

2024-06-05T04:04:19.913907

2010-8-25

PMC3053477

INTRODUCTION ============ A diagnosis of cancer and its subsequent treatment may result in physical, emotional, psychological and spiritual distress that negatively impacts quality of life ([@b41]). This has been observed across cancer sites ([@b31]; [@b41]), patient characteristics ([@b16]; [@b41]) and countries ([@b41]; [@b35]). Psychosocial morbidity can occur at any stage of the disease trajectory, even when the cancer has gone into remission ([@b31]; [@b41]; [@b35]). Suggested steps for improving psychosocial outcomes for cancer patients and survivors include identifying the prevalence of psychosocial morbidity, understanding causes and predictors and intervening appropriately ([@b5]). Once prevalence has been established, the second step is to understand possible causes and predictors of psychosocial morbidity. There are several categories of predictors that could play a role in determining psychosocial outcomes: (1) individual variables such as demographic characteristics, traits ([@b1]; [@b16]), disease and treatment characteristics ([@b15]; [@b10]); (2) social variables such as social support or network of the person with cancer ([@b40]); and finally (3) treatment centre variables related to the organisation and delivery of care within a setting. Individual characteristics of patients and providers ---------------------------------------------------- Some individuals may be predisposed to experience greater psychosocial morbidity due to their psychological make-up ([@b1]), demographic characteristics such as socio-economic status ([@b36]) and personality traits including coping styles ([@b72]). Combinations of demography and psychology may also predict psychosocial morbidity. For example, women with breast cancer who are younger and have a pre-existing history of depression have reported greater psychosocial distress than women who do not share these characteristics ([@b56]). As well as individual characteristics of the patient, individual characteristics of providers may also influence outcomes for the patients they treat. Patients of providers who have undergone communication skills training may experience less distress ([@b28]; [@b53]) and better coping behaviours ([@b28]) than those of providers who have not undergone this type of training. Disease and treatment characteristics ------------------------------------- The patient\'s cancer type, stage of disease and treatment (surgery, radiotherapy, chemotherapy and medication), may also predict the likelihood of experiencing poor psychosocial outcomes ([@b10]). For example lung cancer patients have reported worse global quality of life scores than patients with other types of cancer ([@b67]). Similarly, within the same cancer diagnosis, different treatment regimes such as radiotherapy and chemotherapy versus either alone can lead to different psychosocial outcomes ([@b3]; [@b2]). Social support -------------- Social support can be defined as the provision of non-professional practical assistance, information, emotional empathy and comfort ([@b43]; [@b18]) to patients by others within their social network. Research suggests that low levels of support may be linked to high levels of psychosocial morbidity ([@b17]; [@b47]; [@b49]). In a study by Parker *et al.*, cancer patients with poor social networks had worse mental functioning, higher levels of distress and lower overall quality of life, than patients with good social networks ([@b62]). Treatment centre characteristics -------------------------------- Features of the care environment such as procedure volume ([@b6]) and access to specialist care ([@b30]) have been associated with morbidity and mortality outcomes for patients. The association between volume and patient survival may reflect that specialists who treat many similar patients may have greater clinical skills and may be more up-to-date with best evidence ([@b14]). Further, because they see many similar patients they may be better systems in place to support the delivery of best evidence care for that patient group ([@b14]). Similarly, receipt of treatment at an institution that provides access to clinical trials has been associated with better outcomes ([@b58]). This has been attributed to the rigorous follow-up and care protocols that are applied to clinical trials patients ([@b23]). Given these findings, it is possible that characteristics of the treatment environment related to staff numbers, staff training, care protocol and systems may also be linked to psychosocial outcomes. Research on the role of systems may assist in developing a more sophisticated understanding of which structures and processes of care may influence psychosocial outcomes, leading to research to understand how this occurs and how such factors can be modified for the benefit of the patients served by the system. Once prevalence and predictors have been identified, the next step is to develop and test interventions designed to ameliorate the conditions that contribute to psychosocial morbidity. While disease predictors and many demographic predictors are not modifiable, they may enable attention to be directed to those patients who are most likely to be at risk of poor psychosocial outcomes. Some individual variables such as behaviours, coping strategies and cognitions may be modifiable. Psychological therapies aimed at modifying individual-level variables such as a patient\'s behaviours or cognitions, or social support have been developed ([@b5]; [@b42]). However, effect sizes are often modest ([@b59]; [@b50]), and individual-focused interventions are often too resource intensive for routine implementation ([@b50]). There are other aspects of the cancer patient\'s experience that can be modified. Treatment centre variables related to the environment where care is provided are potentially modifiable ([@b26]). Through policy and practice change, characteristics of the treatment centre have potential to be modified to achieve systematic benefits for all patients who receive care within a particular setting. This suggests that the relationship between treatment centre variables and psychosocial outcomes for cancer patients is an important area of investigation. The aim of this paper is to investigate the proportion of the published psychosocial literature over the last 10 years that has investigated the role of treatment centre variables as potential predictors of psychosocial morbidity in cancer patients. It is hypothesised that the most frequently identified predictors of psychosocial morbidity will be individual variables (individual traits, disease and treatment characteristics, etc.) followed by social support variables. METHODS ======= Data sources and extraction --------------------------- The Ovid search engine was used to search Medline for literature published between 1999 and 19 November 2009. The search string used was as follows: (cancer OR neoplasms) AND (psycho\$.mp or Anxiety or Depression or Quality of Life) AND (randomised controlled trial or Intervention Studies or Cross Sectional Studies or Longitudinal Studies). Inclusion and exclusion criteria -------------------------------- Papers that reported quantitative primary data (either descriptive or intervention studies) were published in English and that were relevant to psychosocial outcomes for cancer patients were retained. Dissertations, books, case studies, qualitative studies, commentaries/letters and review papers were excluded. Due to the large volume of publications identified, a 20% random sample was selected for classification using a random number generation. One coder categorised predictor variables for the entire sample of publications. The first author independently recoded 20% of the sample. Any disagreements were resolved by mutual discussion. Kappa statistics were calculated to determine inter-rater agreement. Coding of study design ---------------------- Studies were initially classified as either intervention or descriptive studies. Studies were classified as descriptive if they used a cross-sectional or longitudinal design. Studies were classified as intervention if they reported on the evaluation of an intervention designed to improve psychosocial outcomes in cancer patients. Classification of predictor variables in descriptive studies ------------------------------------------------------------ Publications reporting descriptive studies were examined to determine which variables were considered as potential predictors of psychosocial morbidity. Variables were grouped into four broad categories. ### Individual predictors (patients) This category included papers where variation in psychosocial outcomes according to individual patient characteristics was explored. Individual characteristics included the following demographic variables, traits, behaviours or experiences of the individual, disease characteristics and characteristics of the treatment received by the individual, as well as co-morbid conditions and cancer side effects. ### Individual predictors (providers) This category included papers where provider variables were linked to psychosocial outcomes. These could include provider demographic characteristics, attitudes, knowledge, skills or behaviours. ### Social support predictors This category included papers where characteristics of relationships or social networks were used to predict psychosocial outcomes, such as network size or quality of social support. ### Treatment centre predictors This category included papers where characteristics of the treatment centre were used to predict psychosocial outcomes. These could include both structures of care in the treatment centre where care is provided (e.g. patient volume, staff to patient ratios, equipment, services available, etc.) and processes of care used within the treatment centre (quality monitoring procedures, screening for distress, etc.). A description of variables in each of these categories is presented in [Table 1](#tbl1){ref-type="table"}. ::: {#tbl1 .table-wrap} Table 1 ::: {.caption} ###### Description of predictor variable categories ::: Predictor category Description ------------------------------------------------------------------------------------------- ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Individual predictors: characteristics of the patient Patients Demographics Age, gender, education, marital status, socio-economic status, other demographic variables Traits/behaviours Coping style, self-esteem, self-efficacy, outlook, locus of control and other individual characteristics Treatment characteristics Treatment regimes (surgery, chemotherapy, radiation therapy, courses of specific drugs, hormone therapy); and any other variables related to the medical treatment of cancer. Including complementary therapies and psychological therapies Disease characteristics Cancer type, stage of cancer Cancer side effects & co-morbid conditions Physical and psychological consequences of the diagnosis, disease and treatment of cancer, or the presence of unrelated physical or mental health conditions, cancer-related fatigue pain, hair loss etc. Individual-level predictors: characteristics of providers Providers Demographics, attitudes, skills, knowledge of providers Social support: characteristics of social support provided to patient Social support structure Living arrangements, network size Social support quality Dynamics within family units, peer relationships, social networks, support from co-workers and managers, and any other variables related to social support Treatment centre predictors: characteristics of the environment of where care is provided Structure of care Volume (e.g. number of cancer patients); setting (e.g. cancer care centre); presence of a cancer training programme; presence of specific types of equipment (e.g. radiation machines); presence of and composition of a multidisciplinary team; staff to patient ratios; teaching status; and any other variables related to the structure of cancer treatment units Process of care Delivery of treatment; case management and decision-making; diagnosis and staging; initial clinical management; patient involvement in decision-making; referrals and coordination of care; management of treatment toxicity; use of guidelines and monitoring of best practice; surveillance after initial therapy, and any other variables related to the manner in which care is provided ::: ### Scoring of descriptive papers Each paper received a score of one for each category of variable used to predict psychosocial outcomes identified within the paper. Scores for each category of predictor were summed across papers to determine the frequency of each predictor category. Classification of interventions ------------------------------- Intervention studies were coded according to whether they sought to modify characteristics of the individual patient or provider, characteristics of the patient\'s relationships or social support, or characteristics of the treatment centre. ### Individual patient-focused Interventions aimed at changing the knowledge, attitudes, traits, cognitions, behaviours or treatment of the cancer patient. In these studies, the unit of intervention and analysis was the patient. Studies that examined medical treatments that included psychosocial outcomes as either a primary or secondary outcome were included. These studies are denoted as individual patient-focussed (medical). ### Individual provider-focused Interventions aimed at changing provider knowledge, attitudes or behaviour including communication skills without changing any other aspect of the care environment. In these studies the unit of intervention was the provider, however, outcomes for individual patients were measured. Interventions aimed at assisting the provider with assessment of needs were included here. ### Social support-focused Intervention aimed at changing relationships or support structure within a family or other small social network. Interventions aimed at improving doctor--patient communication were included in this category if the intervention targeted the doctor\'s and the patient\'s behaviour, knowledge or skills. In these studies the unit of intervention and analysis was the social unit (e.g. family, dyad, etc.). ### Treatment centre-focused Interventions aimed at changing the structure, organisation or delivery mechanism of care. In these studies the unit of intervention and analysis was the system of care (e.g. hospital, clinic or ward). ### Scoring of intervention papers A score of 1 was assigned for each foci of the intervention (individual patient, individual provider, social environment, treatment centre). Scores in each category of intervention were summed across papers to give the frequency of each predictor category. Statistical analysis -------------------- PASW Statistics 18.0 was used to test the distribution of predictor variables across both the intervention and descriptive papers. *χ*^2^-tests were used to test the hypothesis that patient-related characteristics would be the most frequently reported predictors of psychosocial outcomes. RESULTS ======= A total of 4453 publications were identified in the literature search. From these, a 20% sample of publications was randomly selected. Of these 891 papers, 479 (53.8%) papers did not meet the inclusion criteria. Of the excluded papers, 214 (44.7%) were irrelevant to psychosocial outcomes, 134 (28.0%) were not focused on cancer patients, 71 (14.8%) did not report quantitative primary data, 43 (9.0%) were not published in English, and 17 (3.5%) were duplicates. The remaining 412 publications were coded for both study type and predictor type. For classification of included papers, inter-rater agreement as determined by kappa statistic was 0.82. Of the 412 papers that were included for further coding, 169 (41.0%) were classified as descriptive and 243 (59.0%) as interventions. The descriptive studies were further coded to indicate the type of predictor variables used to predict psychosocial outcomes. The number of descriptive papers reporting each type of predictor is shown in [Table 2](#tbl2){ref-type="table"}. A significantly greater number of papers examined individual predictors compared with those which examined social or treatment centre predictors (*χ*^2^= 212.0, d.f. = 3, *P*= 0.005). ::: {#tbl2 .table-wrap} Table 2 ::: {.caption} ###### Number of descriptive studies reporting each descriptor category ::: Predictor type Number (%) ----------------------------------------------- ------------ Individual  Demographics 56 (16.7)  Traits/behaviours/experiences 56 (16.7)  Disease 48 (14.3)  Treatment 77 (22.9)  Co-morbid conditions and cancer side effects 82 (24.4)  Individual level (provider) 0 (0.0) Social support  Support structure 1 (0.3)  Support quality 16 (4.7) Treatment centre characteristics  Structure of care 0 (0.0)  Process of care 0 (0.0) ::: The distribution of intervention types can be found in [Table 3](#tbl3){ref-type="table"}. As with the descriptive studies, the number of studies evaluating interventions to modify individual characteristics was significantly greater than the number evaluating interventions to change social or treatment centre characteristics (*χ*^2^= 145.67, d.f. = 4, *P*= 0.005). Of the 241 interventions concentrating on individual-level variables, 97 (40.2%) used psychosocial strategies, 140 (58.1%) used medical techniques, and 4 targeted the provider (1.7%). ::: {#tbl3 .table-wrap} Table 3 ::: {.caption} ###### Number of interventions by primary focus of intervention ::: Intervention type Number (%) ----------------------------------------- ------------ Individual-focused (patient) (provider) 241 (99%) Social support-focused 2 (0.8) Treatment centre-focused 0 (0.0) ::: DISCUSSION ========== This review sought to identify the proportion of psychosocial literature that has examined the potential role of treatment centre variables as predictors of psychosocial morbidity in cancer patients. No descriptive or intervention studies relevant to treatment centre predictors of psychosocial well-being were identified among the studies reviewed. This is surprising, given that strong attention has been directed towards the importance of health services as a predictor of other health outcomes including survival ([@b30]; [@b6]; [@b23]). What accounts for the predominant focus on individual-level predictors? ----------------------------------------------------------------------- Given that the majority of research focuses on individual predictors of psychosocial well-being, it is important to consider what factors might underlie this. These may include: (1) the theoretical orientation and training of the psychosocial researcher; (2) the greater availability of measures of individual variables compared with system variables; (3) the attention to individual-focused interventions in psychology practice; and (4) the practical advantages of doing individual-focused research compared with system-focused research. ### Theoretical orientation of psychosocial researchers leads to a focus on individual and social support predictors rather than a treatment centre focus Given that many psychosocial researchers have a psychology background ([@b57]), it is plausible that the individual focus in psychological theories and training has contributed to the preponderance of psycho-oncology research focused on individual predictors of distress. While psychological theories acknowledge the role of a range of factors in the aetiology of psychological distress, the emphasis is on the role of factors such as cognitions, attitudes, behaviours ([@b7]), interpersonal relationships ([@b63]) and social factors ([@b61]). This orientation toward individual and, to a lesser extent, social predictors of human behaviour is reflected in the training of psychology professionals ([@b25]). ### Measures of individual characteristics are more readily available than measures of treatment centre characteristics The interest in individual predictors has lead to the development of a number of standardised measures that can be used to measure individual variables ([@b70]; [@b27]; [@b24]; [@b69]; [@b74]). Such tools are widely used and accepted by researchers. In contrast there is a dearth of established methods and measures to assess treatment centre characteristics ([@b55]). Further, there is likely to be much less agreement about what types of treatment centre characteristics may be important to assess. Where measures do exist, for example, ward climate scales ([@b55]) these are likely to be less well known and accepted than measures of individual variables. Therefore, the greater availability of well established and validated tools to assess individual predictors in comparison with treatment centre predictors may perpetuate the focus on individual factors. ### Individual measures lead to individual-focused rather treatment centre-focused interventions If psychosocial research focuses only on the measurement of individual predictors, then these are the only explanatory factors available for the researcher to interpret his or her findings. This will lead naturally to the development of interventions aimed at modifying individual characteristics such as cognitions and behaviours. Congruence between theory and intervention is widely advocated ([@b11]; [@b54]), therefore factors that fall outside the theoretical orientation and expertise of those who do the research, such as systems of care, are not likely to be considered as avenues for intervention. This leads to evaluation of individual-focused interventions, thereby contributing to the focus on the individual rather than the system of care. ### Individual-focused research is easier to do than treatment centre-focused research As cancer is a common disease \[[@b4]; [@b75]\], there are likely to be few problems associated with accessing an appropriate sample of patients for research. As discussed, the assessment of individual variables in such research is commonly accepted and widely practiced. In contrast to this, a focus on system factors would lead to considerable challenges for researchers. These may relate to logistical and cost considerations related to obtaining a large enough sample of treatment centres with which to assess the role of system factors ([@b52]; [@b66]). There may also be significant political sensitivities associated with the collection of data related to treatment centre characteristics, especially where there may be implications for professional and institutional reputations. These factors suggest that it is easier to do individual-focused research than research focused on the system of care. This may contribute to the continued research focus on individual predictors. Why should characteristics of the treatment centre be examined as possible predictors of psychosocial outcomes? --------------------------------------------------------------------------------------------------------------- Treatment centre factors relate to the characteristics of the organisation where care is provided ([@b21]) and may also relate to the broader health care context. Treatment centre characteristics may create an environment that influences the practice of providers ([@b44]). This influence may be exerted through policies, procedures and performance monitoring ([@b14]). In this way the system defines what is expected from providers ([@b19]). As such treatment centre factors can help improve patient outcomes by supporting practices and processes of care that are linked to better outcomes ([@b19]). If we accept that these principles operate to influence outcomes in psychosocial care, then this suggests a need to examine the role of structures and processes of care as predictors of psychosocial outcomes. Treatment centre variables have the potential to provide fruitful avenues for investigation and intervention in improving a range of outcomes ([@b71]), including psychosocial morbidity. There are several advantages to exploring the role of treatment centre variables in psychosocial outcomes. First, many individual demographic or disease variables such as age, gender, type of cancer and stage are non-modifiable. Personality traits such as extraversion or optimism are also difficult to modify. Individual psychological characteristics such as coping strategies and behaviours may be modified; however, uptake of interventions may be low and effects modest ([@b50]). To optimise uptake and effectiveness of individual-focused psychological interventions, implementation will need to be supported by systems within the treatment centre (e.g. training, policies, procedures, coordination of care). Similarly, intervention aimed at changing the practice of individual providers needs to be coupled with strategies at the organisational or treatment centre level ([@b33]). This is because organisational factors may support or hinder systematic implementation of best practice ([@b33]). Systems to monitor patient outcomes or relevant processes of care may be costly to develop and maintain ([@b19]). Hence an approach that enables a range of potential problems to be assessed may be more efficient that multiple individual-focused interventions or services developed and implemented in isolation. Systematic approaches for assessing anxiety, depression as well as a range of other concerns related to information, physical, spiritual and emotional well-being have been trialled previously ([@b51]). A focus on individual predictors suggests that the burden of change to improve psychosocial outcomes rests with the person with cancer. This is at odds with the paradigm employed in other areas of medicine; whereby variation in outcomes are seen as resulting not only from variation in clinical variables, but also from the quality of care received and the treatment centre characteristics that support delivery of quality care ([@b34]). The latter approach places the onus on the treatment delivery setting rather than the individual patient to ensure that the best possible outcomes are achieved for the individual. Third, an approach that takes into account the role of treatment centre characteristics is likely to support equitable care delivery. Adoption of the approach increases the likelihood that all patients, regardless of which provider they see and how well they can communicate their needs, will have access to the best practice psychosocial care. How can characteristics of the treatment centre be measured? ------------------------------------------------------------ There are a number of existing approaches available for considering the effect of treatment centres. Donabedian\'s model involves the considering the role of 'structures' and 'processes of care' on patient outcomes ([@b21]). Structural variables refer to characteristics of the organisation that facilitate delivery of high quality care. These may include size of the organisation, equipment or number of staff ([@b13]). The process domain covers variables related to delivery of care including the presence of policies, procedures and cues to support implementation of best practice care ([@b13]). Other approaches have emphasised the importance of factors such as organisational culture and climate ([@b55]; [@b12]), team functioning ([@b60]) and leadership ([@b64]). Culture and climate relate to a team\'s shared values and beliefs about an organisation\'s policies and practices ([@b37]). These factors are thought to create conditions conducive to the adoption of best practice care ([@b32]). The National Health Service in the UK assesses consumer perceptions of the following domains of care: (1) responsiveness to consumer needs, values and preferences; (2) integration and coordination; (3) physical comfort; (4) emotional support; (5) involvement of family and friends; and (6) information, communication and education ([@b46]). These criteria were developed from consumer views about what is important to quality of care ([@b29]). Notably this focuses predominantly on process of care rather than on structures or team functioning aspects. This perhaps reflects that the latter factors are less likely to be observable to consumers. Each of the approaches to examining treatment centre characteristics leads to different types of data collection strategies. Donabedian\'s model focuses on variables that can be assessed from administrative data. Some process of care data may be examined via medical records audit or through patient or provider report ([@b21]). Assessment of organisational cultural may be done by key informant interviews ([@b20]; [@b21]), or surveys of staff within the institution ([@b12]). In contrast, the NHS approach puts an emphasis on the views of consumers ([@b46]; [@b39]). While each of these theoretical approaches discussed have limitations, each represents a starting point for beginning to examine the role of system characteristics in psychosocial outcomes. How can variation in outcomes between treatment centres be assessed? -------------------------------------------------------------------- The first step in assessing whether variation in outcomes is related to treatment centre characteristics is to determine the level of variation in psychosocial outcomes between organisations. This requires the use of a reliable and valid measure of psychosocial well-being ([@b48]). The measure needs to be administered to a randomly selected number of patients in each treatment centre. The number of patients should be sufficient to represent the performance of the treatment centre. If variation exists, then the relative contribution of patient and treatment centre variables can be examined. How can interventions to modify treatment centre characteristics occur? ----------------------------------------------------------------------- If system factors are shown to influence psychosocial outcomes, then a strategy for intervening is needed. Interventions for changing systems of care include the use of local opinion leaders to influence the culture and practices of others within the organisation ([@b22]); the use of audit and feedback ([@b45]); and implementation of policies and procedures ([@b65]). A commonly used approach is the collaborative method ([@b73]). This involves collection of outcome or process of care data and regular provision of feedback to the clinical team. The team is responsible for setting performance improvement goals and identifying where care can be improved ([@b73]). One of the appealing characteristics of this intervention strategy is that it allows some flexibility for intervention strategies to be tailored to the needs of each participating organisation. More research is needed, however, to develop evidence of effectiveness ([@b68]). How can interventions to modify treatment centre characteristics be evaluated? ------------------------------------------------------------------------------ Where the unit of intervention is the system not the individual patient, randomised controlled trials in which individual patients are allocated to the intervention or control group may be unsuitable ([@b52]). Cluster randomised controlled trials where the unit of allocation is the organisation may be used as an alternative ([@b52]; [@b66]). Data are collected from a sufficient number of patients within each organisation to enable the performance of the organisation to be represented. An adequate sample of both organisations and patients therefore needs to be recruited ([@b66]). Often, however, it is not feasible to recruit the number of organisations needed for a cluster randomised trial ([@b52]; [@b66]). The Cochrane Effective Practice and Organization of Care Group recommends in addition to randomised trial, controlled before and after studies and interrupted time series designs are appropriate for this type of evaluation ([@b9]). Controlled before and after studies involve a control group and intervention group that are not randomly assigned ([@b8]). Baseline and post-test data collection must be collected at the same time in both groups. Interrupted time series studies involve collection of repeated measures data in one site. At least three data points must be collected both before and after the intervention to establish whether there is any change of trend in the data due to the intervention ([@b8]). A variation to this is the multiple baseline design ([@b38]). This involves collection of repeated measures data in several sites. The timing of the intervention is staggered between sites to allow greater control for the effect of external variables on any changes in trend observed ([@b38]). Understanding the prevalence and predictors of psychosocial morbidity, and intervening appropriately, are critical requirements to improve cancer outcomes. This study examined a 20% random sample of psychosocial research literature published over the last 10 years to determine the extent to which research has examined individual-level, social and treatment system predictors. The majority of both descriptive and intervention studies focused on individual-level variables; only 5.0% of descriptive and 0.8% of intervention studies addressed social support variables, and none examined treatment centre predictors. Possible reasons for this discrepancy and suggestions for future research were proposed. Study limitations ----------------- The literature search was conducted using one electronic database and only peer-reviewed papers were included. The search terms used in the review are commonly employed in the psycho-oncology literature; however, it is possible that some relevant articles were missed by these terms. While it is possible that a broader search using additional databases and inclusion of grey literature would have identified additional articles relevant to this review, it is unlikely that this would have substantially changed the proportion of papers examining each type of predictor. Conclusions ----------- Few studies have examined the role of treatment centre characteristics in psychosocial outcomes for cancer patients. There is a need to rigorously assess what types of treatment centre characteristics may influence outcomes and to what degree. This creates a potential avenue for developing interventions aimed improving patient outcomes although the implementation of cohesive and systematic processes and structures to support best practice psychosocial care. [^1]: Funding: This work was supported by the National Health and Medical Research Council (grant number 300749). There are no financial disclosures from any authors. [^2]: Re-use of this article is permitted in accordance with the Terms and Conditions set out at <http://wileyonlinelibrary.com/onlineopen#OnlineOpen_Terms>

PubMed Central

2024-06-05T04:04:19.922103

2011-3-01

PMC3053505

Design and development of drug compounds and new pharmaceutical formulations require full characterization of the chemical and physicochemical events occurring at the level of the single active ingredients or excipients, as well as following their reciprocal interaction. Thermal analysis techniques are among the most widely used methods to this aim; among them, by using the Differential Scanning Calorimetry (DSC) technique, the thermotropic behavior of a single substance or mixtures is analyzed as a function of a controlled temperature program. Differential Scanning Calorimetry is an accurate and rapid thermoanalytical technique, widely used by the pharmaceutical industry and in drug research, to investigate several physicochemical phenomena, such as polymorphism, melting and crystallization, purity, drugexcipient interaction, polymer properties, effects of drying and lyophilization, as well as to characterize biomolecules like peptides, proteins, and genetic material. A recent interrogation of the Pubmed database (December 2010) on publications containing the words 'Differential Scanning Calorimetry' (DSC) gave a total number of about 12,900 items. Out of them, a cross analysis between DSC and the terms 'drug\*' or 'pharm\*' reduced the number to about 4,250 and 2,800 items, respectively. A search for 'DSC' and 'membrane' gave about 2,120 publications, 130 of which were also related to 'biomembranes'. These figures very well corroborate the value and significance of such an analytical technique in the development and optimization of new drug compounds and their final formulations. However, peculiar applications of DSC in biomedical research are also possible, and according to us, not all of them have been explored yet. For instance, DSC can be a very powerful tool and the source of a large amount of information, to study the interactions between drugs and cell membranes. This can be due to both the qualitative and quantitative levels, and using either eukaryotic or bacterial cell membranes, as well as different biomembrane models. In fact, the 3-D nature of these systems allows to put into evidence the different possible mechanisms and degrees of interactions between a biologically active molecule and biomembranes, much better than 2-D (i.e., solvent-solvent) experimental approaches. The database returned only 21 hits when the terms 'DSC' and '(bio)membrane model\*' were put in, and only 14 when the terms 'drug', 'membrane interaction,' and 'DSC' were cross-queried. This meant that a large amount of work could still be done, to make DSC a routine technique in this specific phase of drug design and development. In this special issue we have invited some scientists renowned for their work in the field of DSC applications to drug development and delivery, and especially to drugbiomembrane interaction studies, to contribute in clarifying some of the above-mentioned issues and to highlight areas where uncertainties remain. A fundamental challenge is to combine insights from biochemistry and physiology with those from structural biology and bio-thermodynamics, to obtain a complete depiction of cell membranes and their functions. The review of Prof. Raudino (Department of Chemical Sciences, University of Catania, Italy) brilliantly delineates the physical principles and the thermodynamic techniques staying at the basis of the biological membrane models. The great number of experimental data must be interpreted on the basis of approximate, but not over-simplified, models; Prof. Pignatello (Department of Drug Sciences, University of Catania, Italy) has critically analyzed the literature regarding the *in vitro* biomembrane models and their significance in the development of new drugs and medicines. Prof. Chiu and Prof. Prenner (Department of Biological Sciences, University of Calgary, Canada) have instead reviewed the applications of DSC in biochemical and pharmaceutical studies and highlighted the numerous shortcomings of the approaches used and the results reported in recent literature. An interesting and particular application of DSC for studying the release rate and kinetics of drugs from colloidal nanovectors to biomembrane models has been reviewed by Prof. Sarpietro and Prof. Castelli (Department of Drug Sciences, University of Catania, Italy), whereas Prof. Bastos (Department of Chemistry and Biochemistry, University of Porto, Portugal) has contributed with a research illustrating the application of DSC in characterizing the structure-activity profile of peptide antimicrobials derived from lactoferrin. Finally, Prof. Giatrellis (Department of Medical Biochemistry and Biophysics, Umeå University, Sweden) and Prof. Nounensis (Biomolecular Physics Laboratory, NCSR Demokritos, Greece) have made a critical analysis of the literature on DSC studies on nucleic acid-membrane systems, discussing the experimental data related to the thermodynamics and kinetics of DNA-lipid complexation, and especially to the lipid organization and phase transitions within the membrane model. We are grateful to all the co-authors for the production of the above-mentioned articles and reviews; we are also indebted to the scientists who have peer-reviewed the manuscripts and given valuable advice to the authors and ourselves, whereby all had to schedule their research around a tight time scale. We are also extremely grateful to the Editor-in-Chief, Professor R. K. Khar, Dr. M. Aqil, Editor of the Pharmaceutical Sciences section, Dr. H. Gupta, Head of the Management Board, and the OPUBS group, to have accepted the proposal of this special issue.

PubMed Central

2024-06-05T04:04:19.924984

2011-01-01

PMC3053506

Humans are constantly exposed to a variety of microorganisms, including the major groups, namely, bacteria, fungi (yeasts and molds), algae, protozoa, and viruses. Despite steadily improving public awareness and continued diagnostic and therapeutic advances, human race faces a continuous threat from microorganisms through infections. Many individuals develop a variety of infections but quickly overcome them. In most cases these microorganisms do not produce infection because the skin and mucous membrane surfaces provide effective barriers against invasion. However, some individuals are unfortunate as a few microorganisms can invade through the barrier or through lesions from surgery or trauma and develop chronic or persistent infections. Statistics show 2.5 deaths on an average per 1 million people in the USA as a result of bacterial infections.\[[@CIT1]\] Similarly bacterial infections prevail throughout the world increasing the mortality rate considerably. Antibiotics are used to supplement the body's natural defenses against a bacterial infection. They act either by killing bacteria or by stopping them from growing and multiplying. Antibiotics have historically played a major role in the diagnosis and treatment of infections. Antibiotics have dramatically changed the practice of medicine in this century. They have significantly reduced the occurrences of diseases, such as diphtheria, syphilis, and whooping cough.\[[@CIT2]\] A point to be noted here is that antibiotics are of high demand than any other drugs comparatively, which proves their significance in world economy as well as public health. Even though the topic is very wide through the history and can be hardly told in a nutshell, fewer literature available in this highly specified field forced an inspiration for this topic. History of Antibiotics {#sec1-1} ====================== At the beginning of the 20th century, a metabolic approach was applied to the formulation of new drugs. German bacteriologist Paul Ehrlich contended that since various cells of the body and of various microorganisms could be selectively stained by certain dyes, there must be specific active groups in cells of the human body and in microorganisms to which drugs of the dye type might attach. Such a drug would then act as a "magic bullet," attacking the target cells specifically and killing only the microorganisms, while leaving the human body unaffected.\[[@CIT3]\] Drug development made a great leap forward with the discovery of antibiotics. In 1928, the Scottish scientist Sir Alexander Fleming found a zone in a culture of bacteria that was caused by the invasion of a mold. Penicillin, the extract from that mold, was shown to cure bacterial infections. Meanwhile, in the 1930s sulfa drug became the miracle drug and was used to treat many life-threatening infections. But it tasted bad and was difficult to swallow. As a solution, a US company developed a palatable, raspberry flavored liquid product. However, they used diethylene glycol to solubilize the sulfa, which killed around 107 people, mostly children. This incident made the global countries, including the USA, the Pharma giant, to tighten their regulations. The golden age of antimicrobial therapy started in 1941 when the brilliant research of a group of investigators, led by Howard W. Florey and Ernst Chain, purified penicillin and produced quantities sufficient to permit clinical trials.\[[@CIT4]\] Subsequently, many other antibiotics have been developed. Antibiotics have almost entirely replaced sulfonamides in the treatment of bacterial infection. Generic Pharma Industries / Generic Medicines {#sec1-2} ============================================= The pharmaceutical industry develops and produces a variety of medicinal products that save the lives of millions of people from various diseases and permits many people suffering from illness to recover and to lead productive lives. There are different types of pharmaceutical companies, namely, mainline pharmaceutical companies, which are established firms that have many approved drugs already on the market. These companies often have a significant number of research and development (R&D) laboratories and manufacturing plants throughout the nation and around the world. In contrast, smaller pharmaceutical companies often do not have any approved drugs on the market. In addition to developing their own drugs, some of them may perform contract research for other pharmaceutical companies. Finally, generic pharmaceutical companies manufacture drugs that are no longer protected by patents. Because their products are all established drugs, they devote fewer resources to R&D and more to manufacturing. A generic medicine is a faithful copy of a mature drug---no longer under patent---marketed with the chemical name of the active ingredient. It is a pharmaceutical product intended to be bioequivalent with the innovator or the company which launched the new drug, manufactured without a license from the innovating company and marketed after expiry of a patent or other exclusivity rights. The development of generics markets stems from at least 2 major points. On the demand side, to face the rapid growth of health care expenditures, health care systems have been reformed recently. On the supply side, however, in this decade the slow input of innovative drugs and the expiry of patents on many important medical specialties have led to increasing price competition, favoring generics in many developed countries.\[[@CIT5]\] Generic Antibiotics in Cost Reduction {#sec1-3} ===================================== The generic industries play a major role in the cost reduction of pharmaceutical drugs in general and antibiotics in particular. Historically, the pharmaceutical industry capitalized on the discovery that many microbial secondary metabolites act as antibiotics. Even today, significant portion of the worldwide pharmaceuticals are dedicated to the production of antibiotics. Prescription drugs represent a significant component of increasing costs, with shares ranging from 4% in the USA to nearly 18% in France and Italy.\[[@CIT6]\] The industry data shows that the average brand name prescription price in 2008 was almost 4 times the average generic price (\$137.90 vs \$35.22).\[[@CIT7]\] The Food and Drug Administration (FDA) analysis of 1999--2004 data shows that generic competition is associated with lower drug prices: on average, the first generic competitor prices its product only slightly lower than the brand name manufacturer; the second generic manufacturer reduces the average generic price to nearly half the brand name price; prices continue to fall but more slowly as additional generic manufacturers market the product. For products with a large number of generics, the average generic price falls to 20% of the branded price and lower.\[[@CIT8][@CIT9]\] Innovation in the pharmaceutical industry, spurred in part by competitive market forces, continues to bring enormous benefits to the world. Literature from the USA has shown that brand name manufacturers do not compete on price once generic competitors become available.\[[@CIT10]\] Because generic drugs are typically far less expensive than their corresponding brand name versions, competition from generic drugs can deliver large savings to consumers. United States of America -- Golden Duck for Generic Antibiotics {#sec1-4} =============================================================== The USA market is considered to be one of the biggest markets in the health care sector due to its million dollar business potential and is the primary target for most of the generic pharmaceutical manufacturers. In USA, the pharmaceutical industries have achieved worldwide reputation through R&D on new drugs and spend a relatively high amount of its profits on R&D compared with other industries. Each year, pharmaceutical industry testing involves millions of compounds, yet in the long run, yields only fewer new fruitful medicines. It is critical to maintain appropriate incentives for the development of new drug products, because the necessary R&D is risky and costly. At the same time, expenditures on pharmaceutical products continue to grow and often outpace expenditures for other consumer products. The new drug development, as well as the generic drug availability is well balanced by the current regulations in the USA through implementing the Hatch--Waxman Act. In the USA, oral and parenteral antibiotics were found to possess equal shares of sales. The estimates of antibiotic availability from numerous countries throughout the world serve as an indicator of potential patterns of human antibiotics sales. The extraordinary therapeutic effects of antibiotics, the occurrence of resistance, and the considerable resources spent on antibiotics worldwide were the compelling reasons for concern about adequate and appropriate use of these powerful agents. Antibiotics often accounted for 15%--30% of drug expenditures, the largest share of expenditure among any therapeutic group of drugs. According to the US Centers for Disease Control and Prevention, 1.7 million people per year in the USA face hospital-acquired infections, leading to 5.8% deaths among them.\[[@CIT11]\] Recent figures from the USA indicate that the cost associated with the treatment of these infections in the USA is around \$6.7 billion.\[[@CIT12]\] Such kind of reports infer that hospital-acquired infections in the developed world cost more than \$32.5 billion, higher than the current global sales of antibiotics. Thus antibiotics having the potential impact on preventing mortality in the developed part of the world play a major impact in the global market.\[[@CIT12]\] The costs associated with these infections therefore indirectly provide an important measure of the failure of current antibiotics due to developing resistance, thus encouraging industry to search for newer antibiotics as well. Promotion of Generics: Hatch--Waxman Act {#sec1-5} ======================================== The FDA or USFDA is an agency of the US Department of Health and Human Services, responsible for protecting and promoting public health through the regulation and supervision of prescription and over-the-counter pharmaceutical drugs, vaccines and biopharmaceuticals, and many other commodities, which may influence public health. FDA plays an inevitable role in the approval of generic as well as new drugs, which are to be marketed in the USA. The Drug Price Competition and Patent Term Restoration Act of 1984, usually referred to as the Hatch--Waxman Act, was designed to promote generics in the USA while leaving intact a financial incentive for R&D. It allows generics to win FDA marketing approval by submitting bioequivalence studies. Approvals were generally provided with the following certifications: Paragraph I Certification: The generic applicant certifies that there are no patents listed in the orange book. "Orange book" being a publication of USFDA, lists the patents relating to drugs approved for marketing and sale in the USA, including patents that protect active ingredients.Paragraph II Certification: In case any listed patents have previously expired, the applicant may enter the marketplace immediately upon FDA approval.Paragraph III Certification: The applicant certifies that any listed patent has not yet expired but will expire on a particular date. The FDA may approve the Abbreviated New Drug Application (ANDA) and make it effective as of the patent expiration date.Paragraph IV Certification: The applicant for generic approval intends to market the drug prior to expiration of any patent(s) listed in the orange book; the applicant makes a certification that the patent(s) are not infringed or are invalid and FDA notifies the New Drug Application (NDA) holder and patent owner accordingly.\[[@CIT13]\] It also grants a period of additional marketing exclusivity to make up for the time a patented pipeline drug remains in development. This extension cannot exceed 5 years, and it is in addition to the 20 years exclusivity granted by the issuance of a patent. Another provision of the Hatch--Waxman Act is that it grants a 30-month stay to drug companies that file suits against generic manufacturers who challene their patents. Thus the act maintains a fair balance between the innovator of a new drug and the generic drug producers. History of Generic Antibiotic Regulations in USA {#sec1-6} ================================================ The history of antibiotic regulation clarifies the relationship between regulatory plan and the scientific/regulatory constraints and the marketing conditions in which they operate. Antibiotics and insulin-containing drugs were added to the regulatory scheme beginning with a series of steps in 1941. However, the procedures for establishing safety and efficacy applicable to other "new drug" antibiotics were subject to a far different regulatory scheme. In the case of antibiotics, the monographs were developed on the basis of the first product or the innovator product reviewed and approved in the antibiotic class. Thereafter, any forthcoming vendor merely needed to show that it was bioequivalent to the innovator product for which the monograph was developed and that it followed the specifications of the monograph. In this way, the innovator owned no rights even though the following vendors receive the approval based on his certificate. This seems to be a paradox.\[[@CIT14]\] Along with this, the antibiotic vendor should provide a sample of each batch of the antibiotics to the FDA for laboratory testing and certification. The batches were tested by the agency and if found to meet the standards, the Antibiotic Certificate was issued.\[[@CIT15]\] Also, the batches which were not tested but released prior to such certification were considered as misbranded. Quite unexpected by the agency, the testing of antibiotics became slower due to practical difficulties, such as personnel and facilities limitations. As a result, large quantities of antibiotic products were held in quarantine for many weeks just only for the clearance by the agencies. As an initiative to relax the regulations, in 1980, the FDA announced that testing of topical antibiotics would no longer be required. Finally, in 1982, the batch certification program for antibiotics was eliminated entirely but was considered and regulated as for any other drug to comply with the monograph.\[[@CIT16]\] In 1986, over-the-counter antibiotics that complied with the applicable monograph were excluded from the batch certification process. In contrast to the earlier times where only penicillin was the available antibiotic in the market, several hundreds of antibiotics started getting approval from the agency. As a result of the 1962 Amendments, the FDA required the submission for several antibiotics of scientific evidence of substantial well-controlled clinical studies, demonstrating the effectiveness of the product. Those products that failed to provide such evidence had their certifications overturned. In addition, the FDA cancelled approval of several antibiotics that did not have substantial scientific evidence. In 1985, antibiotic applications were classed as either "New Antibiotic Drug Applications" or "Abbreviated Antibiotic Drug Applications." In line with the new regulatory framework, applicants and manufacturers could now make certain changes in their products without requiring prior FDA approval. In 1997, antibiotics were started to be reviewed and approved on the same basis as any other pharmaceutical product. In 2000, the FDA recounted the term "antibiotic drug" as it refers not only to the active chemical substance, but to any derivative of the substance, such as a salt or an ester of the substance. Under the new scheme, similar to nonantibiotics, antibiotics could also henceforth contain drug substance information in the finished product application itself. Thus, antibiotics entered into the normal stream of pharmaceuticals as any other pharmaceutical categories. Global Scenario of Antibiotics {#sec1-7} ============================== The antibiotics market generated sales of US\$42 billion in 2009 globally, with 14 products recording sales of more than \$1 billion.\[[@CIT17][@CIT18]\] There were 7 blockbuster antibacterial drugs and 8 antivirals. The volume of antibiotic use was increasing in most European countries between 1997 and 2010. In European countries, penicillins were the most prescribed outpatient antibiotics and further enlarged their leading position between 1997 and 2003.\[[@CIT19]\] Likewise, the use of quinolones surged, while the use of another 2 major classes of antibiotics, tetracyclines and sulfonamides, stagnated or decreased in most European countries as newer antibiotics superseded them. More detailed data on the European antibiotic markets are not available due to the wide differences in consumption. Many reasons have been proposed to explain the large differences in the consumption of antibacterial agents among European countries, including the incidence of community-acquired infections, knowledge about antibiotics, and regulatory practices. Striking geographic variations were observed in the use of various antibiotic classes. For instance, the narrowspectrum penicillins and the first-generation cephalosporins were widely prescribed for the treatment of community-acquired infections in many Nordic countries, while they almost disappeared in most Southern European countries. In the latter, an increased use of the newer antibiotics, such as amoxicillin/clavulanic acid, macrolides, and quinolones was also observed. Japan is the second-largest pharmaceutical market in the world. It is well known that Japan is the third country next to the USA and the UK to become self-sufficient in penicillin manufacture as early as in 1948. Besides this, much effort was made in exploratory research on anti-infectives. Starting from colistin, several antibiotics, namely, kanamycin, bleomycin, piperacillin, norfloxacin, meropenem, and others, are from Japan. As of the 12 months to Q1 2009, the Japanese pharmaceutical market accounted for nearly 10% of the world market, and generated total sales of \$71.6 billion. The Japanese government ensures that all citizens and workers have health insurance. Compared with that in 2008, Japan tightened monitoring and inspection of imported antibiotics in 2009 with more stringent requirements for their testing. Regulatory Challenges {#sec1-8} ===================== Several regulatory challenges were involved in the development of a generic antibiotic. Regulatory challenges, such as bioequivalence, patent expiry, newer antibiotic, and the complexity involved in the regulated market, are explained in the following sections. [Figure 1](#F0001){ref-type="fig"} depicts the various tasks involved with the drug development. ::: {#F0001 .fig} Figure 1 ::: {.caption} ###### Antibiotics from generic industry to regulated market :::  ::: Bioequivalence {#sec2-1} -------------- In order to protect consumers, generic products must be demonstrated to be therapeutically equivalent to a previously approved product, typically an innovator product. Nowadays bioequivalence studies are essential part of a company's registration dossier. These bioequivalence studies measure the bioavailability of 2 formulations of the same active ingredient. The plasma concentration of a drug determines the number of drug molecules at a receptor and hence the therapeutic effect. The plasma concentration is governed by absorption, distribution, metabolism, and elimination. The pharmacokinetic parameters measured are area-under-the-concentration (AUC) time curve, the peak concentration (C~max~), and the time to peak concentration (t~max~). Statistically, geometric mean ratio of the test to the reference drug for AUC and C~max~ must fall within 90% confidence limits of 80 and 125.\[[@CIT20][@CIT21]\] Within these statistical limits, these particular parameters will be sufficient for bioequivalence. Substitution of generic drugs for brand name products is highly controversial, especially in antibiotics and often is met with suspicion by health care providers and patients. Bioequivalence issues present more concerns over generic drug substitution, such as consumer perception of risk, differences in product and packaging appearance, and differences in excipients. Also, several antibiotics, such as cephalosporins, have relatively narrow therapeutic window and would undergo stringent bioequivalence testing, because relatively modest changes in the concentration achieved in body fluids might well be associated with large changes in the frequency of therapeutic failure or significant toxicity. It arises several questions as whether is it possible to market various intravenous products with same concentration of active ingredient but with variable quality, even if it does not affect the clinical outcome, it does little to build trust in the generic industry.\[[@CIT22]\] Patent expiry {#sec2-2} ------------- As the earth moves closer to the concept of a global village by elimination of trade barriers, new challenges crop up. A case in point is the market for generic antibiotics. Hatch--Waxman established a regulatory framework that sought to balance incentives for continued innovation by brand name companies and opportunities for market entry by generic drugs. Before 1962, it was observed in the USA that out of the 150 off patent drugs in the market, there were no generic drugs. Many companies did not go in for manufacture of generic drugs because of the impractical and nonscientific manner in which the regulatory authorities viewed the approval process. The Hatch--Waxman Act addressed these issues and proposed many reforms. The underlying objective of the Act is that in the absence of generic drugs, it is difficult to check the profiteering motive of the patent owner of a drug, who may put patients to ransom. However, at the same time it also takes care of the interest of the patent owner and provides relief for undue lengthy process. It is now possible for many generic companies to qualify for the 180-day market exclusivity if several applications are filed on the same day. Under the Hatch--Waxman Act, the government has a system of patent term "restorations" under which monopoly of the original patentee can be extended for a maximum period of 5 years in addition to the initial patent term. In the European Union also there exists a system of supplementary protection.\[[@CIT23]\] Newer antibiotics and their hurdles {#sec2-3} ----------------------------------- Although the need for new antibiotics is increasing, a number of factors make these drugs less economically attractive than drugs that treat chronic diseases. Pharmaceutical companies appear to be less interested in developing anti-infective drugs.\[[@CIT24]\] Reason being antibiotics are typically taken for a week or 2, which makes them cost-effective for the health system, but less lucrative to drug companies than medicines for diseases, such as cancer or diabetes, which might be taken for months together.\[[@CIT25]\] In addition, to prevent the evolution of resistant strains of bacteria, physicians who treat infectious diseases try very hard to limit the overuse of newer antibiotics. These being the reason for the struggle of existence for the newer antibiotics, the discovery of newer antibiotics came to a slower pace due to loss of interest. Such a condition makes a patient suffer from antibiotic resistance and costly medication. Literature search shows that the FDA approvals of new antibiotics declined 56% during the past 20 years (1998-2002 vs 1983-1987).\[[@CIT26][@CIT27]\] Further, to prove the condition, only a few new antibiotics were in the pipeline or already approved by the FDA. [Table 1](#T0001){ref-type="table"}shows the new drug applications approved from 2005 to 2009. ::: {#T0001 .table-wrap} Table 1 ::: {.caption} ###### New drug applications (antibiotics) approved during the period 2005--2009 ::: Antibiotics new drug approvals during 2005--2009^\[[@CIT30]\]^ ---------------------------------------------------------------- ------------- ---------------------------------------------------------- ---------------------- -------------- -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- N022106 Doribax Doripenem Ortho McNeil Janssen Oct 12, 2007 Provides for the treatment of complicated intraabdominal and complicated urinary tract infections caused by susceptible isolates of the designated microorganisms N021821 Tygacil Tigecycline Wyeth Pharms Ltd Jun 15, 2005 Tygacil is indicated for the treatment of complicated skin and skin structure Infections and complicated intraabdominal infections N050786 Pylera Biskalcitrate; metronidazole; tetracycline hydrochloride Axcan candipharm Sep 28, 2006 Provides for the treatment of patients with *Helicobacter pylori* infection and duodenal ulcer disease (active or history of within the past 5 years) to eradicate *H. pylori* N050818 Tobradex St Tobramycin/dexamethasone Alcon Feb 13, 2009 Superficial bacterial ocular infection or a risk of bacterial ocular infection exists ::: Complexity of the market {#sec2-4} ------------------------ Even though it looks simpler, the markets for generic antibiotics are more complex due to several factors. Several factors that influence the generic antibiotic market growth are depicted in [Figure 2](#F0002){ref-type="fig"}. Long-term growth in the global antibiotics market would be affected by 2 major factors: antibiotic resistance and generic competition. ::: {#F0002 .fig} Figure 2 ::: {.caption} ###### Factors influencing the generic antibiotic market growth :::  ::: The evolution of antibacterial resistance in human pathogenic microorganisms is the result of the interaction between antibiotic exposure and the transmission of resistance within and between individuals. In the community, there is cumulative evidence that the antibiotic exposure of populations promotes acquired antimicrobial resistance in community pathogens, such as *Streptococcus pyogenes*, cutaneous staphylococci, and propionibacteria. In a hospital setting, an increased use of antibiotics is often associated with an increase in the frequency of antibiotic resistance. Furthermore, the relationship between antibiotic use and resistance is most evident when resistance is due to mutations selected during therapy, resulting in clinical failure. The inappropriate use or overuse of antibiotics has caused the major problem of polydrug resistance, and resistance patterns of bacteria to the new antibiotics are likely to parallel the extent to which they prescribe. Such general antibiotic resistance affects the market to a greater extent.\[[@CIT28]\] In today's world, even if the innovator obtains regulatory approval to market a newer antibiotic, its commercial success will be limited, as the regulatory agencies are likely to move forward on a recommendation for the sale of cheaper generic copies of the same drug to the public. Furthermore, in reality, newer antibiotics are kept in reserve so as to avoid resistance from the microorganisms. So a physician will not wait around deciding whether to prescribe the newer antibiotic or the older generic. They are likely to prescribe a generic anyway, in the hopes that the patient is among the two-thirds of the population who will get a greater protection against relapse. In the USA, a black box warning is a type of warning that appears on the package insert for prescription drugs that may cause serious adverse effects. It is so named for the black border that usually surrounds the text of the warning. Several antibiotics, including ciprofloxacin, gemifloxacin, levofloxacin, moxifloxacin, norfloxacin, and ofloxacin are affected by these blackbox warnings. Similarly "Dear Doctor" letter is intended to alert physicians to safety precautions that should be taken to reduce the potential risk reported to be associated with the drug products. Even though these warnings may be common for both the generic and branded drugs, these label warnings may create a negative impact on the public perception with the generic antibiotics. Apart from these, drug withdrawal and recall procedures affect the generic market in varying degrees. Recalls are an appropriate alternative method for removing or correcting marketed consumer products, their labeling, and/or promotional literature that violate the laws administered by the FDA.\[[@CIT29]\] Recalls may be conducted on a firm's own initiative, by FDA request or by FDA order, under statutory authority. Such type of FDA's activities affect the generic market mainly due to media coverage and are considered to be serious issues. Pharmaceutical firms that produce an antibiotic are usually given temporary monopoly power through a patent, granted to recover the incurred investment in R&D and by this to encourage future innovation of new drugs. The granting of this monopoly power ignores the fact that this also gives them some control over the levels of the drug's treatment efficacy on the one hand, as well as of the infected population on the other. In turn, a too intensive use of antibiotics within the community may lead to an increase in the bacterial resistance of the drug. Furthermore, literature from the USA has shown that brand name manufacturers do not compete on price once generic competitors become available. The lack of price competition may lead to increased costs in the private market. Private insurance companies generally do not require generic substitution and some provinces do not require generic substitution for cash-paying customers.\[[@CIT17]\] Furthermore, increasingly complex clinical trials are now required to gain approval, with different requirements in the USA and the European Union. These hurdles and drawbacks make the market a more complex one to predict, but still a setback lays for the discovery of newer antibiotics. Discussion and Conclusion {#sec1-9} ========================= Antibiotics are among the most frequently prescribed medications in modern medicine. Most antibiotics have 2 names, the trade or brand name, created by the drug company that manufactures the drug or otherwise known as innovator, and a generic name, based on the antibiotic's chemical structure or chemical class. The antibiotic market has evolved itself to a much stable place in spite of its complexity. The regulations of global health communities or agencies have been strictly developed on these antibiotics due to their potential in curing life-threatening diseases. Previously, a special regulatory care was given to antibiotics rather than any other drugs due to their relatively newness and also due to their significance in usage by public for life-saving purpose. Antibiotics were subjected to a batch certification requirement apart from complying with the monograph. This regulation required that each batch of antibiotic produced be certified to conform to the regulations of identity, strength, quality, and purity before marketing in the USA. This resulted in an aggressive cost increase in the antibiotic sector as well as time delay for approval. Nowadays, after several amendments to the regulations by the FDA, antibiotics are being regulated as any other pharmaceuticals. At the same time, industries are not showing interest in discovering newer antibiotics due to several factors, such as drug resistance, cost involved in the research, comparatively less profit due to industrial competencies, and so on. The introduction of Hatch--Waxman Act has made a paradigm shift in the industry by balancing the benefits for both the innovator and the generic industry. The introduction of only few newer antibiotics in the last 5 years shows the lack of awareness about these antibiotics. Although the infections were inhibited when improved sanitation and drugs were applied to combat microbes, and while many of the terrifying acute diseases, such as typhoid, cholera, and dysentery, have been subdued, many serious microbial diseases have not been eliminated. With the dangerously growing levels of antibiotic resistance and newer type of infections, organisms, and so on, it emphasizes the need for a more streamlined and enhanced means of developing and approving new agents, the need for greater integration of oversight and policy development efforts, and the necessity of greater availability of better data. Rather, regulations should be tightened up to avoid improper prescriptions and the misuse of these antibiotics. However, the introduction of Hatch--Waxman Act has in fact increased the generic antibiotic industries and has a definite effect on their growth. The law requires that these drugs must meet the specifications of the official compendia, as any other pharmaceuticals. Although several efforts have been taken by the regulatory agencies devoted to the tests and methods of assay of antibiotics, including sterility, biological test, microbiological and chemical assays, general and specific chemical tests, and tests on specific dosage forms, the generic industries struggle hard to bring the quality and safe antibiotics complying with these regulations. In future, in spite of the complexity and competency in the field, the wellbeing of the generic as well as the industries involved with the new drug inventions should be well protected to preserve the public health. In this article, generic Pharma industries with respect to antibiotics were explored. Considering the Hatch--Waxman Act, how the generic approvals made as fair has been explained. Also, history of antibiotics, its regulations, current market scenario, and several regulatory challenges facing antibiotic industries have also been discussed with an effort to bring into light a specialized field with challenges. However, with the increase in globalization and the demand for newer antibiotics, industries should be encouraged with focus on the future regarding the antibiotic. In summary, it is critical to devise strategies for antibiotic regulations, to ensure public health with wellbalanced commercial and regulatory aspects.\[[@CIT30]\] **Source of Support:**Nil **Conflict of Interest:** None declared

PubMed Central

2024-06-05T04:04:19.926356

2011-01-01

PMC3053507

Genetic testing is the use of laboratory tests to determine the genetic status of individuals already suspected to be at high risk for a particular genetic disorder based on family history or a positive screening test, and genetic testing and screening are similar in that both involve the use of laboratory tests to reveal the presence of disease-causing genes. The accelerating development of biochemical and DNA-based diagnostic tests for human genetic conditions in the last decade has engendered a revolution in genetic diagnosis. Finally, the pace of development and application of DNA and biochemical genetic tests and their acceptance by the public may be accelerated by the recent widespread media coverage of the work of human geneticists.\[[@CIT1]\] A genetic test is the analysis of human DNA, RNA, chromosomes, proteins, and certain metabolites in order to detect heritable disease-related genotypes, mutations, phenotypes, or karyotypes for clinical purposes.\[[@CIT2]\] Genetic tests also have varied rationale, including the analysis of genetic disease in newborns, children, and adults; the recognition of potential health risks; the prediction of drug responses; and the evaluation of risks to future children. Bioinformatics is a rapidly emerging field of biomedical research. A flood of large-scale genomic and postgenomic data means that many of the challenges in biomedical research are now the challenges in computational science. Clinical informatics has long developed methodologies to improve biomedical research and clinical care by integrating experimental and clinical information systems. The informatics revolution in both bioinformatics and clinical informatics will eventually change the current practice of medicine, including diagnostics, therapeutics, and prognostics. Postgenome informatics, powered by high-throughput technologies and genomic-scale databases, is likely to transform our biomedical understanding forever, in much the same way that biochemistry did a generation ago.\[[@CIT3]\] Characteristics of Genetic Testing {#sec1-1} ================================== In gene tests, scientists scan a patient's DNA sample for mutated sequences. A DNA sample can be obtained from any tissue, including blood. For some types of gene tests, researchers design short pieces of DNA called probes, whose sequences are complementary to the mutated sequences. These probes will seek their complement among the 3 billion base pairs of an individual's genome. If the mutated sequence is present in the patient's genome, the probe will bind to it and flag the mutation. Another type of DNA testing involves comparing the sequence of DNA bases in a patient's gene to a normal version of the gene. Cost of testing depends on the sizes of the genes and the numbers of mutations tested. Both genetic testing and genetic screening involve the same testing processes to examine an individual's chromosomes, DNA, or the biochemical product of a gene, typically a protein to confirm or refute a suspected chromosomal, DNA, or gene product change. Genetic screening is done for a particular condition in individuals, groups, or populations without family history of the condition, and genetic testing is done for a particular condition where an individual is suspected of being at increased risk due to their family history or the result of a genetic screening test. Advantages and Scope {#sec1-2} ==================== Advances in genome technology and other fruits of the Human Genome Project are playing a growing role in the delivery of health care. With the development of new technologies and opportunities for large-scale analysis of the genome, transcriptome, proteome, and metabolome, the genome sciences are poised to have a profound impact on clinical medicine. Cancer prognostics will be among the first major test cases for a genomic medicine paradigm, given that all cancer is caused by genomic instability, and microarrays allow assessment of patients' entire expressed genomes.\[[@CIT4]\] Genomic medicine aims to revolutionize health care by applying our growing understanding of the molecular basis of disease. Research in this arena is data intensive, which means data sets are large and highly heterogeneous. To create knowledge from data, researchers must integrate these large and diverse data sets. This presents daunting informatic challenges such as representation of data that is suitable for computational inference (knowledge representation), and linking heterogeneous data sets (data integration). Fortunately, many of these challenges can be classified as data integration problems, and technologies exist in the area of data integration that may be applied to these challenges.\[[@CIT5]\] Genetic Diagnosis {#sec1-3} ================= The diagnostic techniques outlined briefly above are a powerful new tool in all genetics, but most especially in the arena of human genetics. By application of these tools of biotechnology, and other techniques, molecular biologists and geneticists are providing a basis on which to make a genetic diagnosis. The identification of genetic disorders, and the potential for developing a therapy, is a powerful force in genetics and medicine. Diagnostic techniques can be used to identify specific proteins or fragments of DNA. These techniques can be used to detect viruses of animal and plant diseases, to detect harmful substances in the environment, or to match the DNA left at a crime scene to a possible criminal. Diagnostic techniques often utilize DNA probes, RFLP (restriction fragment length polymorphism) analysis, and mono- or polyclonal antibodies. Advances in genomics have led to mounting expectations in regard to their impact on health care and disease prevention. In light of this fact, a comprehensive research agenda is needed to move human genome discoveries into health practice in a way that maximizes health benefits and minimizes harm to individuals and populations. We present a framework for the continuum of multidisciplinary translation research that builds on previous characterization efforts in genomics and other areas in health care and prevention. The continuum includes four phases of translation research that revolve around the development of evidence-based guidelines. Phase 1 translation (T1) research seeks to move a basic genome-based discovery into a candidate health application (e.g., genetic test/intervention). Phase 2 translation (T2) research assesses the value of a genomic application for health practice leading to the development of evidence-based guidelines. Phase 3 translation (T3) research attempts to move evidence-based guidelines into health practice, through delivery, dissemination, and diffusion research. Phase 4 translation (T4) research seeks to evaluate the "real world" health outcomes of a genomic application in practice.\[[@CIT6]\] Types of Genetic Testing {#sec1-4} ======================== Genetic testing is a complex process, and the results depend both on reliable laboratory procedures and accurate interpretation of results. Tests also vary in sensitivity, that is, their ability to detect mutations or to detect all patients who have or will get the disease. Interpretation of test results is often complex even for trained physicians and other health care specialists. When interpreting the results of any genetic test, one must take into account the probability of false positive or false negative test results. Special training is required to be able to analyze and convey information about genetic testing to affected individuals and their families. Available types of genetic testing are listed in [Table 1](#T0001){ref-type="table"} ::: {#T0001 .table-wrap} Table 1 ::: {.caption} ###### Types of genetic test ::: Genetic test Feature --------------------------------------- ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Newborn screening Newborn screening is used just after birth to identify genetic disorders that can be treated early in life The routine testing of infants for certain disorders is the most widespread use of genetic testing Diagnostic testing Diagnostic testing is used to diagnose or rule out a specific genetic or chromosomal condition In many cases, genetic testing is used to confirm a diagnosis when a particular condition is suspected based on physical mutations and symptoms Diagnostic testing can be performed at any time during a person's life, but is not available for all genes or all genetic conditions Carrier testing Carrier testing is used to identify people who carry one copy of a gene mutation that, when present in two copies, causes a genetic disorder This type of testing is offered to individuals who have a family history of a genetic disorder and to people in ethnic groups with an increased risk of specific genetic conditions If both parents are tested, the test can provide information about a couple's risk of having a child with a genetic condition Prenatal testing Prenatal testing is used to detect changes in a fetus's genes or chromosomes before birth This type of testing is offered to couples with an increased risk of having a baby with a genetic or chromosomal disorder In some cases, prenatal testing can lessen a couple's uncertainty or help them decide whether to abort the pregnancy Preimplantation genetic diagnosis Genetic testing procedures that are performed on human embryos prior to the implantation as part of an *in vitro* fertilization procedure Predictive and presymptomatic testing Predictive and presymptomatic types of testing are used to detect gene mutations associated with disorders that appear after birth, often later in life These tests can be helpful to people who have a family member with a genetic disorder, but who have no features of the disorder themselves at the time of testing. Predictive testing can identify mutations that increase a person's chances of developing disorders with a genetic basis, such as certain types of cancer Forensic testing Forensic testing uses DNA sequences to identify an individual for legal purposes This type of testing can identify crime or catastrophe victims, rule out or implicate a crime suspect, or establish biological relationships between people (e.g., paternity) Parental testing This type of genetic test uses special DNA markers to identify the same or similar inheritance patterns between related individuals Research testing Research testing includes finding unknown genes, learning how genes work, and advancing our understanding of genetic conditions The results of testing done as part of a research study are usually not available to patients or their healthcare providers Pharmacogenomics This type of genetic testing determines the influence of genetic variation on drug response ::: Issues and Ethics {#sec1-5} ================= A comprehensive study of the significance and varieties of genetic discrimination is critical to design strategies to ensure the ethical and appropriate use of genetic testing in the future. The dominant theme noted in the responses in this study is that genetic conditions are regarded by many social institutions as extremely serious, disabling, or even lethal conditions without regard to the fact that many individuals with "abnormal" genotypes will either be perfectly healthy, have medical conditions which can be controlled by treatment, or experience only mild forms of a disease. As a result of this misconception, decisions by such institutions as insurance companies and employers are made solely on the basis of an associated diagnostic label rather than on the actual health status of the individual or family. The appropriate use of genetic testing information to restrict or limit access to public entitlements such as health care or employment has not been established and may not exist. The cost of such labeling is magnified by the fact that errors in testing and interpretation do occur.\[[@CIT1]\] The three basic components in genetic screening, that is, ethical, legal, and social issues, are to be considered and these genetic tests have to be performed with privacy, informed consent, and confidentiality. This brief discussion illustrates public expectations and fears about the effect of genomics, challenges to the goals of antidiscrimination laws and to the nature of the physician--patient relationship, and the contrasting perspectives and legal rules that apply to personal medical care and public health. Acknowledgment and examination of these complex issues are critical for identifying the appropriate ethical principles that should be applied and for creating the necessary legislative and regulatory responses.\[[@CIT7]\] Conclusion {#sec1-6} ========== Given this situation -- powerful and attractive new techniques, social and economic forces pressing for their application, and an incomplete understanding of the potential negative social and personal consequences of genetic testing -- concern about the burdens engendered by widespread utilization of genetic tests seems justified.\[[@CIT1]\] Genetic testing offers important opportunities for diagnosis and assessment of genetic risk. The sensitivity of tests for rare conditions will continue to improve as additional causative mutations are identified. Genetic tests are available to determine the risk of common diseases, but these often have limited predictive values. Evaluating the clinical usefulness of these tests will require a careful assessment of the risks and benefits of testing; the availability of specific measures to reduce risk in genetically susceptible people will be a major consideration.\[[@CIT2]\] With the sequencing of the human genome only months from its finish, the practice of medicine has now entered an era in which the individual patient's genome will help determine the optimal approach to care, whether it is preventive, diagnostic, or therapeutic. Genomics, which has quickly emerged as the central basic science of biomedical research, is poised to take center stage in clinical medicine as well.\[[@CIT8]\] As new genetic tests emerge, their translation into practice will depend not only on their performance based on laboratory standards, but also on their ability to enhance prevention or assist clinicians in diagnosing and treating patients.\[[@CIT9]\] Challenges of translating pharmacogenomics into clinical practice included ethical, social, legal, and economic issues.\[[@CIT10]\] There are many barriers to implementation of genetic medicine, including the cost of testing, the genetic literacy of patients and health care providers, and concerns about genetic discrimination; however, health care providers and patients must have realistic expectations about its predictive power and current limitations.\[[@CIT11]\] Thus it can be concluded that genetic testing is a complex process, and the results depend both on reliable laboratory procedures and accurate interpretation of results. **Source of Support:** Nil **Conflict of Interest:** None declared

PubMed Central

2024-06-05T04:04:19.929947

2011-01-01

PMC3053508

The *Salvadora persica* (Salvadoraceae) tree drives its Persian name, *Darakht-e-miswak* or tooth brush tree, from the fact that wood is much employed for the manufacturers of tooth brush. It is a large much-branched, evergreen shrub or a tree, found in the dry and arid regions of India, and on saline lands and in coastal regions just above the high water mark. Bark is dull grey or grey-white, deeply cracked, and leaves are variable in shape -- elliptic-ovate or ovate-lanceolate -- somewhat fleshy. Flowers are pedicellate, greenish-white or greenish-yellow in lax panicles, drupes are globose or round, smooth, red when ripe. The trees readily regenerate from seeds and coppice well \[Figures [1](#F0001){ref-type="fig"}--[3](#F0003){ref-type="fig"}\].\[[@CIT1]\] ::: {#F0001 .fig} Figure 1 ::: {.caption} ###### Miswak (*Salvadora persica*) leaves and root :::  ::: ::: {#F0002 .fig} Figure 2 ::: {.caption} ###### Miswak (*Salvadora persica*) stem branches with flowers and fruits :::  ::: ::: {#F0003 .fig} Figure 3 ::: {.caption} ###### Miswak (*Salvadora persica*) tooth brush :::  ::: Leaves are eaten as a vegetable in eastern tropical Africa and are used in the preparation of a sauce, and tender shoots and leaves are eaten as salad. Fruits are sweet and edible. A fermented drink is reported to be made from the leaves.\[[@CIT1]\] Fresh root bark is used as a vesicant and is employed as an ingredient of snuff. A paste of roots is applied as a substitute of mustard plaster and its decoction is used against gonorrhea and vesical catarrh. The extract of root is said to relieve the pain due to spleen troubles. A decoction of bark is used as a tonic and stimulant in low fevers and as an emmenagogue. Stem bark is used as an ascarifuge and for gastric troubles.\[[@CIT1]\] Leaves are bitter and possess antiscorbutic,\[[@CIT1]\] corrective, deobstruent, liver tonic, diuretic, analgesic, anthelmintic,\[[@CIT2]\] and astringent\[[@CIT1]--[@CIT2]\] properties and used in piles, scabies, leucoderma, strengthen the teeth, ozoena, and other nose troubles.\[[@CIT2]\] A decoction of leaves is used in asthma and cough, and a poultice made out of them is applied to painful piles and tumors. Leaves are also used as an external application in rheumatism. Dried leaves in small doses are given with copious amount of water for the treatment of flatulent dyspepsia.\[[@CIT1]\] Fruits possess lithontriptic,\[[@CIT1]\] carminative,\[[@CIT1]--[@CIT2]\] diuretic,\[[@CIT1]--[@CIT2]\] aphrodisiac, alexiteric, appetizer, and stomachic\[[@CIT1]--[@CIT2]\] properties and are used in biliousness,\[[@CIT1]--[@CIT2]\] and rheumatism.\[[@CIT1]\] Seeds have a bitter, sharp taste. They are considered as purgative, diuretic,\[[@CIT1]--[@CIT2]\] and liver tonic.\[[@CIT2]\] Seeds oil is applied on the skin in rheumatism.\[[@CIT1]\] Phytochemical Profile {#sec1-1} ===================== A phytochemical investigation of stems from *S. persica* by Khalil resulted in the first isolation of four benzylamides from a natural source. The isolated compounds were identified as butanediamide, *N*^1^, *N*^4^ -bis(phenylmethyl)-2(S)-hydroxy-butanediamide (I), *N*-benzyl-2-phenylacetamide (II), *N*-benzylbenzamide (III), and benzylurea (IV).\[[@CIT3]\] Phytochemical investigation revealed that it contains oleic, linolic, and stearic acids. Among the compounds identified are esters of fatty acids and of aromatic acids, and some terpenoids.\[[@CIT4]\] The major components from the essential oil of the toothbrush tree *S. persica* stem have been identified as 1,8-cineole (eucalyptol) (46%), α-caryophellene (13.4%), β-pinene (6.3%), and 9-epi-(E)-caryophellene.\[[@CIT5]\] GC-MS analysis of the volatile oil extracted from *S. persica* leaves revealed benzyl nitrile, eugenol, thymol, isothymol, eucalyptol, isoterpinolene, and β-caryophyllene as important constituents.\[[@CIT6]\] Sticks from *S. persica* have been analyzed for their soluble and total content of fluoride, calcium, phosphorus, and silica. There was a substantial amount of silica in the ashes of miswak.\[[@CIT7]\] The aqueous extract of stem and root of *S. persica* L. has also been investigated for some antimicrobial anionic components by using capillary electrophoresis techniques. It was reported that the root and stem extracts contain sulfate chloride, thiocynate, and nitrate.\[[@CIT8]\] Physicochemical analysis of air-dried root bark of *S. persica* was carried out by Bhandari in 1990. He found that it contains 27.1% ash, consisting of considerable amounts of salts, mostly as chlorides. The drug has large amount of alkloidal constituents (including trimethyl amine and unidentified alkaloids), small amount of resin and coloring matter, and traces of tannins and saponins. Higher concentration of fluoride and silica, sulfur, vitamin C, small amount of flavonoids and sterols were also reported.\[[@CIT9]--[@CIT11]\] Three lignin glycosides have been reported from the stem of *S. persica*.\[[@CIT12]\] The flavonoids rutin and quercetin were detected in the stem of *S. persica*.\[[@CIT13]\] Salvadourea has been reported in the root of *S. persica*.\[[@CIT14]\] Benzylisothiocynate was also isolated from the root.\[[@CIT15]\] Salvadoricine, a new indole alkaloid, was reported in the leaves of *S. persica*.\[[@CIT16]\] Pharmacological Profile {#sec1-2} ======================= Antimicrobial activities {#sec2-1} ------------------------ Aqueous and methanol extracts of *S. persica* were investigated by Firas *et al*. for its antimicrobial activities against seven isolated oral pathogens -- *Staphylococcus aureus, Streptococcus mutans, Strep. faecalis, Strep. pyogenis, Lactobacillus acidophilus, Pseudomonas aeruginosa, and Candida albicans* -- using disc diffusion and microwell dilution assays. According to both antimicrobial assays, the aqueous extract inhibited all isolated microorganisms, especially the *Streptococcus* spp., and was more efficient than the methanol extract, which was resisted by *L. acidophilus* and *P. aeruginosa.* The strongest antibacterial activity was observed using the aqueous extract against *Strep. faecalis* (zone of inhibition: 22.3 mm; MIC: 0.781 mg/ml). Both extracts had equal antifungal activity against *C. albicans* based on the turbidity test (MIC: 6.25 mg/ml).\[[@CIT17]\] *In vitro* antibacterial effect of miswak pieces without extraction has been found most pronounced on *P. gingivalis, A. actinomycetemcomitans,* and *H. influenzae*, less on *Strep. mutans*, and least on *L. acidophilus*. Miswak embedded in agar, or suspended above the agar plate, had strong antibacterial effects against all bacteria tested. The antibacterial effect of suspended miswak pieces suggested the presence of volatile active antibacterial compounds.\[[@CIT18]\] Miswak (*S. persica*) extract inhibits the growth of some dental plaque bacteria, and antibacterial effect of the herbal toothpaste was significantly greater than that of the placebo.\[[@CIT19]\] Aqueous extracts of miswak and derum enhance the growth of fibroblasts and inhibit the growth of cariogenic bacteria, with the derum extract showing greater activity than miswak.\[[@CIT20]\] Antimicrobial activity of eight commercially available mouthrinses and 50% miswak extract against seven microorganisms was compared by Almas and Ahmad in 2005. Corsodyl, Alprox, Oral-B advantage, Florosept, Sensodyne, Aquafresh Mint, Betadine, and Emoform mouthrinses were used, while 50% aqueous extract of miswak (*S. persica*) was used against *Strep. faecalis, Strep. pyogenis, Strep. mutans, C. albicans, Staph. aureus*, and *Staph. epidermidis.* Mouthrinses containing chlorhexidine (CHX) had maximum antibacterial activity, while cetylpyridinium chloride mouthrinses had moderate, and miswak extract had low antibacterial activity.\[[@CIT21]\] Antimicrobial activity of Neem and Arak chewing stick's aqueous extracts at various concentrations was compared by some research workers. Data suggested that both chewing stick extracts was effective at 50% concentration on *Strep. mutans* and *Strep. faecalis*. Arak extract was more effective at lower concentrations for *Strep. faecalis.*\[[@CIT22]\] Cytotoxic activity {#sec2-2} ------------------ The cytotoxic activity of *S. persica* and CHX *was* evaluated by Rajabalian *et al*. in 2009. The results indicated that both persica and CHX mouthwashes were toxic to macrophage, epithelial, fibroblast, and osteoblast cells in a concentration-dependent manner.\[[@CIT23]\] Tick-repellent properties {#sec2-3} ------------------------- The *S. persica, Pistacia,* and *Juniperus phoenicea* were evaluated by Garboui *et al*. using host-seeking nymphs of *Ixodes ricinus* in the laboratory for tick-repellent effects of the essential oils. Significant tick-repellent effects were observed for the oils of all three species, but the duration of action was short.\[[@CIT24]\] Antidental caries potential {#sec2-4} --------------------------- The efficacy of natural toothbrush or miswak in the prevention of dental caries has been investigated and compared with the efficacy of ordinary toothbrush and toothpaste. The data collected at the end of the study showed that the risk of dental caries for each tooth in the control group was 9.35 times more than the case group.\[[@CIT25]\] Less than two-thirds of the sampled adults followed the recommended toothbrushing frequency of twice daily or more, and the majority of subjects did not have a preventive dental visit in the previous 6 months. Furthermore, most subjects reported multiple oral health problems that are mostly preventable through adequate oral hygiene habits and regular preventive dental visits.\[[@CIT26]\] Rinsing with miswak extract (*S. persica*) stimulated parotid gland secretion and raised the plaque pH, suggesting a potential role in caries prevention.\[[@CIT27]\] Anti-inflammatory and analgesic potential {#sec2-5} ----------------------------------------- Mansour *et al.* evaluated the extract of root and branches of *S. persica* for analgesic activity in mice. It was found that the drug possesses a relatively moderate analgesic effect which might be due to interaction with the central and/or peripheral opiate system.\[[@CIT28]\] The extract of stem of *S. persica* was reported to possess anti-inflammatory activity.\[[@CIT29]\] ACE-inhibiting ability {#sec2-6} ---------------------- *In vitro* screening has reported that *S. persica* possesses high ACE-inhibiting ability.\[[@CIT30]\] Antiplasmodial activity {#sec2-7} ----------------------- Nineteen plant species, used traditionally in Sudan against malaria and similar tropical diseases, were evaluated for pharmacological activity by Ali *et al.* Different extracts of *S. persica* against *P. falciparum* NF54 strain were found to possess antiplasmodial activity.\[[@CIT31]\] Antiplaque activity {#sec2-8} ------------------- It has been observed that miswak was as effective as a toothbrush for reducing plaque on buccal teeth surfaces both experimentally and clinically.\[[@CIT32]\] The water extract (10%) of *S. persica* is an effective antimicrobial agent when utilized clinically as an irrigant in the endodontic treatment of teeth with necrotic pulps.\[[@CIT33]\] Another study compared the oral health efficacy of persica mouthwash (containing an extract of *S. persica*) with that of a placebo. The study showed that use of persica mouthwash improves gingival health and lower carriage rate of cariogenic bacteria when compared with the pretreatment values. Neither the persica nor the placebo reduced the accumulation of dental plaque.\[[@CIT34]\] Scientific evaluation of use of miswak revealed that it is at least as effective as toothbrushing for reducing plaque and gingivitis and that the antimicrobial effect of *S. persica* is beneficial for prevention/treatment of periodontal disease.\[[@CIT35]\] A clinical study was conducted using patients' saliva and measuring the effect of miswak (chewing stick), miswak extract, toothbrush, and normal saline on mutans and lactobacilli by Almas and Al-Zeid. The results showed that there was a marked reduction in *Strep. mutans* among all groups. When the groups were compared, the reduction in *Strep. mutans* was significantly greater using miswak in comparison to toothbrushing and there was no significant difference for lactobacilli reduction. The investigators concluded that miswak has an immediate antimicrobial effect. *Strep. mutans* were more susceptible to miswak antimicrobial activity than lactobacilli.\[[@CIT36]\] Persica mouthwash significantly lowers the gingival index, plaque index, and bleeding index in case group without any reported side effects.\[[@CIT37]\] Effects on fertility {#sec2-9} -------------------- Darmani *et al*. investigated the effects of an extract of miswak for 30 days on the reproductive system of the mouse. The results showed that the exposure to miswak extract did not have much effect on female mouse fertility, although it caused a significant decrease in the relative weights of the ovary and an increase in uterine weights. Exposure of male mice to miswak extract resulted in a 72% reduction in pregnancies in untreated females impregnated by test males. The relative weights of the testes and preputial glands were significantly increased and that of the seminal vesicles was significantly decreased in test males.\[[@CIT38]\] Anticonvulsant and sedative potential {#sec2-10} ------------------------------------- The effect of *S. persica* stem extracts on the potentiation of sodium pentobarbital activity and on generalized tonic-clonic seizure produced by pentylenetetrazol (PTZ) on the rats was observed by Monforte *et al*. The extracts of *S. persica* extended sleeping time and decreased induction time induced by sodium pentobarbital; in addition it showed protection against pentylenetetrazol-induced convulsion by increasing the latency period and diminishing the death rate.\[[@CIT39]\] Antiulcer activity {#sec2-11} ------------------ The antiulcer activity of decoction of *S. persica* has been reported against ASA-induced ulcer in rats. The ulcer index significantly decreased after the treatment with a lyophilized decoction of *S. persica* (500 mg/kg, os), once daily for 7 days, with respect to controls. Moreover, *S. persica* decoction possesses significant anti-inflammatory activity.\[[@CIT40]\] The other study was designed to confirm the antiulcer activity of *S. persica* decoction using optical microscopy. The elements of gastric mucosa tended to be reestablished normally in treated rats.\[[@CIT41]\] Removal of smear layer and occlusion {#sec2-12} ------------------------------------ Soaking the healthy and periodontally diseased root dentine in miswak extract resulted in partial removal of smear layer, and occlusion of dentinal tubules was observed in dentine specimens brushed with miswak solution.\[[@CIT42]\] *S. persica* contains potential antimicrobial anionic components, and the capillary electrophoresis is a convenient method for their identification and quantification.\[[@CIT43]\] Antihyperlipidemic activity {#sec2-13} --------------------------- The effects of prolonged administration of a lyophilized stem decoction of *S. persica* have also been investigated in diet-induced rat hypercholesterolemia. The results showed that the *S. persica* decoction significantly lowered cholesterol and LDL plasma levels in rats.\[[@CIT44]\] Antimycotic potential {#sec2-14} --------------------- Al-Bagieh *et al.* showed that miswak extract at a concentration of 15% and above has a fungistatic effect for up to 48 hours. The antimycotic effect was probably due to one or more of the root contents which included chlorine, trimethylamine, and alkaloid resin, and sulfur compounds.\[[@CIT45]\] Locomotor activity {#sec2-15} ------------------ Mice injected with *S. persica* extracts showed significantly low exploratory locomotor activity.\[[@CIT46]\] Hypoglycemic activity {#sec2-16} --------------------- Trovato *et al*. observed significant hypoglycemic activity of *S. persica* in rats.\[[@CIT47]\] Clinical Study {#sec1-3} ============== The effects of CHX and persica mouth rinses on periodontal status of patients undergoing fixed orthodontic were compared by Poosti *et al*. Gingival index had a significant reduction in all groups after prescribing mouth rinses but this reduction was not significant between groups. Mean pocket depth in CHX group and gingival bleeding index in persica group had significant reduction. Plaque index did not show significant reduction in any of the groups.\[[@CIT48]\] Conclusion {#sec1-4} ========== It is concluded that miswak (*S. persica*) reduces the microbial count in different groups and improves the oral health. The extract possesses antibacterial and antiplaque property and it can be used effectively as a natural tool for teeth cleansing and as a natural analgesic for the disturbing toothache. The drug is also reported to possess anti-inflammatory, anticonvulsant, sedative, antiulcer, hypolipidemic, and hypoglycemic activities. The present review showed that it is useful in a number of diseases. Therefore it is imperative that more clinical and pharmacological studies should be conducted to investigate unexploited potential of this plant. The research workers have isolated many phytoconstituents from the plant. Nevertheless further investigations are required to isolate and purify novel pharmacologically active and industrially important compounds. **Source of Support:** Nil **Conflict of Interest:** None declared

PubMed Central

2024-06-05T04:04:19.931691

2011-01-01

PMC3053509

Animals undergoing external or internal challenge to their state of health mount a vigorous response including activation of both the innate and acquired immune systems. The innate immune system which covers those aspects of the host defense mechanisms not dependent on specific response, such as production of antibody, not only stimulates leukocyte activity but also effects many aspects of the host's metabolic processes. The varied reactions of the host to infection, inflammation, or trauma are collectively known as the acute-phase response (APR) and encompass a wide range of pathophysiological responses such as pyrexia, leukocytosis, hormone alterations, and muscle protein depletion combining to minimize tissue damage while enhancing the repair process.\[[@CIT1]\] Another of these systemic responses to disease is an increase in the production by the liver of a number of plasma proteins which are known collectively as the acute-phase proteins (APP).\[[@CIT2]--[@CIT4]\] The APR is a very complex reaction, involving local and systemic effects. One of these effects corresponds to changes in the concentration of some plasma proteins, mainly synthesized in the liver, which are called APP. The APR is induced by protein hormones called cytokines acting as messengers between the local site of injury and the hepatocytes synthesizing the APPs. Most cytokines have multiple sources, multiple targets, and multiple functions,\[[@CIT5]\] and they have been found in a large number of animal species including mammals, birds, fish, reptiles, and starfish.\[[@CIT6]--[@CIT10]\] The changes in the concentrations of APPs are largely due to changes in their production by hepatocytes. The magnitude of the increases varies from about 50% in the case of C-reactive protein (CRP) and serum amyloid A (SAA). Under the influence of interleukin (IL), i.e., IL-1, IL-2, and tumor necrosis factor -- alpha (TNF-α), liver cells synthesize and secrete APPs. The maximum serum concentration of APPs is typically reached within 24 to 48 h after the initiation. A decline coinciding with the recovery from the infection is seen,\[[@CIT11]\] and generally, feed-back regulations will limit the response leading to its resolution within 4--7 days after the initial stimulus if no further stimulus occurs. When the receptor triggering has repeated pulses, the APR can become chronic. Chronic inflammation (e.g., arthritis) can be perceived as a consecutive series of separate inflammatory stimuli. In such conditions, increased serum concentrations of APPs are generally observed.\[[@CIT12]\] However, the increase is lower than during acute episodes of inflammation or infection. There are also indications that the response to chronic compared to acute inflammation varies from one protein to another.\[[@CIT13]\] The three most important APPs are CRP, serum amyloid P (SAP), and SAA.\[[@CIT14]\] Many APPs, such as CRP and SAA bind to microbial cell walls and they may act as opsonins and fix complement, thus promoting the elimination of microbes. Mechanism of Synthesis of Acute Phase Proteins {#sec1-1} ============================================== An APR is characterized, among other things, by fever and increases the numbers of peripherals leukocytes, in particulars, increasing the numbers of circulating neutrophils and their precursors. At the same time, cellular and biochemical alterations, in particulars the coordinated synthesis, of so-called APPs or APRs by hepatocytes take place in the liver \[[Figure 1](#F0001){ref-type="fig"}\]. ::: {#F0001 .fig} Figure 1 ::: {.caption} ###### Synthesis of acute phase proteins :::  ::: Regulation of Acute Phase Reactions and Synthesis of Acute Phase Proteins {#sec1-2} ========================================================================= According to Beutler and Cerami,\[[@CIT15]\] the APR is stimulated by the release of cytokines such as IL-1, IL-6, and TNF-α from macrophages and monocytes at the site of inflammatory lesions or infections. Inflammatory cytokines such as IL-6, IL-1, TNF, and others such as transforming growth factor (TGF) and interferon (IFN) are produced by inflammatory cells. These proinflammatory cytokines induce local and systemic reactions.\[[@CIT16]\] These mediators are involved in cell activation of leucocytes, fibroblast, endothelial cells, and smooth muscle cells, result in a systemic release of cytokines, increase in the circulation of the cytokines, and then stimulate the hepatic APR.\[[@CIT17]\] Systemic reaction results in activation of the hypothalamus, reduction in growth hormone secretion, and a number of other physiological changes characterized by fever, anorexia, and catabolism of muscle cells. TNF-α, IL-1β, and INF-γ are crucial for the expression of inflammatory mediators such as prostaglandins and leukotrienes and they induce the production of platelet-activating factor and IL-6. After stimulation by proinflammatory cytokines, Kuffer cells in the liver produce IL-6 and present it to the hepatocytes. Thus, IL-6 is the major mediator for the hepatocytic secretion of most of the APPs. Activities are enhanced indirectly by activation of the pituitary/adrenal gland axis, which involves synthesis of adrenocorticotrophic hormone (ACTH) and subsequent production of cortisal. The increase in glucocorticoids during the APR is a result of cytokine stimulation of the pituitary--adrenal axis to produce adrenocorticotrophic hormone.\[[@CIT18]\] As a result, an increase in corticosterone, the main glucocorticoids, is observed later than the appearance of IL-6.\[[@CIT19]\] Cortisol can enhance expression of IL-6 receptors in liver cells and thus promotes IL-6-mediated synthesis of APPs. Glucocorticoids are hormone that can be involved in APR as in mammals.\[[@CIT20]\] The role of glucocorticoids in birds seems to be both stimulatory and regulatory.\[[@CIT19]\] The administration of glucocorticoids to domestic fowl can also stimulate APP synthesis hormone,\[[@CIT18]\] which suggest that glucocorticoids may work independently of cytokines. Negative regulatory loops can involve inhibition of synthesis of IL-6, IL-1, and TNF by cortisol and inhibition of the synthesis of IL-1 and TNF in monocytes by IL-6. Of all mediators participating in the induction and regulation of APP synthesis, IL-6 appears to induce the broadest spectrum of APPs whereas IL-1 and TNF only induce the synthesis of subsets of these proteins. The mechanism for stimulation of hepatic production of APPs by proinflammatory cytokines has been extensively studied. Induction of the APPs by IL-1 following binding to the IL-6 receptor causes phosphorylation and degradation of inhibitor kappa B (IKB). The inhibitor of transcription factor nuclear factor kappa B (NF-kB) leads to the release of NF-kB and subsequent activation of acute-phase gene in nucleus.\[[@CIT21]\] Classification of Acute-phase Proteins {#sec1-3} ====================================== On the basis of protein concentrations {#sec2-1} -------------------------------------- ### Negative acute-phase proteins {#sec3-} The liver responds by producing a large number of APRs. At the same time, the production of a number of other proteins is reduced; these are therefore referred to as "negative" APPs. Negative APPs are albumin, transferring, transthyretin, transcortin, and retinol-binding protein. ### Positive acute-phase proteins {#sec3-2} Positive APPs are CRP, D-dimer protein, mannose-binding protein, alpha 1 antitrpysin, alpha 1 antichymotrypsin, alpha 2 macroglobulin, fibrinogen, prothrombin, factor VIII, von-Willebrand factor, plasminogen, complement factors, ferritin, SAP complement, SAA, ceruloplasmin (Cp), and haptoglobin (Hp). Positive APPs serve different physiological functions for the immune system. Some act to destroy or inhibit growth of microbes, e.g., CAA and Hp. Others give negative feedback on the inflammatory response, e.g., serpins, alpha 2 macroglobulin and coagulation factors affect coagulation. Positive APPs are produced during the APR associated with anorexia and changed metabolism. On the basis of their mode of action {#sec2-2} ------------------------------------ APP classified as below: Protease inhibitors, e.g., alpha 1 antitrypsin, alpha 1 antichymotrypsin.Coagulation proteins, e.g., fibrinogen, prothrombin.Complement proteins, e.g., C2, C3, C4, C5, etc.Transport proteins, e.g., Hp, Cp, hemopexin.Other proteins, e.g., CRP, SAA, SAP, acid glycoprotein (AGP). Classification depending on the basis of their increased / decreased concentration in different species {#sec2-3} ------------------------------------------------------------------------------------------------------- The concentration of most of the APP increases, whereas other plasma proteins show decrease in their basal levels. Some APPs are present at very low concentration in normal state and may show increase up to 100 fold. This is the case of CRP or SAA in humans. Others increase between 2 to 10 times, whereas minor APPs are modified less than twofold. The APP pattern may vary from one species to another. As example, CRP that is a major APP in humans or dogs does not modify its concentration in cattle or cats. In pig, a higher CRP serum concentration was observed in pigs with compared to without clinical signs of acute inflammation.\[[@CIT22]\] Other main APP in the pigs is CRP and Hp (increase of 8--10 and 2--10 times, respectively, in the turpentine model). SAA has also been described as a major APP and Cp is a minor APP in pigs. Beside albumin, fetuin and transferring are negative APPs (decreases of 20-40%). In cattle, Hp and SAA are major APPs while fibrinogen, alpha-AGP, Cp, and alpha-antitrypsin are minor APPs in the cattle. SAA is the most studied APP in cattle. It can increase around 2--8 times during an APR and seems to react faster than Hp after the inflammatory stimuli. In sheep, haptoglobin is a major APP in the sheep. Its concentration was raised up to 100 times after injection of yeast, whereas Cp and fibrinogen increased around four times, and albumin decreased. Increases of SAA of around 10 times normal values have been observed in ewes with mastitis induced experimentally. SAA also increased in milk. In dogs, the behavior of CRP is similar as in humans. The concentration of CRP dramatically rises from undetectable levels to around 100 mg/mL in the first 24 h after surgery, declining after that. Hp, alpha-AGP, and ceruplasmin increased moderately (around twofold), and the concentration of alpha-antitrypsin was not modified.\[[@CIT23]\] General role of APPs in the body {#sec2-4} -------------------------------- The function of positive APPs is regarded as important in optimization and trapping of microorganism and their products, in activating the complement system, in binding cellular remnants like nuclear fractions, in neutralizing enzymes, scavenging free hemoglobin and radicals, and in modulating the host's immune response. CRP is the first described APP in 1930. It binds directly to several microorganisms, and activates the complement system by the classical C1q pathway, and acts as opsonins. SAA was first described in 1994 and is an apolipoprotein of high-density lipoproteins (HDL). An APP is thought to influence HDL--cholesterol transport. In tissues, it attracts inflammatory cells and inhibits the respiratory burst of leukocytes.\[[@CIT24]\] It is also described to bind lipopolysaccharide (LPS), comparable to LPS binding protein. Hp strongly binds to hemoglobin, has anti-inflammatory capabilities, and binds to integrins on leukocytes. Although representing a positive APP, its quantity may decrease on massive erythrolysis and when blood is hemolytic. Cp has histaminase, ferroxidase activity, and scavanges Fe^2+^ and free radicals, while α~2~ macroglobulin (α~2~MG) binds to proteolytic enzymes \[[Table 1](#T0001){ref-type="table"}\]. ::: {#T0001 .table-wrap} Table 1 ::: {.caption} ###### Biological activities of selected acute phase proteins ::: Acute phase protein Biological activity --------------------- ----------------------------------------------------------------------- Haptoglobin Binds with hemoglobin Bacteriostatic effect Stimulation of angiogenesis Role in lipid metabolism/development of fatty liver in cattle Immunomodulatory effect Inhibition of neutrophils respiratory burst activity C-reactive protein Complement activation and opsonization Modulation of monocytes and macrophages, cytokine production Binding of chromatin Prevention of tissue migration of neutrophils Serum amyloid A Transport of cholesterol from dying cells to hepatocytes Inhibitory effect on fever Inhibitory effect on the oxidative burst of neutrophilic granulocytes Inhibitory effect on *in vitro* immune response Chemotexic effect on monocytes, leukocytes, and T cells Induction of calcium mobilization by monocytes Inhibition of platelet activation ::: Pattern recognition molecules, pentraxins, and C-reactive proteins {#sec2-5} ------------------------------------------------------------------ In the past, the innate immune system was considered to be a primitive static system; nowadays delve into its complexity. It is a system that is able to recognize and respond to danger signals represented by a limited number of highly conserved structures of microorganisms \[pathogen-associated molecular patterns (PAMPs)\] and several cell products associated with a breach in defenses. For this purpose, the innate immune system possesses a large number of soluble (e.g., pentaxins), membrane-bound \[e.g., toll-like receptors (TLR)\], and cytosolic (e.g., nod-like receptors) "receptors."\[[@CIT25]\] They are known collectively as pathogen recognition receptors, or pattern recognition receptors, but a more accurate term is pattern recognition molecules (PRMs). Pentraxins are superfamily of proteins, phylogenetically conserved from arachnids to mammals and characterized by the presence of their carboxyl and terminal of a 200 amino acid pentraxin domain. The pentraxin was first assigned to CRP for its ultrastuctural appearance of five subunits. These protein pentraxins have been around in the animal kingdom for some time, since a closely related homolog, limulin, is present in the hemolymph of the horseshoe crab, not exactly a close relative of *Homo sapiens*. Human CRP is composed of five identical polypeptide units noncovalently arranged as a cyclic pentamer around a Ca-binding cavity. Based on the primary structure of the subunits, the pentraxin are divided into short and long pentraxins. The short pentraxins reactive proteins and serum amyloid pentraxins component are produced by liver and represent the main APPs in human and mouse, respectively. The long pentraxins, i.e., PTX3, are produced by innate immunity cells \[e.g., polymorphic mononuclear cells (PMN), macrophages, and dendritic cells\], interact with several ligands, and play an essential role in innate immunity.\[[@CIT26]\] PTX3 provides a paradigm for mode of action of humoral innate immunity. Thus, pentraxins recognize a wide range of exogenous pathogenic substances and altered self-molecules and in species-specific manner behave as APPs.\[[@CIT27]\] A major property of CRP is its ability to bind in a Ca-dependent fashion, as a pattern recognition molecule, to a number of microorganisms which contain phosphorylcholine in their membranes, the complex having the useful property of activating complement. This results in the deposition of C3b on the surface of the microbe which thus becomes opsonized (i.e., made ready for the table) for adherence to phagocytes.\[[@CIT28]\] CRP was originally discovered by Tillett and Francis\[[@CIT29]\] in 1930 as a substance in the serum of patients with acute inflammation that reacted with the C-polysaccharide by the liver and by adipocytes. ### Functions {#sec3-3} CRP levels rise dramatically during inflammatory processes occurring in the body. CRP rises up to 50,000 fold in acute inflammation, such as infection. It rises above normal limits within 6 h, and peaks at 48 h. CRP binds to phosphorylcholine on microbes. It is thought to assist incomplete binding to foreign and damaged cells and a cell enhances phagocytosis by macrophages, which express a receptor for CRP. It is also believed to play an important role in innate immunity, as an early defense system against infections. ### Diagnostic use {#sec3-4} CRP is used mainly as a marker of inflammation and infection. Measuring and charting CRP values can prove useful in determining disease progress or the effectiveness or treatments. Viral infections tend to give a lower CRP level than bacterial infection. ### Role in cardiovascular disease {#sec3-5} Patients with elevated basal levels of CRP are at an increased risk of diabetes, hypertension, and cardiovascular disease. CRP can exacerbate ischemic necrosis in a complement-dependent fashion and that CRP inhibition can be a safe and effective therapy for myocardial and cerebral infarcts; this has only been demonstrated in animal models. ### Diagnostic test {#sec3-6} Various analytical methods are available for CRP determination, such as enzyme linked immunosorbent assay (ELISA), immunoturbidimetry, rapid immunodiffusion, and visual agglutination. To measure the CRP level, a "high-sensitivity" CRP or hs-CRP test needs to be performed and analyzed by a laboratory. This is an automated blood test designed for greater accuracy in measuring low levels of CRP, which allows the physician to assess cardiovascular risk. If a result in the low-risk range is found (\<1 mg/L), it does not repeating. Higher levels need repeating, and clinical evaluation as necessary. Relevance of genetic vs environmental determinants of CRP {#sec2-6} --------------------------------------------------------- Elevated plasma levels of CRP are associated with increased risks of ischemic heart disease and ischemic cerebrovascular disease.\[[@CIT30]--[@CIT33]\] The random assortment of genes that occurs during gamete formation provides a relatively unbiased method of assessing whether risk factors that have a genetic component are in fact causally related to clinical outcomes. This phenomenon has sometimes been termed "mendelian randomization." Thus, genetic variants that specifically increase plasma levels of CRP\[[@CIT34][@CIT35]\] provide an ideal system to assess the consequences of lifelong high CRP levels, independently of other risk factors.\[[@CIT36]\] According to the study conducted by Zacho *et al*.,\[[@CIT37]\] on genetically elevated CRP and ischemic vascular disease showed that CRP genetic variation was associated with elevated CRP levels without predicting an increased risk of ischemic vascular disease. Genetic variants that are associated with lifelong increases in plasma CRP levels are not associated with an increased risk of ischemic heart disease or ischemic cerebrovascular disease. The increase in the risk of ischemic vascular disease associated with higher plasma CRP levels observed in epidemiological studies may not be causal, but rather that increased CRP levels are simply a marker for atherosclerosis and ischemic vascular disease. Serum amyloid A {#sec2-7} --------------- Serum amyloid A (SAA) proteins are a family of apolipoproteins and produced by the liver. These proteins play a highly essential role in all animals. Acute phase SAA proteins (A-SAAs) are secreted during the acute phase of inflammation. These proteins have several roles, including the transport of cholesterol to the liver for secretion into the biles, the recruitment of immune cells to inflammatory sites, and the induction of enzymes that degrade, such as amyloideosis, atherosclerosis, and rheumatoid arthritis. Several isotypes of SAA are found; types 1 and 2 represent positive APPs. In the bovine, also a negative protein crossreacting with anti-SAA serum has been described.\[[@CIT38]\] The acute phase SAA isoforms have been reported in mice, called SAA1, SAA2, and SAA3. Besides the acute phase SAAs, constitutive variants are described.\[[@CIT39]\] Human SAA4 is normally present in serum.\[[@CIT40][@CIT41]\] Rabbit SAA3\[[@CIT42]\] is formed by synoviocytes, fibroblasts, and macrophages, and is not a blood protein. The mammary gland is a well-known source of an SAA3 variant\[[@CIT43]--[@CIT45]\] occurring in colostrum and in mastitis milk that should have beneficial functions for the gut mucosa of the offspring.\[[@CIT46]--[@CIT48]\] Haptoglobin {#sec2-8} ----------- Haptoglobin (Hp) is a protein in the blood plasma that binds free hemoglobin released from erythrocytes with affinity and thereby inhibits its oxidative activity. The haptoglobin--hemoglobin complex is used to screen for and monitor intravascular hemolytic anemia. ### Clinical significance {#sec3-7} Haptoglobin is produced mostly by hepatocytes but also by other tissues, e.g., skin, lung, and kidney. Reticuloendothelial system will remove the haptoglobin--hemoglobin complex from the body; haptoglobin levels will be decreased in hemolytic anemia. In the process of binding hemoglobin, haptoglobin sequesters the iron within hemoglobin, preventing iron-utilizing bacteria from benefiting from hemolysis. Haptoglobin is ordered whenever a patient exhibits symptoms of anemia, such as pallor, fatigue, shortness of breath, along with physical signs of hemolysis, such as jaundice or dark-colored urine. Decreases in haptoglobin can support a diagnosis of hemolytic anemia, especially when correlated with a decreased RBC count, hemoglobin and hematocrit, and also an increased reticulocyte count. If the reticulocyte count is increased but the haptoglobin level is normal, this may indicate that the cellular destruction is occurring in the spleen and liver, which may indicate a drug-induced hemolysis or a red cell dysplasia. The spleen and liver recognize an error in the red cell and destroy the cell. This type of destruction does not release hemoglobin into the peripheral blood, so the haptoglobin cannot bind to it. Thus, the haptoglobin will stay normal. If there are symptoms of anemia, it is most likely not due to hemolysis but instead some other error in cellular production, such as aplastic anemia. Haptoglobin levels which are decreased but do not accompany signs of anemia may indicate liver damage, as the liver is not producing enough to begin with \[[Table 2](#T0002){ref-type="table"}\]. ::: {#T0002 .table-wrap} Table 2 ::: {.caption} ###### Haptoglobin level in various species ::: Species Normal range (Mg/dL) Increase in APR (Mg/dL) --------- ---------------------- ------------------------- Bovines 0.0--0.5 1.0--3.0 and \< Canines 0.3--3.6--5 4.0--9.0 Porcine 0.0--2.2 3.0--8.0 Felines 0.7--2.0 3.0--10 Ovine 0.0--1.0 0.0--3.0 Humans 1.0--3.0 4.3--7.8 ::: Mannose-binding protein {#sec2-9} ----------------------- The most important acute phase opsonin is the Ca-dependent mannose-binding protein (MBP), which can react not only with mannose but several other sugars, so enabling it to bind with an exceptionally wide variety of Gram-negative and -positive bacteria, yeasts, viruses, and parasites; its subsequent ability to trigger the classical C3 convertase through two novel associated serine proteases (MASP-1 and MASP-2) qualifies it as an opsonins. MBP is a multiple of trimeric complexes, each unit of which contains a collagen-like region joined to a globular lectin-binding domain. This structure places it in the family of collectins (collagen + lectin) which have the ability to recognize "foreign" carbohydrate patterns differing from "self" surface polysaccharides normally decorated by terminal galactose and sialic acid groups, while the collagen region can bind to and activate phagocytic cells through complementary receptors on their surface.\[[@CIT49]\] Collectins are a group of proteins containing C-type carbohydrate recognition domains (CRD) attached to collagen-like regions via a-helical coiled-coil regions.\[[@CIT50]\] The group includes mannan-binding lectin (MBL), surfactant proteins A and D (SP-A and SP-D), conglutinin, 43-kDa collectin (CL-43), and the recently identified CL-L1 and CL-P1.\[[@CIT51][@CIT52]\] Conglutinin and CL-43 have so far only been identified in the *Bovidae*. MBL, conglutinin, and CL-43 are plasma proteins synthesized in the liver. The collectins play an important role in the nonadaptive immune defense, as demonstrated by the finding that SP-A- or SP-D-deficient mice are susceptible to a variety of infections.\[[@CIT53][@CIT54]\] The collectins, especially MBP and the alveolar surfactant molecules SP-A and SP-D, have many attributes that qualify them for a first-line role in innate immunity. These include the ability to differentiate self from nonself, to bind to a variety of microbes, to generate secondary effector mechanisms, and to be widely distributed throughout the body including mucosal secretions. MBL is the only collectin that activates the complement system. After binding to microorganisms, the MBL-associated serine proteases cleave and activate C4, C2, and C3.\[[@CIT55]\] This may lead directly to complement-mediated lysis of the microorganisms or may indirectly increase the opsonization mediated by deposition of C3. MBP is an APR, and its deficiency is associated with the common opsonic defect and susceptibility to infections and atopic constitution. The high concentration of MBP in infants may best be explained by exposure to novel environmental antigens in early childhood, which suggests a protective role for MBP during the period of immaturity of the immunosystem. In older children, the high level of MBP can probably be explained by childhood infections and the ensuing need of MBP.\[[@CIT56]\] Role of phycolins and collectins as acute phase proteins {#sec2-10} -------------------------------------------------------- When infection exceeds the capacity of the local cells and mediators for containment and/or elimination of an organism in a tissue site, a systemic host response can ensue. This response involves release of numerous APPs from the liver in response to pathogen products (e.g., endotoxins) and cytokines (e.g., IL-1, TNF-α, and IL-6 generated locally and systemically). The liver produces complement, collectins, and pentraxins together with numerous other classes of molecules involved in host defense, inflammation, clotting, cardiovascular function, and so forth. Probably because of the presence of repeated and severe infections, chronic obstructive pulmonary disorder (COPD) is characterized by the elevations of APPs, including CRP.\[[@CIT57][@CIT58]\] Systemically, these molecules may contribute to disease, because they can have inflammatory actions caused by activation of leukocytes and activation of complement. Locally, however, the antimicrobial effects of opsonins are likely to be protective. There is a growing realization that local cells in the airways can produce collectins and APPs, including complement proteins and pentraxins.\[[@CIT59]--[@CIT63]\] Components of this local APR appear to be induced by cytokines and TLR ligands.\[[@CIT59]\] Pentraxins recognize a wide range of exogenous pathogenic substances and altered self-molecules and in species-specific manner behave as APPs. A recent study showed that CRP is highly expressed by airway epithelium and that CRP in sputum and nasal lavage fluid is capable of killing bacteria.\[[@CIT63]\] Future studies are needed to determine the relative importance of local and systemic APRs in host defense in the airways. A newly recognized family of molecules, the intelectins, has been identified and may play a role similar to that of pentraxins and collectins in the airways.\[[@CIT64]\] Transferrins {#sec2-11} ------------ Transferrin is a blood plasma protein for iron ion delivery. Transferring is a glycoprotein, which binds iron very tightly but reversibly. Transferrin is also associated with the innate immune. Transferrin is found in the mucosa and binds iron, thus creating an environment low in free iron, where few bacteria are able to survive. The levels of transferrin decrease in inflammation, seeming contradictory to its function. A decrease in the amount of transferrin would result in hemosiderin in the liver. Transferring has a bactericidal effect on bacteria, in that it makes Fe^3+^ unavailable to the bacteria. A transferrinemia is characterized by anemia and hemosiderosis in heart and liver. The iron damage to the heart can lead to heart failure. The anemia is typically microcytic and hypochromic (red blood cells are abnormally small and pale). APPs in Veterinary Diagnosis {#sec1-4} ============================ Bovine respiratory syncytial virus {#sec2-12} ---------------------------------- The sperm concentrations reached for SAA and haptoglobin during the BRSV-induced APR were generally the same or higher than bacterial infections in calves. The magnitude and the duration of the haptoglobin response was found to correlate well with the severity of clinical signs (fever) and with the extent of lung consolidation while SAA responded most rapidly to infection.\[[@CIT65]\] Prostate cancer penitents with bone lesions {#sec2-13} ------------------------------------------- Prostate cancer has a propensity to metastasize to the bone. Correctly, there are no curative treatments for this stage of the disease. Sensitive biomarkers that can be monitored in the blood to indicate the presence or development of bone metastases and/or response to therapies are lacking. The cluster of unique proteins in the sera of patients with bone metastases was identified as isoforms of SAA.\[[@CIT65]\] Hematological and neoplastic diseases of the dog {#sec2-14} ------------------------------------------------ Serum concentrations of APPs, Hp, Cp, SAA, and CRP were determined in healthy dog and dogs with different diseases, grouped as acute inflammation, hematological neoplasias \[hemotologic tumor (HT)\], including epithelial, mesenchymal, and mixed and autoimmune hemolytic anemia. Measurement of APPs may be helpful to assess clinical evolution and monitor treatment of these processes.\[[@CIT66]\] Growing calves suffering from bronchopneumonia under filed conditions {#sec2-15} --------------------------------------------------------------------- Blood samples were taken from calves with respiratory disease the first day of examination for determination of the serum concentration of haptoglobin, fibrinogen, α-2- and γ-globulins, and albumin. The two serum proteins haptoglobin and fibrinogen, and especially haptoglobin, were useful for the identification of calves requiring an anti-inflammatory treatment.\[[@CIT67]\] Multiple myeloma {#sec2-16} ---------------- Long-lasting APR occurs in patients with chronic inflammation and cancer. IL-6 was negatively correlated with five out of nine (C1-INH, C8, C9, AGP and haptoglobin) positive APPs, but positively correlated with CRP.\[[@CIT68]\] Endotoxin mastitis {#sec2-17} ------------------ A crossover study was conducted to investigate the effect of intramammarily infused lipopolysaccharide (LPS) on the APR in early (EL) and in late (LL) lactation. Nine cows received intramammary injections of 100 *µ*g of *Escherichia coli* LPS during EL and LL. The milk TNF-α is on average higher in EL. SAA concentration was not correlated being on average higher in EL. SAA concentration was not correlated with changes in milk appearance.\[[@CIT69]\] Mastitis {#sec2-18} -------- In a well-managed dairy herd, in addition to clinical mastitis, subclinical mastitis should be efficiently detected. The most promising parameters for monitoring subclinical mastitis are milk *N*-acetyl-[D]{.smallcaps}-glucosaminidase activity, lactose and electrical conductivity, along with some other indicators such as optical and milk flow measurements, preferably with an interquarter evaluation included in the test. APPs, Hp, and SAA are also potential candidates for mastitis monitoring.\[[@CIT70]\] The concentration of Hp in serum has been shown to dramatically increase in cows with experimental and spontaneous coliforms mastitis. The first APPs measured from milk and used as indicators of inflammation are bovine serum albumin and α-1 trypsin inhibitor. Hp and SAA were measured from milk and serum, and compared as tests to detect intramammary infection (IMI). A significant correlation was found between the concentrations of Hp in the serum and milk, but the concentrations of SAA in the serum and milk were not related. No correlation was found between Hp and SAA levels in milk. SAA could distinguish between mild and moderate mastitis. Using a threshold value of 0.02 mg/mL for milk Hp and 0.55g/mL for milk SAA, both tests has a high specificity (100%) with no false positive results, and a reasonable sensitivity\[[@CIT71][@CIT72]\] for the diagnosis of mastitis. Hp and SAA concentrations below the detection limit were considered as good indicators of healthy udder quarters. A substantial variation in Hp and SAA concentrations in milk was observed in udder quarters with chronic subclinical mastitis.\[[@CIT73]\] The CRP is not regarded as an APP in cattle, but has been tested as an indicator for mastitis. The concentration of CRP was shown to increase in bovine milk during mastitis. The capacity of milk CRP to distinguish between healthy and mastitic quarters was found to be poor, and the correlation between the concentration of the CRP in milk and somatic cell count (SCC) was low (*r*=0.32). It seems that the CRP does not have the best potential to be used in the detection of mastitis.\[[@CIT70]\] SAA and Hp for the detection of bovine mastitis clinical and subclinical mastitis can be revealed by high serum concentrations of Hp and SAA. It is also of interest that the concentration of APPs in milk from infected quarters is higher than that in noninfected quarters. By testing milk, a large number of samples are easily obtained in a way that is less stressful than obtaining a blood sample. If APPs are produced locally in the udder as a response to mastitis, they might be more rapid and sensitive markers of acute inflammation than the somatic cell count. However, future studies on the applicability of APP in milk as markers of mastitis are needed. *Streptococcus suis* infection in the pig {#sec2-19} ----------------------------------------- In order to measure serum transthyretin (TTR) in the pig during an APR, an assay was developed using anti-human TTR antibodies which crossreacted with porcine TTR. Following *Streptococcus suis* type-2 infection TTR showed a negative APR with serum concentrations reaching a significantly lower level at 2 days following infection.\[[@CIT74]\] Starvation {#sec1-5} ========== Negative reacting proteins are normally present in healthy animals, but will decrease in concentration due to the APR. Albumin is generally accepted as negative APP present in most species. The negative reacting protein transferrin is possibly involved in the innate immunity, perhaps by sequestering ferric ions to prevent pathogens and parasites from using nutrients. Retinol-binding protein (RBP) is a small-molecular-weight protein which is the exclusive protein for the transport of vitamin A (retinol) in the body. The synthesis and secretion of RBP by parenchymal hepatocytes is mainly controlled by the combination with the larger, tetramer protein, transthretin. The complex formation appears to be necessary to prevent extensive loss of the low-molecular-weight RBP through glomerular filtration.\[[@CIT75]\] During starvation, there is no full positive response, and a general depression of hepatic protein synthesis occurs. Malnutrition and the anorectic effects of pro-inflammatory cytokines in the brain result in a negatively changed hepatic synthesis. The major three of these cytokines (TNF-α, IL-1, and IL-6) have a profound behavioral, neuroendocrine, and metabolic effect.\[[@CIT76]--[@CIT79]\] Moreover, there is evidence that cytokines and their cognate receptors are present in the neuroendocrine system and brain. In laboratory animal species, IL-1, IL-6, and TNF-α have been found to modulate intermediary metabolism of carbohydrate, fat, and protein substrates, regulate hypothalamic--pituitary outflow, and act in the brain to reduce food intake.\[[@CIT76][@CIT78]\] On starvation and negative energy balance associated with most diseases, muscle proteins are catabolized for amino acid supply of the hepatic APP formation and as source of energy. Especially for those APPs which rapidly and quantitatively increase in blood, their formation may have amino acid impact. An increased hindquarter protein catabolism exceeding the hepatic protein synthesis, and efflux of glutamine and alanine from the hindquarter was measured during a porcine-induced endotoxemia study.\[[@CIT80]\] For growth during and after recovery from a disease, food requirements for amino acids thus may differ from the formula in ordinary food. Some pig studies indicate positive influences of additional dietary tryptophan\[[@CIT81]\] or [L]{.smallcaps}-arginine.\[[@CIT82]\] Lymphatic Neoplasia {#sec1-6} =================== Median CRP concentration was increased in all groups with neoplastic lymphatic disorders like lymphomas, malignant lymphoma, and multiple myeloma. Hp level was specially increased in dogs with acute lymphoblastic leukemia (ALL) and malignant lymphoma. The median values in the dogs with ALL were significantly higher than in dogs with other neoplastic lymphatic disorders.\[[@CIT83]\] APPs in dogs and cats: Current knowledge and future perspectives {#sec2-20} ---------------------------------------------------------------- The APR and clinical application of monitoring APPs in dogs and cats include proper and adequate clinical interpretation. In addition, the diagnostic use of APPs and their possible application in monitoring treatment can be considered as one of the most interesting and promising practical applications of these proteins. New and cheaper automated assays for determination of the main APPs in small animals will contribute to a wider use of these proteins as biomarkers of infection and inflammatory lesions.\[[@CIT80]\] Some application of acute phase proteins as diagnostic tool in animals {#sec2-21} ---------------------------------------------------------------------- APP is applied as unspecific markers of clinical and subclinical infections, to discriminate between acute and chronic disease and for prognostic purposes, since the duration and magnitude of the response reflect the severity of the disease and the effect of treatment. Haptoglobin: A marker of herd health status in pigs {#sec2-22} --------------------------------------------------- Canadian and American researchers showed that in immunologically naive boars moved to new facilities, an increase in Hp concentration was observed before the clinical signs of the disease were evident.\[[@CIT84]\] The Hp concentration remained high, and the animals subsequently showed clinical signs of the disease (depression, respiratory distress, and cyanosis). Higher Hp serum concentration prior to the clinical signs could be due to early, subclinical pathological conditions. Also, lower gaining pigs were found to have higher Hp levels than gaining pigs. The serum Hp concentration increased significantly with age in conventional slaughter pigs without clinical signs but not in slaughter pigs from high health (SPF-[X]{.smallcaps}) herds, indicating that subclinical disease in conventional herds may be the cause of higher serum Hp concentration in older pigs. Therefore, Hp seems to be a promising marker of health status by reflecting a broad spectrum of ongoing clinical as well as subclinical diseases. Serum amyloid A as a prognostic marker in equine respiratory disease {#sec2-23} -------------------------------------------------------------------- SAA is useful in the management of bacterial and viral infections in horses by large-scale monitoring in stables and as the prognostic tool in relation to clinical severity and the recovery of individual horses. Serum concentrations of SAA have been found to increase in foals during infection with *Rhodocoocus equi*, equine influenza serotype A2 (H3N8), equine herpes virus serotype 1, and *Streptococcus equi*. A statistically significant association between SAA serum concentration and severity of clinical signs of respiratory disease as well as rectal temperature has been observed. However, future studies of APP in equine medicine are needed before the applicability can be assessed.\[[@CIT85]\] There is an increase in APPs, particularly Hp, SAA, in chronic respiratory diseases in calves, and their evaluation could be useful in the determination of prognosis in sick calves.\[[@CIT86]\] Acute Phase Index {#sec1-7} ================= When APPs are used to assess nonhealthy animals versus healthy ones, values of single reactants are often not sensitive enough to detect a special patient or subject in a population of livestock. However, the acute phase signal obtained for an individual animal can be enhanced when the values of positive APP (rapid and slow) are combined with those of rapid and slow negative AP in an index. In starvation especially a decrease in the reactants may be expected. $${Nutritional}\ {and}\ {acute}\ {phase}\ {index}\ \left( {NAPI} \right)\ = \ \frac{Value\ of\ a\ rapid\ positive\ APP\ \times \ Value\ of\ a\ slow\ positive\ APP}{Value\ of\ a\ rapid\ negative\ APP\ \times \ Value\ of\ a\ slow\ negative\ APP}\ $$ The index has been used as prognostic inflammatory and nutritional index (PINI) for human patients and as acute phase index (API) for cattle. NAPI enhances sensitivity and specificity of the APP to detect nonhealthy subjects in a population of normal animals. Determination of APP can help to monitor herd and individual health, especially when several acute phase variables are combined in an NAPI.\[[@CIT87]\] Technology to quantitatively measure proteins {#sec2-24} --------------------------------------------- Radioimmunoassay (RIA) and ELISA used for APP measurement in particular of CRP are developing methods for rapid measurements of APP values. Turbidimetry is developed for APP in the dog, horse, and for the cat. Two-dimensional electrophoresis with mass spectrometry has been shown to be applicable to animal samples with the aim to measure APRs. A protein chip has been developed for the measurement of Hp and SAA in human patients. Protein microarray methodology on slides has been proposed for APP in pigs. Preliminary experiments with a monoclonal antiporcine CRP and pig acute-phase sera using methodology as described offered the possibility to measure more than 1000 pig blood sample spots on a single slide. Indirectly, APP formation may be measured in biopsies by methods to assess upregulation of protein synthesis \[quantitative polymerase chain reaction (PCR)\]. Especially the technique may be applied on samples after slaughter, or in histopathology and together with the assessment of cytokines. These technological developments may have crucial importance in the future if done rapidly, and at low costs, many samples can be handled, the APPs have a good future in diagnostics. This technique is for general assessment just as the erythrocyte sedimentation rate is used in internal medicine, but more sensitive and for special groups of patients such as hoses after castration or laprotomy.\[[@CIT87]\] Conclusion {#sec1-8} ========== Determination of animal health is important. APPs may provide an alternative means of monitoring animal health. An increased focus on the application of APP for this purpose has recently been developed. Due to a relatively short life in serum and high response in diseased animals, APP serum responses constitute a valid measure of a systemic response in diseased animals; APP serum responses constitute a valid measure of a systemic response to an initiating stimulus at the time of blood sampling. Like rectal temperature, APP levels are not suitable for establishing a specific diagnosis but can provide information about the extent of ongoing lesions in individual animals. At the herd level, APP might be useful for determining from where the disease is spreading by providing information about the prevalence of ongoing clinical and subclinical infections indicated by the high serum concentration of selected APP and by serving as the prognostic tool, with the magnitude and duration of the APR reflecting the severity of infection. Important points to consider before using APP as markers of animal health are the possible influence of environmental factors, handling, and other types of stress in the absence of disease. APPs have their possible use as markers of domestic animal health alone or at the herd level, for the detection and as a prognostic marker of different diseases or infections. However, an international standardization of APP assays is needed before they can be applied for the systematic health monitoring in veterinary medicine. **Source of Support:** Nil **Conflict of Interest:** None declared.

PubMed Central

2024-06-05T04:04:19.933472

2011-01-01

PMC3053510

Exposure of humans to excess nitrosating species from the diet, cosmetics, pharmaceuticals, the environment or from overproduction of endogenous nitric oxide (NO) is commonly linked to mutagenic and carcinogenic events.\[[@CIT1][@CIT2]\] The reactive nitrogen species (RNS), which comprise of nitric oxide, dinitrogen trioxide, nitrite, nitroxyl and nitrosonium ions, and nitrous acid, are interconnected through a series of reactions that can be admitted to start either with the generation of NO from arginine by nitric oxide synthase or with the transformation of nitrite ions into nitrous acid in the acidic environment of the stomach.\[[@CIT1][@CIT3]\] It was recently proven that dinitrogen trioxide has a biological relevance, as it originates not only from the reaction of nitric oxide with oxygen, but can also be formed by the reaction of nitric oxide with a superoxide.\[[@CIT4]\] Nitrosative deamination of the DNA bases, involving either a direct reaction with RNS or a nitroso group transfer from NO donors such as *N*-nitrosamides and *N*-nitrosamines,\[[@CIT5]--[@CIT8]\] can lead to mutagenesis through misincorporation by DNA polymerase, misrepair, or no repair of the resulting deamination products, which may result in the formation of abasic sites. The diazonium ions thus formed in the bases with aminic nitrogen may undergo nucleophilic substitution by water, leading to the transformation of cytosine, guanine, and adenine in uracil, xanthine, and hypoxanthine, respectively.\[[@CIT5]--[@CIT7][@CIT9][@CIT10]\] Published results point to guanine being more reactive than adenine toward sodium nitrite in acidic medium and to cytosine not reacting at all.\[[@CIT11]\] Moreover, *N*-nitrosamines, upon metabolic activation by cytochrome P450, form DNA oxidizing and alkylating agents.\[[@CIT12]\] Phenolic compounds are known to act as natural antioxidants and antinitrosating agents. A regular intake of (poly)phenolic compounds widely found in fruits, vegetables, tea, and red wine is believed to decrease the incidence of certain forms of cancer, and for that reason they are commonly regarded as chemopreventive agents.\[[@CIT1][@CIT13]\] The antioxidant properties of phenols are determined by their radical scavenging ability and consequent inhibitory action on lipid peroxidation under oxidative stress conditions, which correlate with their substitution pattern.\[[@CIT14]\] The antinitrosating activity of phenols is thought to be due to their action as RNS scavengers, thus preventing both direct nitrosation of DNA bases\[[@CIT11]\] and endogenous formation of carcinogenic *N*-nitroso compounds,\[[@CIT15][@CIT16]\] by reacting with the above-mentioned species more rapidly than most amino compounds. However, along with this long-known inhibitory effect on the formation of N-nitrosamines, phenols and some polyphenols have been reported to act as catalysts rather than inhibitors of such reactions.\[[@CIT17]\] A mechanism is proposed, in which the quinone monoxime tautomer of a nitrosophenol reacts with nitrous acid to produce an intermediate nitrosating agent, which undergoes an attack by the amine to produce *N*-nitrosamine and regenerate the catalyst,\[[@CIT18]\] although it has been stressed that catalytic activity is only observed when the nitrosating agent concentration significantly exceeds the concentration of phenol, which is rare *in vivo* or in environmentally significant situations.\[[@CIT19]\] It has also been observed that the introduction of a nitroso group in phenols brings about a loss in their antioxidant activity.\[[@CIT14]\] The purpose of this study is to acquire a deeper understanding of the mechanism of DNA base transnitrosation by NO donors, and on the possible action of phenols in avoiding that reaction, by comparing the relative rates of the nitroso group transfer to both classes of nucleophiles, using *N*-methyl-*N*-nitroso-4-metilbenzenesulfonamide (MeNMBS) as model compound. This and other *N*-methyl-*N*-nitrosobenzenesulfonamides have been used before as nitrosating agents in the study of the transnitrosation of amines\[[@CIT20]\] and phenols.\[[@CIT21]\] In addition, we expect to collect more evidence on the mechanism of nitrosophenol formation, as there has been some discussion on whether the reaction occurs directly at the carbon atom\[[@CIT19][@CIT22][@CIT23]\] or the previous formation of an *O*-nitroso intermediate is involved.\[[@CIT21][@CIT24]\] The nitrosation of OH-blocked 'phenols,' such as anisole, is possible when the mechanism involves the nitrosonium ion polar reaction,\[[@CIT25]\] but it will not be observed when the reaction follows the oxidative pathway via the phenoxy radicals, because the formation of those are hindered, as observed by Daiber *et al*.\[[@CIT26]\] We must also remark that the team of Luis García-Rio was developed a rather elegant methodology to distinguish reactivity in ambident nucleophiles. In their case they have been able to determine that for the nitrosation of enols, the reaction follows two parallel pathways, one involving C-nitrosation and the other O-nitrosation.\[[@CIT27]\] Materials and Methods {#sec1-1} ===================== Reagents and equipment {#sec2-1} ---------------------- Reagent grade guanine, cytosine, adenine, uracil, hypoxanthine, guaicol (2-methoxyphenol), 3-methoxyphenol, 2-bromophenol, 3-bromophenol, 2,3-dimethoxyphenol, 2-chlorophenol, 2-fluorophenol and 2,6-di-*tert*-butylphenol, sulfanilamide and *N*-1-naphtylethylenediamine from Aldrich, xanthine from Sigma, 4-methoxyphenol, 4-chlorophenol and *N*-methyl-*N*-nitroso-4-metilbenzenesulfonamide (MeNMBS) from Merck, carvacrol (2-methyl-5-*iso*propylphenol), 4-bromophenol, 3,5-dimethoxyphenol and syringol (2,6-dimethoxyphenol) from Fluka, 3,5-di-*tert*-butylphenol from Enamine, *N*-methyl-4-metilbenzenesulfonamide (MeMBS) from Ega-Chemie were used; thymol (5-methyl-2-*iso*propylphenol) and anisole were from Riedel-de-Haën; the latter was previously distilled and the second fraction (41--42°C)\[[@CIT24]\] was used. HPLC grade acetonitrile and methanol were from Lab-Scan, and 1,4-dioxane from Riedel-de-Haën and bidistilled water from a quartz bidistiller. An UV-Vis spectrophotometer (Varian Cary 50 Bio), an HPLC (Agilent 1100) with RP-18 LiChrospher column (250 × 4 mm, 5 *µ*m), and diode array UV-Vis detector, a thermostated bath (Julabo F12), with ± 0.1°C precision and a pH meter (Thermo Orion 4 Star pH-ISE Benchtop) were used. In the kinetic studies, the experimental points were adjusted to equations using GraFit^®^ 5.0.5. (Erithacus Software Limited). Solutions {#sec2-2} --------- The following solutions were prepared: phosphate buffer 50 mM, pH 3 and 4, and 0.1 M, pH 9.5, 10, 10.25, 10.75, 11, and 11.5; guanine 1.75 × 10 ^-4^, 2.6 × 10 ^-4^, 3 × 10 ^-4^, and 3.5 × 10 ^-4^ M, in pH 3 and pH 4 phosphate buffer; cytosine 1.75 × 10 ^-4^, 2.5 × 10 ^-4^, 3.5 × 10 ^-4^, 1 × 10 ^-3^, 5 × 10 ^-3^, and 1 × 10 ^-2^ M, in pH 3 phosphate buffer; adenine 2 × 10 ^-4^, 1 × 10 ^-3^, 3 × 10 ^-3^, 5 × 10 ^-3^, 7 × 10 ^-3^, and 1 × 10 ^-2^ M, in pH 3 phosphate buffer; 0.1 and 0.3 M of each phenol in dioxane; 0.01 M of MeNMBS in dioxane; 0.01 M of 2,2,6,6-tetramethyl-1-piperidinyloxy (TEMPO) in water. Kinetic study of the acidic transnitrosation of DNA bases {#sec2-3} --------------------------------------------------------- In the study of the DNA base concentration effect on the reaction rate, the appropriate volumes of pH 3 solutions were placed in a vial in order to get the following final concentrations: guanine 1.74 × 10 ^-4^ to 3.48 × 10 ^-4^ M; cytosine 1.74 × 10 ^-4^ to 9.93 × 10 ^-3^ M; and adenine 1.99 × 10 ^-4^ to 9.93 × 10 ^-3^ M. After stabilization at 35°C for 10 minutes, MeNMBS solution was added, to a final concentration of 6.67 × 10 ^-5^ M, in a total volume of 3 mL. To determine the pH effect on the reaction between guanine and MeNMBS, the previous study was repeated at pH 4. The decay of MeNMBS concentration with time was obtained from the peak areas at 247 nm in the HPLC chromatograms of the reaction mixture. In the analysis of the reaction mixtures of guanine and cytosine, a step gradient from a sodium phosphate buffer (50 mM, pH 3), to a mixture of 40% buffer and acetonitrile was used. Retention times were: guanine 7.0 minutes, xanthine 11.2 minutes, cytosine 3.1 minutes, uracil 5.0 minutes, MeMBS 19.9 minutes, and MeNMBS 22.9 minutes. For the reactions of adenine, elution was carried by a step gradient from a mixture of 87.5% phosphate buffer (50 mM, pH 4 and 2 mM in triethylamine) with methanol to 30% buffer. Retention times were: hypoxanthine 3.7 minutes, adenine 9.3 minutes, MeMBS 15.6 minutes, and MeNMBS 17.9 minutes. Kinetic study of the basic transnitrosation of phenols {#sec2-4} ------------------------------------------------------ In the study of the phenol concentration effect on the reaction rate, appropriate volumes of solutions to get final concentrations ranging from 1 × 10 ^-3^ to 6 × 10 ^-3^ M, were pipetted to a spectrophotometer quartz cell containing pH 10.75 buffer solution. The cells were placed in the spectrophotometer and allowed to stabilize for 15 minutes at 25°C and MeNMBS solution was added to the final concentration of 1 × 10 ^-4^ M, in a total volume of 3 mL. To determine the pH effect, the study was carried in buffer solutions from pH 9.5 to 11.5. The formation of nitrosophenol was monitored by measuring the solution absorbance at the appropriate wavelength \[[Table 1](#T0001){ref-type="table"}\]. The quantification of the by-product phenol *p*-toluenesulfonate was made by HPLC analysis of the reaction mixture in a 60% acetonitrile / 40% water mixture and the product distribution was established by conjugation of this data, with quantification of the initial MeNMBS, using the Shin's method.\[[@CIT29]\] Retention times were: NO-thymol 2.7 minutes, thymol 6.7 minutes, NO-carvacrol 3.7 minutes, carvacrol 6.3 minutes, 3-methoxyphenol 2.8 minutes, 3-methoxyphenol tosylate 9.3 minutes, 3,5-dimethoxyphenol 2.9 minutes, 3,5-dimethoxyphenol tosylate 10.5 minutes, and MeMBS 3.1 minutes. ::: {#T0001 .table-wrap} Table 1 PhOH NO-Ph λ~max~ /nm p***K***~a~ lit. p***K***~a~ exp. *k*/**M-1.s-1** ------ ------------------ -------------------- ------------------ -------------------- 1a 416 8.45 \[[@CIT33]\] \- \* 1b 410 8.56 \[[@CIT33]\] \- \* 1c 404 8.70 \[[@CIT33]\] \- \* 1d 403 9.03 \[[@CIT33]\] \- \* 1e 410 9.17 \[[@CIT33]\] \- \* 1f 339 9.34 \[[@CIT33]\] 9.01 ± 0.02 (1.9 ± 0.3) × 10-3 1g 426 9.41 \[[@CIT33]\] \- \* 1h 348 9.65 \[[@CIT33]\] 10.29 ± 0.05 (2.8 ± 0.5) × 10-2 1i 464 9.97 \[[@CIT33]\] \- † 1j 392 9.98 \[[@CIT33]\] \- † 1k 437 10.10 \[[@CIT33]\] \- † 1l 428 ‌‌ \- † 1m ‡ 10.30 \[[@CIT33]\] \- 1n 390 10.38 \[[@CIT33]\] 10.58 ± 0.14 (3.0 ± 0.7) × 10-2 1o 390 10.60 \[[@CIT33]\] 10.71 ± 0.10 (3.4 ± 0.6) × 10-2 1p 376 11.70 \[[@CIT33]\] \- † \*The reaction proceeded very slowly and the data did not fit properly. A first order integrated equation; ^†^Product formation was not observed; ‡Failed to react with acidic nitrite; ^§^Impossible to solubilize in the reaction medium; ‌‌Not found in the literature. ::: Synthesis of nitrosophenols and phenyl *p* -toluenesufonates {#sec2-5} ------------------------------------------------------------ In order to determine the appropriate wavelength to follow the transnitrosation reactions, authentic samples of nitrosophenols were synthesized by acidifying a mixture of phenolate and sodium nitrite with sulfuric acid. \[[@CIT31]\] In the reaction of 3,5-di-*tert*-butylphenol no isolable product was formed. The hindered 2,6-dimethoxyphenol was nitrosated by dissolution in an ethanolic hydrochloric acid solution, to which sodium nitrite was gradually added. \[[@CIT32]\] Nitrosation of anisole was performed with sodium nitrite in a mixture of dichloromethane and trifluoroacetic acid. \[[@CIT30]\] The product had a λ~max~ at 350 nm. Phenols were tosylated with *p*-toluenesulfonyl chloride in pyridine to afford the corresponding phenyl *p*-toluenesufonate to be used as standards in HPLC analysis of the reaction mixtures. \[[@CIT34]\] Kinetic study of the basic transnitrosation of phenols in the presence of a radical trap (TEMPO) {#sec2-6} ------------------------------------------------------------------------------------------------ Volumes of 50 *µ*L of 0.3 M solution of 3-metho × yphenol, 90 *µ*L of TEMPO solution, 70 *µ*L of 1,4-dio × ane, and 2760 *µ*L of pH 10.75 phosphate buffer solution were pipetted to a quartz cell. After temperature stabilization, 30 *µ*L of MeNMBS solution was added and the formation of 4-nitrosophenol monitored. A control experiment was set at the same time and the radical trap solution replaced by the same volume of buffer, which amounted to 2850 *µ*L. The same tests were made with thymol. Results and Discussion {#sec1-2} ====================== The transnitrosation of DNA bases \[[Figure 1](#F0001){ref-type="fig"}\] was studied in acidic medium due to their low solubility at higher pH values, guanine being the least soluble of them, which limited the range of concentrations that could be used. At low pH, *N*-Nitrosobezenesulfonamides undergo hydrolysis, so a competition between both processes is expected. The observed rate constant for the consumption of MeNMBS should be the sum of the two contributions: transnitrosation (k~T~ ) and hydrolysis (k~H~ ^+^ ), according to eq. 1. ::: {#F0001 .fig} Figure 1 ::: {.caption} ###### Transnitrosation step in the nitrosative deamination of guanine :::  ::: $$k_{obs}\ = \ k_{T}\ \left\lbrack {DNA\ base} \right\rbrack\ + \ k_{H +}\ \left\lbrack H^{+} \right\rbrack\ $$ Although UV / Vis spectroscopy is broadly used to monitor the progress in organic reactions, mainly due to spectral similarities of the involved species, HPLC with diode array detection proved to be a more reliable quantitative methodology. Reaction of MeNMBS (66.7 *µ*M) with excess guanine (174 to 348 *µ*M), corresponding to *pseudo* first-order conditions, was carried out in dioxane/water (\~1 : 150), at 35°C and a pH 3 phosphate buffer. The decay of the former was monitored by HPLC analysis of the reaction mixture. The corresponding peak areas in the chromatograms were adjusted to a first-order integrated equation for each guanine concentration and the corresponding value of k~obs~ calculated. The plot of k~obs~ against the guanine concentration proved the reaction to be independent of DNA base concentration, which meant that MeNMBS was being hydrolyzed and no transnitrosation occurred, at least to a measurable extent. The estimated value of k~H~^+^ compares to the value determined in our laboratory following the hydrolysis of MeNMBS at the same temperature by UV/Vis spectroscopy, 5.5×10 ^-2^ M ^-1^ s ^-1^ . At pH 4, although hydrolysis was slower, it still prevailed over transnitrosation. The same procedure was applied to cytosine and adenine, but because of their higher solubilities, wider ranges of concentrations could be used: 174 *µ*M to 9.93 mM for the former and 199 *µ*M to 9.93 mM for the latter. In both cases results were the same as those obtained with guanine. In the reaction of phenols with nitrosating agents, the major product is the corresponding 4-nitrosophenol 2, the most stable isomer, due to a tautomeric equilibrium with the quinone monoxime form. A small percentage of the 2-nitroso compound may also form, except when position 4 holds a substituent, in which case the latter is formed exclusively. When the phenol bears a methoxy substituent in position 3, and because of electronic factors, the opposite distribution of products is observed.\[[@CIT31]\] It has been discussed whether the reaction occurs directly at the carbon atom or if the previous formation of an *O*-nitroso intermediate is involved. Casado *et al*,\[[@CIT19][@CIT22][@CIT23]\] have shown unambiguously that the reaction with nitrous acid occurs by the first of these mechanisms. However, other nitrosating agents, like nitrososulfonamides\[[@CIT21]\] and alkyl nitrites,\[[@CIT24]\] in basic media, present a different behavior. In both cases, the possibility of the formation of an unstable aryl nitrite, which, upon homolysis followed by recombination, afforded the C-nitroso product, was invoked. A similar combination between nitroso and pheno × y radicals was observed by EPR.\[[@CIT32]\] In order to further enlighten the reaction mechanism, kinetic experiments involving the transnitrosation of a series of substituted phenols 1 by *N*-Methyl-*N*-Nitroso-4-metilbenzenesulfonamide (MeNMBS) were run at 25°C in 5% dioxane / water in the pH range 9.5 -- 11.5. The ambident character of the electrophile leads, in some cases, to the occurrence of the phenol *p*-toluenesulfonate 3 along with 2\[[Figure 2](#F0002){ref-type="fig"}\]. In such cases, the product distribution was established by the conjugation of HPLC analysis of the reaction mixture with quantification of the initial MeNMBS, using Shinn's method.\[[@CIT29]\] The obtained results are summarized in [Table 1](#T0001){ref-type="table"}. ::: {#F0002 .fig} Figure 2 ::: {.caption} ###### Nitroso group transfer from MeNMBS to substituted phenols :::  ::: Nitrosation of chloro- and bromophenols was reported to be negligible.\[[@CIT19]\] In fact, the halogenated phenols studied (1a-e and 1g) reacted very slowly with MeNMBS and the absorbance-time data deviated significantly from the best-fitting first order integrated equation. Kinetic constants estimated from the available data were in the order of 10 ^-4^ M ^-1^ .s ^-1^, which was the magnitude of the basic hydrolysis constant of MeNMBS determined in our laboratory, so a competition between the two reactions should be expected. In the case of phenols with methoxy substituents in positions 2, 4, and / or 6 (1i-l), the corresponding nitrosophenol (2i-l) was not detected. A probable cause would be a stabilization effect of the intermediate radical, formed by homolysis of the initial aryl nitrite, preventing the recombination step \[[Figure 3](#F0003){ref-type="fig"}\]. Another possible explanation was the fact that methoxy substituents in these positions put a negative charge by resonance in the carbon bearing the hydroxyl group, thus destabilizing the phenoxide ion, although that effectis not very patent in the order of pK~a~ values. ::: {#F0003 .fig} Figure 3 ::: {.caption} ###### Possible stabilization of the pheno × yl radical by the metho × y groups :::  ::: As for 3,5- and 2,6-*t*-buthylphenols (1m and 1p), the former did not react due to steric hindrance and the latter was not soluble in the medium basicity media. A pH of 12.5 was needed to achieve complete solubilization, at which point hydrolysis of MNTS was much faster than transnitrosation. The fact that 2,6-*t*-buthylphenol was nitrosated by nitrous acid, but the 3,5-isomer was not, clearly emphasizes the difference in mechanism operating in acidic nitrosation and basic transnitrosation. 3,5-dimetho × y- and 3-metho × yphenols as well as carvacrol and thymol (1f, 1h, 1n, and 1o) exhibited fittable absorbance-time data and measurable rate constants. In the last two compounds, nitrosophenol (2n and 2o) was the only product, while in the first two formations, *p*-toluenesulfonate (3f and 3h) was also observed. In a basic medium \[[Figure 4](#F0004){ref-type="fig"}\], MeNMBS may get hydrolyzed (k~OH~ ^-^ ), transfer its nitroso group to the phenolate ion (k~N~ ) or undergo attack by the former in its sulfonyl group (k~S~ ): ::: {#F0004 .fig} Figure 4 ::: {.caption} ###### None :::  ::: The observed rate constant for the formation of nitrosophenols should account for all the processes (eq 2): $$\begin{array}{l} {k_{obs}\ = \frac{k_{Nu}K_{a}}{\left\lbrack H^{+} \right\rbrack + K_{a}}\left\lbrack {PhOH} \right\rbrack + k_{OH^{-}}\left\lbrack {OH^{-}} \right\rbrack} \\ \\ {{with}\ k_{Nu}\ = \ k_{N}\ + \ k_{S}.} \\ \end{array}$$ The plot of the *pseudo*-first-order rate constant against the concentration of phenol showed that the reaction was first-order in the former. The existence of an intercept was indicative of the occurrence of hydrolysis \[[Figure 5 (a)](#F0005){ref-type="fig"}\]. The influence of the medium pH on the reaction rate was accounted for in the first term of the previous equation. A linear plot of such term, according to eq. 3, showed a decrease in the rate constant with increasing acidity, meaning that the phenolate ion was the active nucleophile. From the slope and intercept, values of pK ~a~ consistent with those reported in the literature were found \[[Figure 5 (b)](#F0005){ref-type="fig"}\] and [Table 1](#T0001){ref-type="table"}. ::: {#F0005 .fig} Figure 5 ::: {.caption} ###### Reaction of 3,5-dimetho × yphenol with MeNMBS in 5% dio × ane / water \[MNTS\] = 1 × 10-4 M, 25°C. (a) First-order plot relative to phenol concentration, pH = 11.38; (b) Influence of acidity on the *pseudo*-first-order rate constant, \[PhOH\] = 5 × 10^-3^ M :::  ::: $$\frac{1}{k_{obs}}\ = \frac{1}{K_{Nu}\left\lbrack {PhOH} \right\rbrack} + \frac{1}{K_{Nu}K_{a}\left\lbrack {PhOH} \right\rbrack}\left\lbrack H^{+} \right\rbrack$$ The existence of a linear relation between k~N~ and the pK~a~ of those phenols and four others, whose reaction with MeNMBS was reported\[[@CIT21]\] is patent in [Figure 6](#F0006){ref-type="fig"}. The 3-MeO derivative was out of the correlation, as observed by Leis *et al*.,\[[@CIT24]\] in the case of nitrosation by alkyl nitrites. The obtained value for β~nucl.~ (1.4 ± 0.2), and the *P* value (-1.7 ± 0.2) obtained from the Hammett plot \[[Figure 7](#F0007){ref-type="fig"}\] showed that all the reactions of monophenols occurred by the same mechanism, involving the development of a positive charge at the transition state. The fact that no reaction occurred between MeNMBS and anisole (metho × ybenzene) under the same conditions, strongly suggested the involvement of the oxygen atom, with the formation of an aryl nitrite, followed by a Fisher-Hepp-like rearrangement. However, when the reaction was run in the presence of TEMPO, a radical trap, no difference in the values of k~obs~ was noticed, indicating that either radical species were not involved in the *O*-nitroso to C-nitrosophenol rearrangement or that this occurred after the rate limiting step. The choice of TEMPO as a radical trap could be tricky, as the parent compound TEMPOL could show catalytic activity in the nitrosation of phenols, namely, in reactions with peroxynitritrite.\[[@CIT34]\] ::: {#F0006 .fig} Figure 6 ::: {.caption} ###### Brønsted plot for the reactions of phenolates with MeMNTS nitroso group (Insert: Brønsted relationship for monophenols) :::  ::: ::: {#F0007 .fig} Figure 7 ::: {.caption} ###### Hammett plot for the reactions of phenolates with the MeMNTS nitroso group :::  ::: Nevertheless, absence of the kinetic influence of the radical trap supported our previous remark. In fact the possibility of reaction via the nucleophilic carbon cannot be ruled out, and this subject should be considered in further studies. **Source of Support:** Nil **Conflict of Interest:** None declared

PubMed Central

2024-06-05T04:04:19.937999

2011-01-01

PMC3053511

The slow but unremitting growth of pediatric audiology over the last five decades has culminated in the actuality of delivering services to the youngest and most vulnerable population. This makes preventive audiology a practicable and fundamental foray in current times. Early detection and intervention for hearing impaired infants has become a progressively imperative aspect of neonatal and infant care and has magnified the audiological scope of practice significantly, as a form of secondary prevention.\[[@CIT1]\] This change in scope of practice has produced a host of new challenges in the delivery of effective and reliable hearing care services to newborns and young infants. This has also resulted in large scale research initiatives that address the rising surge of questions regarding the improvement of methodologies for identification and intervention of hearing loss, most especially in developing countries.\[[@CIT2][@CIT3]\] One of the questions that the current study aims to explore is whether epidural anesthesia has any effect on the hearing screening results in newborns. The process of identifying the section of population at the highest risk of hearing loss is an essential component of audiological practice and serves as the initial step toward delivering effective audiological services to the pediatric population.\[[@CIT4]\] The screening of children and infants for hearing loss is a steadily advancing process that has accelerated significantly over the past few years. Even though a number of different methods of detecting hearing loss were evaluated earlier, it was only during the 1990s that substantial progress was made in lowering the average age at which significant hearing loss was identified.\[[@CIT5]\] This delayed identification of hearing loss was predominantly due to a lack of methodical screening programs and the shortcomings of subjective behavioral screening methods. Fortunately, the emergence of a more accurate, noninvasive, and rapid means of screening hearing loss as well as more concerted and rigorous efforts by professional bodies, such as the Health Professions Council of South Africa,\[[@CIT6]\] has resulted in the prospect of screening prior to hospital discharge being urged and supported. As hearing loss cannot be promptly and effortlessly identifiable by routine clinical examinations such as behavioral observation,\[[@CIT7]\] screening with more objective electrophysiological measures such as otoacoustic emissions and auditory brainstem responses is advocated for the universal screening of all newborns and infants.\[[@CIT5][@CIT6]\] Universal new-born hearing screening programs are aimed at obtaining hearing screening results for every new-born at any time prior to discharge from the hospital.\[[@CIT8]\] This is believed to possibly be the major way forward to guarantee that early identification and early access to services, including amplification and individualized family-centered early communication intervention, ensues.\[[@CIT9]\] Early hearing detection through universal newborn screening has taken on exceptional reputation as being the best practice in child healthcare in the developed world;\[[@CIT10]\] and this highlights the value of ensuring that the methodologies employed are effective and accurate; and yield minimal false positive results in a time frame that is as early as possible.Therefore, it is vital to isolate and categorize factors that may influence the success or failure of new-born audiological screening programs, such as epidural anesthesia, in neonates born to mothers who undergo elective Cesarean section. False positive results are obtained when a condition is not present, but the test results indicate that it is present. For example, Owen *et al*,\[[@CIT11]\] document the high number of false positive results when OAEs are obtained in the first 24 hours after birth. Referral rates for OAEs are reported to be 5 -- 20% when conducted earlier than 24 hours after birth and less than 3% when conducted 24 to 48 hours after birth.\[[@CIT12]\] Differing results exist regarding the specificity of OAEs measured at various time intervals after birth. Korres,, *et al*,\[[@CIT13]\] report a high rate of screening OAE recordings after 24 hours. Research conducted by Levi *et al*,\[[@CIT14]\] indicates that OAEs can be measured reliably earlier than 48 hours after birth, while Pennsylvania Health Care Cost Containment Counsel\[[@CIT15]\] suggests that OAEs recorded after 48 hours are more reliable. Unfortunately, an early discharge from maternity wards may contribute to a significant increase in false positive results due to vernix in the ear canal.\[[@CIT16]\] Such reports underscore the need for establishing other conditions present early during the postnatal period; such as epidural anesthesia, which may increase the rate of false positive hearing screening findings. Stuart and Moretz\[[@CIT17]\] suppose that the presence of false positive results may be an important factor hindering the implementation of new-born hearing screening programs. False positive results are also assumed to incur a significant cost to screening procedures;\[[@CIT16]\] hence, the significance of the current study. According to Gorga *et al*.,\[[@CIT18]\] an efficient new-born audiological hearing screening program aims to identify newborns with a hearing loss in a cost-effective manner, and incorporates tests that have a low false positive rate. Mehl and Thompson\[[@CIT19]\] assert that the expense of hearing screening programs also need to be deliberated in terms of the cost of special education and support programs. A delay in the diagnosis of a hearing loss leads to a cost to the general public, the child's family, and the child with the hearing loss. This is thought to be particularly important considering the documented numbers of hearing impaired individuals. There is an increasing prevalence rate of hearing loss reported globally.\[[@CIT20]\] This increasing prevalence has resulted in the customary practice of universal new-born hearing screening in developed countries. Despite South Africa having a comparatively well-developed infrastructure compared to other regions in the Sub-Saharan Africa, new-born hearing screening programs are a long way from common practice. This may be due to the availability of very little contextual research on infant hearing screening. In addition, this lack of data and the increasing priority toward addressing the overwhelming burden of infectious diseases such as HIV / AIDS has raised obstacles in cultivating support, funding, and political activism for infant hearing screening.\[[@CIT20]\] Therefore, additional research on Early Hearing Detection and Intervention (EHDI) in a South African context is vital for the collation and development of appropriate and efficient neonatal hearing screening guidelines and protocols; hence, the current study. With the documented prevalence of one to four in every 1000 live births globally, hearing impairment is one of the most common congenital abnormalities in new-borns.\[[@CIT21]\] Globally, it is reported to be twice as prevalent as other neonatal conditions, screened for at birth.\[[@CIT22][@CIT23]\] The current literature suggests that globally, approximately six in every 1000 infants present with permanent hearing loss at birth or within the neonatal period.\[[@CIT24]\] This documented rise in the prevalence rate of hearing impairment worldwide correlates with that reported by the World Health Organization (WHO), which maintains that the estimate for incapacitating hearing loss has increased from 120 million to approximately 278 million in the decade between 1995 and 2005.\[[@CIT25]\] This increased prevalence is more in developing countries where it is reported that more than 90% of all infants with congenital or early-onset hearing loss reside.\[[@CIT24]\] In these developing countries, Olusanya and Newton\[[@CIT24]\] assert that environmental risks are more prevalent, and that early identification programs are exceptionally scarce. Moreover, this reported increased prevalence rate is reported to be even higher, if mild and unilateral hearing losses are also incorporated.\[[@CIT9]\] Literature has identified universal new-born hearing screening as the recommended protocol for Early Hearing Detection and Intervention (EHDI), particularly in developed countries.\[[@CIT26]\] Factors influencing the implementation of universal new-born hearing screening in South African tertiary hospitals include: practicality, ergonomics and economics (cost effectiveness), and the availability of equipment and manpower.\[[@CIT26]--[@CIT28]\] If a South African prevalence estimate of 10% is used, an estimated 4.5 million individuals are present with sensorineural hearing loss.\[[@CIT28]\] This reportedly results in each audiologist being required to serve a significantly larger number of patients when compared to their counterparts in developed countries. Most importantly, the majority of these audiologists operate within the private healthcare sector, where only a small minority of individuals who can afford these services can be seen.\[[@CIT28]\] Therefore, when matching population size with the number of qualified audiologists in South Africa, there is an apparent scarcity of manpower in the public healthcare sector.\[[@CIT28]\] This situation is far from being remedied, as formal training in the profession of audiology is also lacking in most tertiary institutions in developing countries.\[[@CIT27]\] In comparison to first world countries, the aforementioned factors may influence the ability of audiologists in South Africa to effectively implement universal new-born hearing screening on all infants, prior to hospital discharge. Therefore, the use of targeted screening in high-risk neonates, as well as a clear understanding of the influence that certain postnatal factors (such as epidural anesthesia) may have on the implementation of early hearing screening measures may be more cogent. A need exists to determine the audiological findings in neonates born to mothers who have undergone epidural anesthesia during elective Cesarean sections. It is possible that anesthesia may depress the functioning of the auditory system, as also the integrity of the hearing screening, thereby causing false positive results.\[[@CIT29]\] Epidural anesthesia is a procedure that entails the injection of a substance outside the dura mater of the spinal cord, and this causes an autonomic and partial central nervous system blockade. It is commonly used for elective Cesarean section.\[[@CIT30]\] It has the advantage of allowing the mother to remain awake, minimizes the risk of maternal aspiration, and reduces drug effects on the new-born.\[[@CIT31]\] The long-term effects of epidural anesthesia, which are rare and minimal, have been well-documented;\[[@CIT30]\] however, if or how the hearing abilities of new-born infants are affected by epidural anesthesia is not well established. Some evidence of the influence of anesthesia on new-born infant's hearing was reported as early as 1988 by Diaz *et al*,\[[@CIT32]\] examined the effects of maternal lidocaine hydrochlorideanesthesia on the brainstem auditory evoked responses (BAERs)in neonates born by Cesarean delivery. In their study, the findings indicated the effect that maternal anesthesia had on the auditory brainstem response, with a significant delay noted in the central neuralcomponent of the BAER at 90 dB for the experimental group when compared to the control group. In another study by Bozynski *et al*,[@CIT33] the mean wave I-V intervals were prolonged when testing was conducted at less than four hours when compared to findings at 48 hours or longer; and these researchers concluded that changes in the serial auditory brainstem-evoked response tests occurred after maternal lignocaine epidural anesthesia in newborn infants, and that these changes correlated with the blood lignocaine concentrations. As many studies conducted in the United Kingdom, Israel, and a majority of the developed countries have already established the feasibility of hospital-based hearing screening programs despite early postnatal hospital discharge;\[[@CIT14][@CIT34]\] studies investigating improved methodological strategies need to be prioritized before the realization of effective and efficient low-cost universal new-born hearing screening programs. This includes the appropriate timing of such programs in the postnatal period, to minimize false positive results, while ensuring early identification of neonatal hearing loss prior to hospital discharge. Materials and Methods {#sec1-1} ===================== Main aim {#sec2-1} -------- To establish if epidural anesthesia has an influence on new-born hearing screening results in newborns born via elective Cesarean section in healthy pregnancies. Specific objectives {#sec2-2} ------------------- To determine hearing screening results in a group of newborns born to mothers who had undergone epidural anesthesia during Cesarean section childbirth (experimental group);To compare hearing screening findings of the experimental group with those of newborns born to mothers who had undergone natural vaginal delivery without epidural anesthesia (comparison group);To establish if the time of the hearing screening following delivery has any effect on the screening results. Research design {#sec2-3} --------------- This study employed a prospective quasi-experimental repeated measures design with a comparison group.\[[@CIT35][@CIT36]\] This design gave the researchers an opportunity to compare the findings of the experimental group with those of the comparison group at different screening times. Participants {#sec2-4} ------------ ### Sample size {#sec3-1} The sample comprised of 40 newborns that were born at the chosen private hospital in Johannesburg. The newborns were divided into 20 born to mothers who had chosen to undergo epidural anesthesia during Cesarean section childbirth (experimental group) and 20 born to mothers who had undergone a natural delivery without the use of epidural anesthesia (comparison group). ### Sampling strategy {#sec3-2} A nonprobability sampling strategy, purposive sampling strategy was adopted. The participants were selected from a location convenient to the researchers, where the reported numbers of elective Cesarean section births were high. Participants meeting the inclusion criteria were identified through the Obstetrics and Gynecology Department at the research site and approached to participate in the study. ### Inclusion criteria {#sec3-3} Inclusion criteria for participation were: (i) mothers who had carried the newborns to their full gestational term; (ii) healthy pregnancy; (iii) maternal age less than 35; (iv) no identified risk factors, such as, a number of natural abortions, illness or condition/s that required admission to the Neonatal Intensive Care Unit, craniofacial anomalies, *in-utero* infections, family history of hearing loss; (v) Cesarean section with epidural anesthesia or normal vaginal delivery; and (vi) no epidural anesthesia during natural vaginal delivery. ### Ethical considerations {#sec3-4} Prior to the study being conducted, permission to conduct the study was secured from the Medical Human Research Ethics Committee --- Protocol number M070307. Following ethical clearance, the researchers presented the research proposal to the Research Coordinator at the research site for approval to conduct the study. The Review Board at the research site reviewed and approved the study. Thereafter, the participants were invited to volunteer to participate in the study with the assistance of the nursing staff in the maternity ward. The following ethical practices were adhered to during this study: Permission to conduct the research was obtained from the gynecological specialist and pediatrician.Informed consent was obtained in writing from the mothers of the participants before hearing screening.The participants' rights and worth were respected.The voluntary nature of participation was made clear to the participants, and they were notified of their right to withdraw from the study at any point without any negative consequences.Confidentiality was ensured through the anonymity of participants and safe storage of information during and after the completion of the research. Research codes instead of participant identifying information were used.The researchers were sensitive toward maternal anxiety and when required, in-house counseling services were approached to assist.Participants who required further diagnostic testing were provided with both private and public health sector referral details, to ensure that the newborns received early intervention. Data collection {#sec2-5} --------------- ### Materials and procedures {#sec3-5} All mothers in the study completed a questionnaire pertaining to the birth time, pregnancy, and birth and family medical history. Information obtained from this questionnaire and from the medical chart reviews allowed the researchers to ensure that the participants met the inclusion criteria and were without risk indicators for hearing loss as defined by the Health Professions Council of South Africa (HPCSA).\[[@CIT6]\] Screening measures that were recommended for newborn hearing screening were required to be physiological or objective in nature. These included transient evoked otoacoustic emissions (TEOAE), distortion product otoacoustic emissions (DPOAE), and automated auditory brainstem response (AABR).\[[@CIT27]\] TEOAEs were low intensity sounds originating from active amplification of the outer hair cells of the cochlear, whereas, DPOAEs were generated by two continuous pure tones presented simultaneously to the ear.\[[@CIT27][@CIT37]\] In contrast, the AABR was an electrical response to auditory stimuli and assessed the integrity and function of the eighth cranial nerve and auditory pathway. OAE and AABR testing modalities were considered to be complementary in nature; hence their combined use in the current study.\[[@CIT38]\] All participating newborns were screened using automated TEOAEs (through the use of GSI AUDIO screener) followed by automated ABR (through the use of Maico MB11- MAICO Diagnostic) measures, while resting quietly in open bassinets in an empty new-born nursery. Although both these measures did not require testing to be conducted in a soundproof environment, as they possessed advanced digital signal processing that reduced the effect of ambient noise; the ambient noise levels were monitored through the use of a sound level meter during data collection, to ensure that the findings were valid and reliable and were not negatively influenced by noise. Nursing and sanitary staff were informed about the importance of a quiet environment to obtain accurate results.\[[@CIT18]\] Every effort was made to minimize physiological noise, by screening newborns when they were resting or immediately after feeding.\[[@CIT39]\] For both test measures, the results were recorded as either *pass* or *refer* for each participant at each testing session. ### Data analysis {#sec3-6} *Pass / refer* criteria for the analysis of TEOAE and AABR results was adopted. Due to the reported high ambient noise levels in a hospital,\[[@CIT37][@CIT40]\] which primarily affected low frequencies, 250 Hertz (Hz) and 500 Hz were not included within the *pass* / *refer* criteria. Gorga *et al*,[@CIT41] reported noise levels to affect 1 kHz as well. Following consultation with a statistician, data were analyzed through both descriptive and inferential statistics, utilizing the statistical computer program SPSS. Inferential quantitative analyses of the audiological results were performed using inferential statistical methods, which includedthet-test andFishers Exact test due to their precision in showing relationships in sample sizes below 30.\[[@CIT42]\] The p-value (0.05) was selected to test the hypothesis with *P*\<0.05 indicating rejection of the null hypothesis. The first null hypothesis for the current study was that epidural anesthesia did not have an effect on the hearing screening results of newborns, while its alternate hypothesis was that epidural anesthesia did have an effect. The second null hypothesis was that the time of testing did not have an effect on the results, while its alternate hypothesizes an effect. ### Reliability and validity {#sec3-7} The following measures were adopted to improve the reliability and validity of this study: Ambient noise levels were minimized and monitored to ensure accuracy of the results\[[@CIT18]\] and the physiological noise was minimized by only screening the infants when they were resting or immediately after feeding. Furthermore, for TEOAE screening, frequencies below 1 kHz were excluded from the analysis due to these frequencies being most affected by acoustic ambient noise, and external and internal artifacts.The audiological equipment used during the study had undergone the annual calibration prior to data collection, as per the manufacturer's guidelines, and the biological calibration of the instruments was also performed prior to each testing session. Test administration and control of patient variables was consistent throughout the data collection time, to ensure reliability.\[[@CIT43]\] Results and Discussion {#sec1-2} ====================== Hearing screening results for the experimental group are depicted in [Table 1](#T0001){ref-type="table"}. These findings indicate that a large majority (90%) of the newborns obtained *refer* results for TEOAEs at sessions 1 and 2 for the experimental group, while 40% obtained the *refer* findings at session 1 for the AABR. ::: {#T0001 .table-wrap} Table 1 ::: {.caption} ###### Hearing screening results in the experimental and comparison group at the two testing sessions ::: Screening session and measure Experimental group (n = 20) Comparison group (n = 20) ------------------------------- ----------------------------- --------------------------- ------ ------- Pass Refer Pass Refer Session 1 TEOAEs 10% 90% 20% 80% Session 1 AABR 60% 40% 90% 10% Session 2 TEOAEs 10% 90% 20% 80% Session 2 AABR 100% 0% 100% 0% Session 3 TEOAEs 100% 0% 100% 0% Session 3 AABR 100% 0% 100% 0% AABR: *P* = 0.00001 (\< 0.05). The *P* value \[*P* = 0.00001 (\< 0.05)\] rejected the first null hypothesis, thus confirming that epidural anesthesia did have an effect on the AABR hearing screening results of newborns in the current study ::: The aforementioned results for session 1, although consistent with some previous reports of epidural anesthesia causing delayed latencies on the ABR should be interpreted with caution as the presence of vernix in the neonates could have had an additional influence on the AABR. Nonetheless, the fact that vernix could have also had a similar influence on the AABR in the comparison group, but did not, raises a strong index of suspicion about the role of epidural anesthesia, which was the only differentiating factor between the two groups. Of particular interest is the significant improvement in *pass* AABR screening results at sessions 2 and 3; possibly indicating that the effects of the anesthesia may have worn off by then. The same cannot be said for the OAE screening results, which seemed to remain fairly the same at sessions 1 and 2 for both groups; with clear significant changes at session 3. The TEOAE findings confirm documented evidence that OAE screening is more reliable 24 hours after birth, due to vernix. Comparison of hearing screening findings of the experimental group with the comparison group {#sec2-7} -------------------------------------------------------------------------------------------- Overall, as depicted in [Table 1](#T0001){ref-type="table"} and [Figure 1](#F0001){ref-type="fig"}, a large majority of newborns in both groups obtained *refer* TEOAE findings at the earlier testing sessions. The TEOAE hearing screening findings of both groups only changed when screening was conducted after 24 hours (at discharge). The AABR findings indicated a higher (40% at session 1) *refer* rate in the experimental group when compared to the 10% in the comparison group. The AABR findings positively changed with all newborns who had obtained *refer* results at session 1 passing at sessions 2 and 3. Figure 1Comparing the experimental and comparison results These findings may indicate that even though the effective use of OAEs and AABR as screening measures has been well established, it is important to establish factors that may influence the reliability of these measures. In the current study, epidural anesthesia seems to have had an influence with regard to increasing false positive findings when testing was conducted earlier than four hours after birth. Establishing if the time of hearing screening following delivery has any effect on the screening results {#sec2-8} -------------------------------------------------------------------------------------------------------- The p-values of 0.00014 for a two-sided exact significance and 0.00007 for a one-sided exact significance were found when examining if time of screening following delivery had any effect on the results. These values reject the null hypothesis (*P*\<0.05), hence indicating that time of testing did have an effect on the screening results in the current study. From the results obtained, it can be concluded that TEOAE testing earlier than four hours after birth, as well as between four and six hours after birth is unfavorable. This was postulated to be possibly due to the vernix present in the external auditory canal at such times. According to Korres *et al*,\[[@CIT13]\] and the Pennsylvania Health Care Cost Containment Counsel,\[[@CIT15]\] TEOAEs are viable tools during new-born hearing screening between 24 and 48 hours after birth, because at that time the external auditory canal would be free of vernix. An index of suspicion about the influence of epidural anesthesia was raised. Conclusions {#sec1-3} =========== Despite the fact that the sample size for the current study was small; and therefore limited the generalizability of the results, findings from the current study have significant implications for the implementation of universal new-born hearing screening programs; particularly in developing countries, where allocation of resources is driven by priorities such as management of infectious conditions such as the HIV / AIDS pandemic. Knowing where and when to focus the available resources for the best and effective EHDI programs would not only improve service delivery; but may improve access by the general South African population to the services of audiologists; which are currently scarce, particularly in the public healthcare sector. From the current findings, evidence points to the reliability of performing hearing screening with AABR on the day of delivery, as long as that happens four hours following the birth, in newborns where anesthesia was used during delivery. The timing of universal hearing screening, especially in babies born with epidural anesthesia, is important, as the use of epidural anesthesia could lead to increased false positive results, which may therefore cause undue maternal anxiety. The use of OAEs seems to be significantly influenced by vernix in the first few hours following birth, highlighting the need for ensuring that fluids in the external auditory canal are actively cleared before reliable use of OAEs can be implemented as a screening measure before 24 hours. Findings from the current study have particular relevance in developing countries such as South Africa, where women attending state hospitals' maternity wards are discharged with their babies a few hours after giving birth. The current study should be replicated within a larger sample size with diagnostic ABR and OAEs, where findings will not be restricted to *pass* or *refer*; but would provide specific findings about the site and degree of the influence of anesthesia on the auditory pathway. The authors would like to acknowledge Azeemah Mayet, who significantly assisted in conducting data collection for the current study. **Source of Support:** Nil **Conflict of Interest:** None declared.

PubMed Central

2024-06-05T04:04:19.940110

2011-01-01

PMC3053512

During the early stages of the human immunodeficiency virus / acquired immune deficiency syndrome (HIV / AIDS) pandemic, treatment strategies did not seem to have a positive influence on patients' lives, and therefore, hearing loss did not seem to be an important manifestation of HIV / AIDS that required characterization. However, hearing loss has become one of a number of sensory disabilities associated with HIV / AIDS that must now compete for attention by the research and medical community. Friedman and Noffsinger\[[@CIT1]\] were among the first to advocate that as primary professionals in hearing healthcare, audiologists have a responsibility to inform both themselves and other relevant healthcare professionals about this issue, hence the current study. Understanding the effects of HIV / AIDS and the treatment of HIV / AIDS on the auditory system is becoming more important, because patients with HIV / AIDS are living longer due to the positive effects of highly active antiretroviral therapy (HAART). The discovery of antiretroviral drugs for the treatment of HIV / AIDS has changed the face of the HIV / AIDS pandemic internationally, and has also led to changes in the medical field, with people who have HIV / AIDS living for longer periods of time experiencing toxic-related morbidity that influences the quality of life indicators.\[[@CIT2]\] There is a concern, however, that HIV-associated auditory disorders may be seriously underreported. Zuninga\[[@CIT3]\] makes a reference to anecdotal reports suggesting that hearing loss and dizziness, which are often the initial symptoms of the underlying auditory system disease may not have been reported by patients prior to HAART because many patients focused on the life-threatening complications of the HIV disease rather than on the quality of life issues. This situation is yet to be fully realized in developing countries, such as South Africa, where ARVs have not been available for a long time; and where available, access is not the same for the entire population infected by the virus. People who will benefit from these drugs in these countries may in the near future become more conscious of the quality of life issues and complain about them. Numerous clinical, and mostly medically oriented studies have demonstrated the occurrence of hearing loss and other auditory manifestations in HIV / AIDS. Auditory manifestations may be one of the issues that the population will have to deal with; therefore, over and above the management of the known side effects of ARVs, research into the identification and monitoring of all other manifestations of the disease is required. According to the research literature, auditory abnormalities associated with HIV / AIDS and their treatment has been reported in persons with varying degrees of HIV infection, in both symptomatic and asymptomatic patients, as well as in patients on antiretroviral treatment. Indications exist that the HIV effects on the auditory system can be direct as well as indirect; however, this distinction is not always clear and consistent. Early reports in the literature demonstrated that the HIV might directly affect the auditory function due to the fact that the virus is neurotropic and commonly manifests itself neurologically,\[[@CIT4]\] which may be what Kallail *et al*.\[[@CIT5]\] term primary causes of auditory system disorders in HIV / AIDS. These direct causes have been reported to possibly give rise to the central pathology observed in this population.\[[@CIT6][@CIT7]\] More commonly, however, reports in the literature focus significantly on the indirect effects of the virus on the ear. It is believed that indirect causes that result in hearing loss stem from opportunistic infections that require suppressive therapy, thereby leading to ototoxicity;\[[@CIT6]--[@CIT8]\] which Kallail *et al*.,\[[@CIT5]\] refer to as iatrogenic sources. It is important to note that these findings are mainly from developed countries where the presentation and management of HIV / AIDS is different from that in the developing countries, suggesting a need for more research in this area, particularly as the number of adults living with HIV / AIDS in developing countries such as South Africa is still high, and also because the context is different. With regard to auditory manifestations, both identification and monitoring of ototoxicity require rigorous research, to enhance the patients' quality of life, particularly as internationally, a link has been established between ARVs and ototoxicity. On account of all the diseases and infections that the population with HIV / AIDS present with, it is not surprising to find patients with hearing loss due to ototoxicity, as this population goes through a drug regimen that often involves potentially ototoxic medications.\[[@CIT9]\] Bankaitis and Schountz\[[@CIT8]\] report that the use of experimental antiretroviral drugs, with undocumented or unknown side effects, contributes to this hearing loss. In addition, ototoxic drugs that are often used in the treatment of opportunistic infections such as tuberculosis may increase the potential for a drug-induced hearing loss in this population.\[[@CIT10]\] Internationally, iatrogenic hearing loss has been associated with many of the drugs used to treat HIV / AIDS and its associated complications. As early as 1998, the potential for a drug-induced hearing loss in an HIV-infected individual at any stage of the disease was reported to be relatively high.\[[@CIT8]\] With all the medications that individuals with HIV are taking and the continual developments in HIV therapies, it is challenging to acquire and maintain a comprehensive knowledge base of HIV-related drugs and their associated ototoxicities. Although the side-effects of many antiretroviral drugs are yet to be determined, HIV-infected individuals are often prescribed medications as a prophylaxis or treatment of opportunistic infections that have been long associated with the development of audiological and vestibular changes. Antineoplastic medications such as vincristine, antifungal agents, including amphotericin B, flucytosine and ketoconazole, immune modulators, aminoglycoside antibiotics, erythromycin, and azidothymidine (AZT) are all widely used in the management of HIV and are all reported to be associated with significant ototoxicity or decreased hearing.\[[@CIT7][@CIT8][@CIT11]--[@CIT14]\] These medications are associated with hearing loss, tinnitus, and vertigo. Frequently administered medications for *Pneumocystis* pneumonia (PCP) (Pentamidine, TMP / SMX, Primaquine) may cause tinnitus, vertigo, dizziness, auditory disturbances, deafness, decreased hearing, hearing loss, and otalgia.\[[@CIT8]\] Moreover, the use of experimental medications with relatively unknown toxicity as well as the use of ototoxic drugs such as anti-tuberculosis (TB) medications, in combination, adds to the overall effect on hearing.\[[@CIT15]\] In South Africa, one of the most frequently administered treatments for the HIV / AIDS population was TB treatment. South Africa, like many sub-Saharan countries, witnessed a dramatic upsurge of TB cases over the past decade.\[[@CIT16]\] This upsurge in the number of TB cases was expected to continue, largely due to co-infection with HIV, with the emergence of drug-resistant TB\[[@CIT17]\] also being reported. This co-occurrence of HIV/AIDS and TB raises serious implications for the audiologist with regard to the possible association between TB treatment and antiretroviral therapy (ART). As some of the drugs used in the treatment of TB fall under the umbrella term 'aminoglycosides',\[[@CIT18]\] interactions between these treatments need to be explored. Examples of these aminoglycosides include amikacin, gentamicin, kanamycin, netimicin, paromomycin, streptomycin, tobramycin, and apramycin.\[[@CIT19]\] These antibiotics are most notorious for being ototoxic, primarily targeting the renal and cochleovestibular system.\[[@CIT12]\] This impact of medications on the hearing function are being reported, although not extensively, with nucleoside analogue reverse transcriptase inhibitors (NRTIs). Although a variety of adverse effects have been attributed to treatment with NRTIs for HIV-1 infection, only a small number of cases of ototoxicity have been reported in literature. Simdon reported three subjects who experienced ototoxicity, all of whom were over the age of 45 and received combination ART with two-to-three NRTIs plus a non-nucleoside reverse transcriptase inhibitors (NNRTI) or a protease inhibitor (PI). All three subjects had prior hearing problems, prior exposure to occupational noise, and all developed significant tinnitus.\[[@CIT20]\] Clearly, the presence of these confounding variables (prior hearing loss, noise exposure history, and older age) needs to be taken into consideration when interpreting the findings from these cases. The authors suggest that NRTIs must be used cautiously in patients with pre-existing hearing loss. Again, the ability to generalize these results is limited as they are based on case reports and not on large samples. These authors suggest that reductions in mitochondrial DNA content induced by NRTIs, as well as mitochondrial DNA mutations associated with aging and HIV-1 infection, may all contribute to auditory dysfunction in older patients with HIV-1 infection. They highlight the fact that prospective studies are necessary to determine the incidence of tinnitus and hearing loss among HIV-1-infected patients and their relation to the use of NRTIs.\[[@CIT20]\] Several cases of ototoxicity have been reported in HIV-infected patients treated with zalcitabine;\[[@CIT21]--[@CIT23]\] didanosine;\[[@CIT24]\] zidovudine;\[[@CIT20]\] and combinations of zidovudine and didanosine;\[[@CIT25]\] stavudine and lamivudine;\[[@CIT15]\] stavudine, lamivudine, didanosine, and hydroxyurea;\[[@CIT15]\] and post exposure prophylaxis with stavudine, lamivudine, and nevirapine.\[[@CIT26]\] Moreover, a study of 99 HIV-infected individuals who received antiretroviral drugs showed that hearing loss was common in this population. Hearing loss was significantly associated with being 35 or older and with a history of ear infection, and there was a trend toward an association with a documented receipt of therapy with antiretroviral drugs in the preceding six months.\[[@CIT27]\] As illustrated earlier, previous cross-sectional studies and case reports have shown an association between hearing loss and NRTI therapy.\[[@CIT15][@CIT27][@CIT28]\] There have been two case reports of hearing loss in persons receiving ART regimens that included NRTIs and a second class of antiretroviral drugs; one with an NNRTI (Nevirapine) and one with a PI (lopinavir / ritonavir), each combined with NRTIs, (both these subjects also received stavudine and lamivudine). One case reported sudden hearing loss two weeks subsequent to the person completing one month of post-exposure prophylaxis, which resulted in long-term hearing loss.\[[@CIT26]\] The other case described hearing loss in a subject with extensive HIV pre-treatment, which suggested a possible relationship with the protease inhibitor, although there were other possible explanations noted in Simdon's reply to this case report.\[[@CIT20][@CIT29]\] One should note that not all of the aforementioned studies utilized sensitive ototoxicity monitoring protocols such as ultra-high frequency audiometry and / or otoacoustic emissions. Furthermore, some of these studies did not follow longitudinal research designs either, which could have allowed the researchers to investigate within-subject changes; but they rather followed the cross-sectional methodology designs. In addition, the reports that other factors such as age, drug interactions, concomitant noise exposure, and so on may have an influence on the ototoxicity of ARVs should be taken into consideration when reviewing the effects of ARVs on hearing. Although ototoxic hearing loss has been described in HIV-infected people after beginning NRTIs, there have been extremely limited prospective studies, with one published example of a prospective study by Schouten, Lockhart, Rees, Collier, and Marra.\[[@CIT30]\] Hence, there still need to be extensive investigations to clearly establish and confirm this relationship. The study by Schouten *et al*.\[[@CIT30]\] investigated hearing changes longitudinally in treatment-naïve HIV-infected subjects, following initiation of regimens containing NRTIs. The goal of their study involved performing a prospective assessment of the contribution of zidovudine (ZVD) and didanosine (ddI) to hearing loss. Changes in hearing levels at all frequencies and in low and high frequency pure tone averages were measured at baseline, 16, and 32 weeks after initiating antiretroviral therapy. In Schouten *et al.'s*\[[@CIT30]\] study, treatment with ZVD and ddI did not result in loss of hearing, even after taking into account noise exposure, immune status, and age. The results of this prospective pilot study did not support the view that treatment with nucleoside antiretroviral drugs, damages hearing. This finding contradicts reports from previous cross-sectional studies and case reports that have indicated that hearing loss may be common among HIV-infected people due to ototoxic drug therapy.\[[@CIT27][@CIT31]\] The results of the prospective study by Schouten *et al*.\[[@CIT30]\] did not corroborate this relationship and are consistent with the report from the Adult / Adolescent Spectrum of HIV Disease Project Group that demonstrated no association between hearing loss and drugs used in the treatment of HIV. Of note, however, the Adult / Adolescent Spectrum of HIV Disease Project Group study was centred on a retrospective chart review for International Classification of Diseases (ICD) - 9 coding for hearing loss and not on a formal audiometry.\[[@CIT28]\] This represents a significant weakness in the methodology for a study attempting to determine the ototoxic effects, which can be subclinical in nature, hence requiring sensitive audiological monitoring tools. There are at least three criticisms that can be leveled against the aforementioned study by Schouten *et al*.\[[@CIT30]\] First, this study did not incorporate otoacoustic emissions (OAEs) as part of their monitoring battery, and this could have had a significant impact on their results, as OAEs have been shown to be sensitive to cochlear damage in ototoxicity monitoring. Second, only 33 participants were included in their study, a small sample size that significantly reduces the strength of the study in terms of the ability to generalize the findings. Moreover, a small sample size limits the power of this study to detect a difference and also limits the ability to accurately interpret the results. Third, there was no control group, although the researchers did acknowledge that this was a pilot study. To their credit, these authors' pure tone testing included 12 kHz, which is an ultrahigh frequency. Ultrahigh frequencies have been reported to be finely tuned to the effect of damaging environmental factors such as noise and ototoxic drugs.\[[@CIT12]\] Replication of studies, such as Schouten *et al.'s*\[[@CIT30]\] longitudinal study, in developing countries such as South Africa, may be challenging due to a number of factors. First, the nature of the HIV / AIDS disease and the population being studied may preclude complete control over the confounding variables that could have had an influence on the results such as interactions of ARVs with other therapies; especially traditional medicine in the form of '*ubhejane*,' which has been reported to be in widespread use.\[[@CIT32]\] Although, the current researcher is of the opinion that isolating all the possibly contributing confounding variables may provide a more accurate answer, it may not necessarily provide a practical, relevant, and context-sensitive finding. Within the South African AIDS population for example, it may be impossible to find participants who are only exposed to just one strict ARV regimen without any other medications coming into play. Second, securing a decent-sized comparison groups may be difficult, thereby preventing randomized matching of participants in the comparison group with those in the experimental group. Challenges in obtaining large enough sample sizes for control groups may be due to factors such as, attrition, on account of patients commencing treatment during the study as well as loss to follow-up. Third, ultra-high frequency audiometry, which does not form part of the routine audiological test battery, may influence the type of results found; and this may result in the clinical changes in the ultrahigh frequencies depicted on the audiogram being entirely missed. Finally, the length of time for which audiological monitoring occurred may be too short, due to attrition, to allow for the clinical hearing loss possibly caused by ART to manifest and therefore be detected on the audiogram. Nevertheless, such longitudinal studies of patients on various regimens of ARVs need to be conducted. These need to be carried out in order to determine if any hearing changes occur during the period when the patients receive ARVs. Both clinically significant and statistically significant changes need to be investigated, as the presence of statistically significant changes does not necessarily translate to clinically significant findings. It is also critical that measures such as DPOAEs, which are sensitive to microcochlear changes, form part of the methodologies employed, as DPOAEs have been shown to be superior to pure tone audiometry in this regard;\[[@CIT33]\] hence the current study, which aims to monitor the auditory status in a group of adults with AIDS receiving HAART (lamivudine, stavudine, and efavirenz), in a hospital outpatient clinic, in Gauteng. Materials and Methods {#sec1-1} ===================== Research, aims, and objectives {#sec2-1} ------------------------------ ### Primary aim {#sec3-1} The primary aim is to monitor the auditory status in a group of adults with AIDS receiving HAART (lamivudine, stavudine, and efavirenz), in a hospital outpatient clinic, in Gauteng. ### Specific objectives {#sec3-2} To longitudinally assess hearing function in AIDS-infected adults on HAART (experimental group);To longitudinally assess hearing function in AIDS-infected adults not on HAART (comparison group);To compare the results of the experimental group with those of the comparison group;To analyze the hearing function in the group of adults with subclinical hearing loss. The null hypothesis was that the participants' hearing function before and after antiretroviral drug-use would remain the same. The alternative hypothesis was that it would not remain the same, that is, the participants would present with changes in their auditory function.\[[@CIT34]--[@CIT36]\] Design of the study {#sec2-2} ------------------- As an extensive literature search yielded a paucity of both South African and internationally published data on this topic, the study was exploratory and longitudinal in nature. The design utilized was a repeated measures, quasi-experimental design, with pre- and post-treatment testing, and a control group.\[[@CIT37]\] A quasi-experimental design was reported to be the best design where there were practical and ethical barriers to conducting randomized controlled trials.\[[@CIT38]\] In the current study, although there was manipulation of the independent variables (the antiretroviral drugs) and a control group, there was no random allocation of participants, which led to the study being quasi-experimental instead of experimental in nature.\[[@CIT34]\] The antiretroviral medications and other therapies were the independent variables, with the audiological measures (otoscopy, impedance audiometry, pure tone audiometry, otoacoustic emissions) featuring as dependent variables. The aim was to investigate and monitor the auditory status in a group of adult patients with AIDS receiving ART and other therapies before and during antiretroviral treatment --- with measures taken before commencement of ARVs, three months after initiation of treatment, and six months into therapy. All of the objectives were examined at baseline (before initiation of ARVs) and with repeated measures (three and six months into treatment) for both the control and experimental group. A comparison of results of the control group and experimental group was done for all objectives. One methodological limitation of the current study was related to the time when the last measure was carried out (six months after treatment). This could have limited the type of results obtained, as ototoxicity could present long after six months; however, the researcher decided on this time in an attempt to control for variables such as: patients changing medications, participant attrition via patients leaving the study for various reasons, and so forth. Participants {#sec2-3} ------------ A total sample of 54 adults (between the ages of 18 and 50 years) in the experimental group and 16 participants in the comparison group participated in the study. The patients selected for this study were recruited from the Hospital's Adult HIV / AIDS clinic. Patients attending this clinic had already been diagnosed with HIV / AIDS and were seen there for general medical management as well as antiretroviral treatment and monitoring. At the time of the study all patients with CD4+ counts below 200 cells / mm^3^ had access to ARV treatment at this clinic --- and this was the group that was targeted for the experimental group. Participant selection criteria - Inclusion criteria {#sec2-4} --------------------------------------------------- Given the fact that little, if any, published research had been conducted on this aspect of HIV/AIDS in South Africa, the researcher believed that it was crucial to have a high degree of control over the variables, which could confound the results of the study (e.g., noise exposure, syphilis, and so forth).\[[@CIT37][@CIT39]\] Consequently, the participant inclusion criteria that were adopted following baseline testing are depicted in [Table 1](#T0001){ref-type="table"}. ::: {#T0001 .table-wrap} Table 1 ::: {.caption} ###### Summary of participant inclusion criteria following baseline measures ::: Criterion Experimental group Control group --------------------------------------------------------------------- -------------------- --------------- HIV / AIDS positive serology Yes Yes On ARVs Yes No Age between 18 and 50 years Yes Yes Alert and oriented Yes Yes Normal pure tone audiometry (thresholds better or equal to 25 dBHL) Yes Yes ::: Participant selection criteria - Exclusion criteria {#sec2-5} --------------------------------------------------- The following criteria were strictly observed for persons who participated in the repeated measures following baseline \[[Table 2](#T0002){ref-type="table"}\]: ::: {#T0002 .table-wrap} Table 2 ::: {.caption} ###### Summary of participant exclusion criteria following baseline measures ::: Criterion Experimental group Control group ---------------------------------------------------------------------------------------- -------------------- --------------- Noise exposure Yes Yes Recent (less than three years) or current history of treatment for TB and radiotherapy Yes Yes Positive clinical or serological evidence of syphilis Yes Yes Middle ear pathology Yes Yes Presence of tinnitus Yes Yes Recent (less than three years) history of previous ARV use Yes Yes ::: Recruitment and sampling procedure {#sec2-6} ---------------------------------- A nonprobability convenience sampling technique was utilized in the study, as the sample was restricted to a part of the population that was readily available,\[[@CIT37][@CIT39]\] and true random sampling would have been difficult to achieve due to time, cost, and equipment limitations. Participants for the control group were recruited from the wellness section of this clinic, where patients who refused treatment were seen by Dieticians and Social Workers. At the time of the study, the researcher was already providing service to patients at this clinic, as she was on an honorary appointment as an Audiologist at the research site. Research procedures and materials {#sec2-7} --------------------------------- Participants underwent case history interviews and medical record reviews, otoscopy and tympanometry, as well as conventional pure tone audiometry and distortion product otoacoustic emission testing. Baseline data were collected by assessing the participants' dependant variables before administration of ARV therapy. These baseline data were then compared with two other measures that were taken three and six months after commencement of therapy. The same procedures were followed for the control group. For those participants who did not attend all three sessions of testing, their data were excluded from the inferential statistical analysis, as this required data from all three sessions. Following infection control measures proposed by Kemp and Roeser,\[[@CIT40]\] all testing was conducted in a sound-proof booth. Basic audiological testing followed by DPOAE measurements for all participants was undertaken and systematically recorded. ### Case history {#sec3-3} A case history form that targeted the signs and symptoms of auditory manifestations was utilized, in order to gather all the important case history information, audiological data, and some medical variables that could have exerted an impact on the results of the study. ### Otoscopy {#sec3-4} The researcher evaluated the participants' ears for the presence of impacted wax; otitis externa; possible otitis media; perforated tympanic membranes; collapsed ear canals; presence of any growths, or any other ear disorders.\[[@CIT41]\] These otoscopic abnormalities were reported to have a significant effect on DPOAE and therefore needed to be documented before testing commenced.\[[@CIT33]\] ### Impedance audiometry {#sec3-5} Impedance audiometry in the form of tympanometry (through the use of Inter-AcousticAZ26 audiotympanometer) was utilized to assess the status and integrity of middle ear functioning. A standard single frequency tympanometry using an 85 dB SPL tone, set at 226 Hz was done. The primary purpose of impedance audiometry was to determine the status of the tympanic membrane and middle ear via tympanometry. The researcher ensured that all participants undergoing DPOAE testing had normal (type A tympanogram) tympanometry results. ### Pure tone audiometry {#sec3-6} Conventional (250 Hz -- 8000 Hz) pure tone audiometry was performed on all participants through the use of an Inter-Acoustic AC 40 diagnostic audiometer. The criteria used to define normal hearing, was that of pure tone thresholds of 25 dBHL or lower across all frequencies, with the absence of an air-bone gap.\[[@CIT42]\] If pure tone air conduction and tympanometry were abnormal at any test frequencies, in the pre-treatment phase, the participants were excluded from continuing in the study, and were referred to the Ear, Nose, and Throat Specialists for assessment and management, and were subsequently offered appropriate audiological rehabilitation. Participants presenting with normal pure tone audiometry at the baseline were advanced to sessions two and three of the study, where ototoxicity monitoring was conducted. Using pure tone data, a change in the hearing level of 10 dB at one or more frequencies was commonly taken to be indicative of some significant change,\[[@CIT43]\] hence, this protocol was followed in the current study. ### Distortion product otoacoustic emissions {#sec3-7} All participants with normal middle ear functioning underwent DPOAE measurements, as a crucial aspect of ototoxicity monitoring, through the use of a Biologic Scout Otoacoustic emissions meter. OAE testing is often used as a screening tool to determine the presence or absence of cochlear function, and analysis can be performed for individual cochlear frequency regions, therefore, they are regarded as an excellent tool for the early detection of cochlea damage due to ototoxicity.\[[@CIT12][@CIT33]\] OAEs can detect cochlear dysfunction before it is evident on pure tone audiometry, and in ototoxicity monitoring this factor is critical, as the main aim of monitoring is the early detection of adverse effects of the drug before it causes clinical damage.\[[@CIT12][@CIT33][@CIT44]\] One methodological limitation of the current study was that the ultra-high-frequency audiometry did not form part of the ototoxicity monitoring protocol due to lack of equipment available at the time of the study. Ultra-high-frequency testing has been reported to be sensitive to ototoxicity. As far as DPOAEs are concerned, the current study only monitored frequencies up to 8000 Hz - and this was another acknowledged methodological limitation, as literature has indicated greater sensitivity when using higher frequencies in ototoxicity monitoring. The following DPOAE test protocol was employed: ::: {#d32e612 .table-wrap} ------------------ ------------------------------------------------------ Test parameters Diagnostic / High frequency Stimuli  Intensity level L~1~-L~2~ = 10 dB (e.g., L~1~ = 65 dB, L~2~ = 55 dB)  Ratio f~2~/f~1~ = 1.22  Frequency range 750 to 8000 Hz ------------------ ------------------------------------------------------ ::: The presence of the DPOAE was determined by comparing the amplitude of the DPOAE to that of the noise floor to calculate the size of the emission. A DP amplitude that exceeded the noise floor by at least 7 dB across all frequencies measured was regarded as indicative of a normally functioning cochlea.\[[@CIT33]\] The size of the emission at the different frequencies measured was then monitored over the three testing sessions. Validity and reliability {#sec2-8} ------------------------ Test reliability was controlled and maintained at a high level by standardizing test administration, ensuring proper equipment calibration, and controlling patient variables. For all audiological assessments precautionary measures advocated by Bess and Humes\[[@CIT45]\] and Hall\[[@CIT33]\] were followed in terms of proper maintenance and calibration of the equipment; optimizing testing environment; correct earphone and bone vibrator placement, and proper probe placement for DPOAE. All testing was conducted in a soundproof booth or sound-treated room with equipment that was calibrated on an annual basis, with biological calibration conducted before every test session. All the participants were tested by the same researcher using the same test procedure at all three sessions. Furthermore, all the patients were tested in the mornings to reduce the effect that fatigue could have on the patients' responses to behavioral audiometry testing. However, threats to validity in the current study were present, and they included the fact that the study was not a double-blind study, as the researcher was aware of which participants were in the control group and which were in the experimental group. There was no random selection of participants to reduce bias in the sample; and there was limited control over the confounding variables such as interaction of ARVs with previous exposure to ototoxic drugs, and interaction of ARVs with other routine medications and supplements that were being prescribed at the time of the study.\[[@CIT34][@CIT38]\] Finally, due to the sample size and the fact that the data were collected in one hospital in Gauteng, South Africa, the researcher's ability to generalize the results from the sample studied to the total population of adults with AIDS in South Africa is limited. Data analysis and statistical procedures {#sec2-9} ---------------------------------------- Both descriptive and inferential statistics were used to analyze data from the study. Inferential statistics in the form of repeated measures, using analysis of variance (ANOVA), Multivariate analysis of variance (MANOVA), and Tukey-Kramer post-test were used to establish statistical significance levels, and to determine when the statistically significant changes occurred within the longitudinal design of the study. Furthermore, the clinical significance of the findings was also analyzed. First, a statistical comparison was done, basing the results on the average change from the baseline. Each frequency's mean change from the baseline for the ears, individually, was computed and then combined for both pure tone audiometry and distortion product otoacoustic emissions. Repeated-measures analysis of variance\[[@CIT34]\] was used to compare the mean change from the baseline for the control group and the experimental group, from session to session. In the analysis of the DPOAE data, the baseline DPOAE levels in decibels SPL for each f2 value tested were compared with the corresponding measurements in sessions 2 and 3. The signal-to-noise difference was used as the measure of DPOAE amplitudes. For pure tone audiometry data, the baseline thresholds in decibels HL for each frequency were compared with the corresponding results in sessions 2 and 3 as well. To statistically test the hypothesis, a threshold *P* value (alpha) of 0,05 was selected.\[[@CIT46]\] Finally, as part of a statistical analysis of the data, a post-test in the form of the the Tukey-Kramer multiple-comparison post-test was conducted. This test was conducted to compare pairs of group means so as to identify where, precisely, statistically significant changes occurred along the time continuum (baseline to six months) - if they did. Second, for the purposes of the current study, clinical significance (changes that are deemed significant enough to indicate structural and functional changes in the ear) over and above the statistical significance was examined. There is a growing recognition that assessing an intervention's effect should not only focus on the statistical significance of the differences between the experimental and control groups, but should also focus on the relevance or importance of these outcomes.\[[@CIT47]\] For pure tone testing, most often a change of 10 dB at one or more frequencies is commonly taken to be indicative of some significant change;\[[@CIT43]\] and this is the protocol followed in the current study. As far as DPOAEs are concerned, change is only regarded as significant in the DPOAE measures if there is a change of at least more than 6 to 9 dB in the DPOAE level between consecutive measures.\[[@CIT48][@CIT49]\] Absence of clinically significant changes in the DPOAEs confirm intact cochlear functioning and the absence of cochlea damage, even microcochlear pathology - which is often evident on OAEs, long before being depicted on the pure tone audiogram.\[[@CIT33]\] During exploratory data analysis, it was discovered that there were participants in the experimental group who presented with changes on DPOAE measures, without changes in pure tone audiometry. This presentation of results necessitated an additional analysis step where these participants (referred to as the *subclinical hearing loss group*) were analyzed separately; hence this formed an additional set of results. The following table \[[Table 3](#T0003){ref-type="table"}\] provides a summary of all collection material and test procedures used in this study. ::: {#T0003 .table-wrap} Table 3 ::: {.caption} ###### A summary table of collection materials and test procedures used in the current study ::: Equipment Function Pass criteria Fail criteria --------------------------------- ------------------------------------------------------------------------ -------------------------------------------------------------------------------------------------------------------------- -------------------------------------------------------------------------------------- Case history form Gather important case history data Refer to inclusion and exclusion criteria Refer to inclusion and exclusion criteria Welch Allyn Otoscope Visual inspection of the ear Clear outer ear with normal tympanic membrane Obstruction; abnormal tympanic membrane, pathologies of the outer and middle ear AZ26 Interacoustic tympanometer Middle ear functioning assessment Type A tympanogram Other tympanograms, but type A AC40 Diagnostic audiometer Conventional pure tone audiometry (250 -- 8000 Hz) Monitoring function Thresholds at and better than 25 dBHL and no Air-Bone Gap No 10 dB threshold change at one or more frequencies over time Thresholds worse than 25 dBHL A change of 10 dB at one or more frequencies over time Biologic Scout OAE machine Diagnostic DPOAE measurement Greater than 7 dB DP amplitude at frequencies assessed Less than 7 dB DP amplitude Ototoxicity monitoring No DPOAE level change of more than 6 to 9 dB between consecutive measures Change of more than 6 to 9 dB in the DPOAE level between consecutive measures ::: Ethical consideration {#sec2-10} --------------------- Prior to the commencement of the study, permission to conduct the research project was sought from the University of the Witwatersrand Human Research Ethics Committee (Medical), which gave unconditional ethical clearance in the form of protocol number M041131. The researchers ensured that permission to conduct the study was obtained from the hospital management and from the Heads of the Audiology and HIV / AIDS clinics at the research sites. Written informed consent to participate in the study was obtained from all the participants before the study was conducted, with an assurance that confidentiality of all records would be maintained. Furthermore, to ensure anonymity, the researcher ensured that no personal or identifying information was included in the research report and research coding numbers instead of identifying information were used. The current study also reduced risks to the participants to a minimum, by conforming to the ethical principles\[[@CIT50]\] and observing provisions of the Nuremberg Code of ethics\[[@CIT34]\] during the study. Finally, the hospital and participants could request to see the research results if they were interested. Results {#sec1-2} ======= As indicated in [Table 4](#T0004){ref-type="table"}, the current investigation revealed noteworthy findings. As the researcher looked at both the statistical and clinical significance, the findings highlighted the importance of including both these means of establishing the significance in any longitudinal audiological study. Specifically, pure tone audiometry results in the comparison group revealed hearing within normal limits, with the average PTAs being above the level regarded as indicative of normal hearing across all frequencies evaluated, as depicted in [Figure 1](#F0001){ref-type="fig"}. These mean results were normal for all three testing sessions. With regard to the MANOVA tables, all changes in the comparison group were found not to be statistically significant over the three testing sessions. Moreover, these changes were also not audiologically clinically significant changes. Most often a change of 10 dB at one or more frequencies is taken to be indicative of a clinically significant change.\[[@CIT43]\] None of the mean changes in pure tone results of the comparison group in the current study were greater than 10 dB; suggesting that they were not clinically significant changes. ::: {#T0004 .table-wrap} Table 4 ::: {.caption} ###### Summary of ototoxicity monitoring findings from the current study ::: Factor Comparison group (N=16) Experimental group (N=54) --------------------------------------------------- --------------------------------------------------- -------------------------------------------------------------- Pure tone audiometry (PTA) Normal throughout the three testing sessions Normal throughout the three testing sessions Clinical analysis No clinically significant changes No clinically significant changes Statistical analysis No statistically significant changes Statistically significant changes at 8000 Hz Distortion product otoacoustic emissions (DPOAEs) Normal OAEs throughout the three testing sessions Reduced / absent OAEs at session 3 Clinical analysis No clinically significant changes Clinically significant changes at 6 and 8000 Hz at session 3 Statistical analysis No statistically significant changes Statistically significant changes at session 3 Key: PTA change \> 10 dB = Clinically significant change; DPOAE change \> 6dB = Clinically significant change. *P* \< .05 = statistically significant change ::: ::: {#F0001 .fig} Figure 1 ::: {.caption} ###### Mean bilateral pure tone audiometry results (in dBHL) and their standard deviations for the control group at the three different sessions (n=32 ears) :::  ::: Furthermore, DPOAE measures for the comparison group \[[Figure 2](#F0002){ref-type="fig"}\] revealed cochlea functioning to be within normal limits, with the average DPOAE emission size being above the level regarded as indicative of normally functioning cochleas, across all frequencies evaluated for all three sessions. The MANOVA tables indicated that changes were not statistically significant (*P*\>.05). Furthermore, these changes were also not clinically significant. Change is only regarded as clinically significant in DPOAE measures if there is a change of at least 6 to 9 dB in the DPOAE level between consecutive measures.\[[@CIT48][@CIT49]\] In the comparison group, none of the mean changes in DPOAE results were greater than 6dB - indicating that no clinical changes occurred in the cochlear function. The absence of clinically significant changes in pure tone audiometry and DPOAEs in the comparison group, in the three different sessions, over a period of six months, confirmed intact hearing functioning and the absence of any evidence of cochlea damage, even microcochlear pathology - which is often evident in OAEs long before being depicted on the pure tone audiogram.\[[@CIT33]\] ::: {#F0002 .fig} Figure 2 ::: {.caption} ###### Mean bilateral distortion product otoacoustic emission results (in dBSPL) and their standard deviations for the control group at the three different sessions (n=32 ears) :::  ::: For the experimental group, however, the ototoxicity monitoring phase yielded different results to those of the comparison group, in that, some results indicated the presence of both statistically and clinically significant changes \[[Table 4](#T0004){ref-type="table"}\]. For pure tone audiometry data, the criteria for clinically significant change were not met --- none of the mean changes in the pure tone results were greater than 10 dB \[[Figure 3](#F0003){ref-type="fig"}\]. The MANOVA tables for repeated measures analysis of variance (within group) revealed no statistically significant changes (alpha was greater than 0.05), with the exception of significant changes at 8000 Hz \[left ear (*P*=0.04)\]. ::: {#F0003 .fig} Figure 3 ::: {.caption} ###### Mean bilateral Pure Tone audiometry results (in dBHL) and their standard deviations for the experimental group at the three different sessions (n=108 ears) :::  ::: However, results of the DPOAE analysis revealed the cochlea function to be normal at sessions 1 and 2, at all frequencies evaluated, with changes at repeated measures, indicating declining DPOAE values \[[Figure 4](#F0004){ref-type="fig"}\]. These changes were found to occur at all evaluated frequencies, but were more clinically significant at 6 and 8 kHz bilaterally in session 3. The DPOAE results at these two frequencies at session 3 were in fact below the norm, in that, the DPOAE value did not exceed the noise floor by at least 7 dB, as expected in a normally functioning cochlea. Statistically, the MANOVA \[within group (time)\] results indicated extremely significant *P* value (*P*\<.001) for all frequencies assessed - implying that the cochlear function changed after ARV initiation. ::: {#F0004 .fig} Figure 4 ::: {.caption} ###### Mean bilateral DPOAE results (in dBSPL) and their standard deviations for the experimental group at the three different sessions (n=108 ears) :::  ::: The presence of significant changes on DPOAEs in the three different sessions, over a period of six months, in the experimental group, indicated possible microcochlear pathology that was not necessarily indicated on pure tone audiometry. Hence, it would seem to be a possibility that based on the objective nature of the DPOAE measures, subclinical auditory changes occurred after a six-month period. This finding also highlights the crucial need for the use of such sensitive measures (DPOAE) in monitoring the possible effects of toxins on the ear, as DPOAEs have been shown to be superior to pure tone audiometry in this regard. For the statistically significant results for DPOAEs in the 'within group' analysis, the Tukey-Kramer test indicated that generally, significant changes occurred between baseline measures and session 2 (at three months) for the lower frequencies, with the higher frequencies being significantly affected all through the three sessions of testing. This early onset of symptoms (although subclinical) again highlights the need for early involvement of the audiologist in the assessment and management of patients with AIDS. Findings from the *subclinical hearing loss* group further illustrate that subtle subclinical changes can only be identified by DPOAEs and not pure tone audiometry. [Figure 5](#F0005){ref-type="fig"} depicts the findings of 45 participants who presented with normal pure tone audiometry results, but showed significant changes on DPOAEs over the three testing sessions. ::: {#F0005 .fig} Figure 5 ::: {.caption} ###### Mean bilateral DPOAE results (in dBSL) and their standard deviations for the group with normal PTA at the three different sessions (n=90 ears) :::  ::: The presence of DPOAE changes indicating subclinical changes in hearing status suggested hearing loss that could have possibly been due to medications used by the participants in the experimental group (lamivudine, stavudine, and efavirenz). These results were consistent with the reports that have associated iatrogenic hearing loss with many of the drugs used to treat HIV / AIDS.\[[@CIT11][@CIT14]\] Although the current study may not be able to draw a definitive conclusion that ARVs used in the sample had a direct effect on hearing - because of the difficulty in controlling the other interacting factors - an index of suspicion is raised by the fact that the results of the participants in the experimental group seem more consistent with ototoxic hearing loss than those in the comparison group, and the fact is that the only identifiable difference between the two groups is that of ARV use. However, the current researcher is of the opinion that isolating all the possible contributing confounding variables may provide a more accurate answer, but may not necessarily provide a practical, relevant, and context-sensitive finding. Within the South African AIDS population, it may be impossible to find participants who are only exposed to just one strict ARV regimen without any other medications coming into play. The results of this study are felt to be valuable and applicable to the South African context, and may provide more realistic and context-specific implications; those that are needed for establishment and implementation of ototoxicity monitoring protocols as part of the routine clinical management of adults with HIV / AIDS. Discussion and Conclusion {#sec1-3} ========================= This study displays and highlights an expanded role of the audiologist in Food and Drug Administration processes of drug development, drug approval, and drug monitoring. The sub-clinical hearing changes (cochlea function changes detected before the hearing loss is seen on an audiogram) highlight the importance of DPOAEs in ototoxicity monitoring studies.This research also demonstrates the important role that the audiologist may have in both the assessment and treatment of patients with HIV / AIDS. Finally, implications raised by the current study can be translated into recommendations for the clinical assessment and management of patients with HIV / AIDS who are taking ARVs; education of team members; policy formulation; as well as further research. Medical awareness of ARV doses, forms of administration, populationsat risk, and possible synergism with other factors is necessary to developappropriate care in the prescription of drugs with possible or established ototoxicside effects. Furthermore, issues such as risk-benefitanalysis, patient-informed consent, and quality-of-life considerations, are also crucial factors to beconsidered in the management of patients with HIV / AIDS. Regardless of whether the effects of the drug are negligible or not, these effects still need to be determined so that proper patient adherence counseling can take place. It is fundamental that audiologists establish and become aware of the ototoxic effects of medications used to manage chronic conditions such as HIV / AIDS, as also medications prescribed to significant numbers of people - such as 11% of the population afflicted by HIV / AIDS in South Africa.\[[@CIT51]\] This awareness is critical, to ensure that proper patient education occurs, as patients may not notice ototoxic hearing loss until a communication problem becomes evident, signifying that hearing loss within the frequency range, which is vital for understanding speech, has already occurred. Likewise, by the time the patient complains of dizziness, permanent vestibular system damage may have already occurred. Clinically used drugs and chemical agents may potentially causeadverse effects to the human auditory and vestibular systems.\[[@CIT52]\] Many of these drugs can playa critical role in the treatment of serious or life-threateningdiseases. others offersuch important therapeutic effects compared to the ototoxicside effects that the ototoxicity risk can be considered tobe of minor importance, and such may be the case with HIV / AIDS (a sentiment echoed by some physicians). The problem of ototoxic side effectsis reported to be more critical in developing countries, where highly effectiveand low-cost drugs are more easily prescribed without adequatemonitoring.\[[@CIT53]\] It is possible that such a situation may exist in some parts of Africa, particularly with the high numbers of patients on treatment for HIV / AIDS. An additional concern in the management of HIV / AIDS patients, who may be on potentially ototoxic medication without being audiologically monitored, is that noise exposure following treatment with ototoxic drugs can act synergistically with the drugs that have not been fully cleared from the inner ear.\[[@CIT54]\] Increased susceptibility to hearing loss can continue for several months after completion of treatment or therapy. Due to this likelihood, it is imperative to implement hearing conservation in the form of advising patients to avoid excessive noise exposure for at least six months. In addition, patients who use amplification in the form of hearing aids may need to be counseled and warned to closely monitor and control the hearing aid maximum output during this critical time.\[[@CIT55]\] Given this scenario, it seems more pressing than ever to endeavor to prevent or ameliorate the possible ototoxic hearing loss in this population, by ensuring ototoxicity monitoring as part of the routine clinical management; particularly as the treatment regimen is varied and the WHO ART guidelines continue to be modified, as some drugs get phased out, such as the recent suggestion by WHO ART to phase out d4T. When life-threatening illness necessitates treatment with ototoxic drugs, preserving the quality of the patients' remaining life is customarily a treatment goal. Early detection of ototoxic hearing loss provides physicians with critical information and the opportunity necessary to minimize further impairment, and in some cases, prevent hearing loss from progressing to the point where permanent damage occurs. Although hearing loss is not regarded as a life-threatening condition, it does become a severe threat to the essential quality of life indicators unless intervention occurs early during treatment. The adverse effects of hearing loss on cognitive-linguistic skills and psychosocial behavior are well documented, as also the serious vocational, social, and interpersonal consequences for the patient. Findings of the current study, although important, should be interpreted after taking the identified methodological limitations into account. The main limitations of the current study included, first, the nature of the HIV / AIDS disease and the population being studied precluded complete control over the confounding variables, which could have had an influence on the results such as interactions of ARVs with other therapies; especially traditional medicine in the form of '*ubhejane*,' which has been reported to be in widespread use.\[[@CIT32]\] Second, the sample size for the comparison group was small, thereby preventing randomized matching of participants in the comparison group with those in the experimental group. The small sample size of the comparison group was due to factors such as attrition, due to patients commencing treatment during the study, as well as loss to follow-up. Third, ultra-high frequency audiometry did not form part of the test battery and this may have influenced the type of results found, in that, clinical changes in the ultra-high frequencies depicted on the audiogram could have been missed. Finally, the length of time for which audiological monitoring occurred (six months) may have been too short to allow the clinical hearing loss, possibly caused by ART, to manifest, and therefore, be detected on the audiogram. Nevertheless, findings do justify an intense pursuit of the answer to the question: Does HAART *sound toxic?* The known effects of HIV / AIDS on the auditory system that have been reported in the literature are mainly based on cross-sectional studies and case reports conducted internationally in industrialized countries, with very limited information coming from third world countries where the presentation of the virus and its treatments may be different. Furthermore, because this evidence may not be viewed to be contextually relevant to the developing world, its incorporation into routine clinical assessment and management lags behind significantly. Hence, the need for categorizing the ototoxic effects of HIV / AIDS treatment, in an effort to ensure that ototoxicity monitoring protocols are established and implemented as part of the routine clinical management among infected patients. Research in ototoxicity in HIV / AIDS needs to be locally relevant, should include large sample sizes and longitudinal follow-up of cases, and should also utilize sensitive audiological test measures to improve the validity and reliability of the findings. **Source of Support:** Nil **Conflict of Interest:** None declared.

PubMed Central

2024-06-05T04:04:19.942664

2011-01-01

PMC3053513

Impressive progress has been made in detecting and imaging structural properties of biological systems. Structure data, however, represent only the first step toward an understanding of physiological processes. A deeper insight into the functions of biological macromolecules and their supramolecular assemblies requires additional information both on the interactions and on the dynamics governing their behavior. Nowadays, there is renewed interest in addressing the collective behavior of the biological system, shifting the focus from a detailed description of the single isolated molecule to the properties of assemblies of idealized simple objects. Such issues are typically tackled by bio-thermodynamics. At the variance of the classical thermodynamics, where the ultimate goal is the macroscopic properties of a single system (sometimes isotropic and macroscopically homogeneous, such as a liquid solution), biological phenomena involve a variety of multiple scale subsystems, each of them defined over a particular size and time scale. These subsystems, spanning from the angstroms to the micron, and from the pico-second to hours, are not isolated, but strongly interact with each other giving rise to new and challenging phenomena. In this review we focus on a typical collective system, the biological membrane, selected both for its fundamental role in cell biology and for the different, but closely connected, space and time scales. In order to be more specific, we list here, a few open questions in membrane science that could be answered only by considering a multiscale approach: Phase diagrams and phase transition kinetics in multicomponent lipid systems --- how do we combine observation and modeling of molecular rearrangements on \> 100 *nm* length scales during domain formation and / or phase transitions?Coupling between different fluctuating fields (e.g., shape and composition) --- how can continuum elastic theories, mean field models, and particle-based simulations be combined so as to capture membrane behavior from 1 *nm* to 10 microns?Cooperative phenomena in membranes --- how do membranes and proteins interact collectively in processes that span multiple length and / or time scales, for example, endocytosis?Active lipid transport and non-equilibrium membrane processes in live cells --- how is energy efficiently deposited into a membrane to drive processes such as raft domain formation, pore formation, vesicle fusion, membrane invagination, and protein activity?Hydrodynamic effects on membrane dynamics --- when are hydrodynamic effects indispensable in membrane dynamics, and how can their effects be quantitatively captured across different scales?Large-scale membrane remodeling events studied through a hierarchy of scales --- how do we connect single-molecule diffusion studies to the collective migration of lipid domains or patches?Cross-coupling between lipids and proteins --- Membranes move proteins and proteins reshape membranes: how do we systematically improve the minimal protein models and dynamics currently employed in coarse-grained simulations and parametrize them using atomistic modeling?Connecting single / multiple particle tracking experiments with nanoscale spatial resolution in living cells to the underlying collective membrane dynamics. What do such experiments reveal about membrane structure and dynamics?. The above-mentioned issues are tackled by a combination of theoretical / computational approaches and thermodynamic techniques. On the experimental side, excellent microcalorimeters and other techniques measuring heats, volumes, pressures, and related properties have been developed over the last decades and are now available to a broad spectrum of users. On the theoretical side, there is an explosion of analytical and computational techniques, which have shown potential usefulness in understanding the collective properties of model membranes. Besides the methods of classical and statistical thermodynamics, new ideas have been proposed, for instance: the theories of phase transitions, the different approaches dealing with out-of-equilibrium thermodynamics, the application of the continuum elasticity, and viscoelasticity theories to lipid membranes and so on. Also on the computational side, a variety of approaches have been suggested in the field of Molecular Dynamics and Dissipative Dynamics. They range from highly idealized coarse-grained pictures of lipids, proteins, and water, to complete simulations at an atomistic level. Simulations are gaining broader and broader applications because they provide, with a steady increasing level of accuracy, information on both the structural details (geometry) and the collective property of the system (e.g., lipid order parameter, total energy, and bilayer elastic constants). This review is mainly directed to researchers working in the field of lipid membranes in biological as well as model (e.g., vesicle) systems.\[[@CIT1]--[@CIT3]\] It aims at providing an overview of the thermodynamic techniques and of the physical principles behind the investigated systems. The broad scope of the review makes it impossible to explain the thermodynamic background or technical details of the methods\[[@CIT4]--[@CIT9]\] or to discuss the results obtained by using them. Instead, the article must be limited to making one aware of the calorimetric assays that are available to tackle a certain problem and to giving a few selected references. One current trend in membrane thermodynamics seems to be the consideration of increasingly complex systems. Vesicles of uncharged DMPC or DPPC (dimyristoyl- and dipalmitoyl-phosphatidylcholine) have yielded important information, but there are many other problems for which these lipids are rather poor model systems. For instance, lipid vesicles made up of ionic lipids and / or pH-modulated vesicle surface potential may represent a useful tool in mimicking the surface potential of real cell membranes. Furthermore, the great interest in lipid rafts has led to a much broader consideration of complex mixtures of glycero-, sphingo-, and glycolipids and sterols. Calorimetry of biological membrane extracts, viruses, organelles or whole cells is being further developed. Another important development is the ongoing introduction of new instruments, techniques, and assays. The crucial challenge is to combine insights from biochemistry and physiology with those from structural biology and from bio-thermodynamics to derive an integral picture of membranes and their functions. The great amount of experimental data must be interpreted on the basis of approximate, but not over-simplified, models. This issue is too large and cannot be contained in the space of a review. We will mention only the main ideas behind the various thermodynamic models developed to investigate the membrane properties. Brief Survey of the Main Thermodynamic Techniques {#sec1-1} ================================================= Calorimeters measure the heat consumed or released by a sample on re-equilibration after a perturbation. Such perturbations can be caused by a change in temperature (Differential Scanning Calorimetry), addition of material (Isothermal Titration Calorimetry), a change in pressure (Pressure Perturbation Calorimetry) or in water activity (sorption calorimetry). For a comparison between different types of calorimeters, such as adiabatic, heat flow, or power compensation instruments, see Wadsö\[[@CIT4]\] and Höhne *et al*.\[[@CIT10]\] Briefly, the fast response time of power compensation instruments makes them more sensitive for measuring the heat of fast effects and for revealing their kinetics. Heat flow calorimeters can provide better long-term stability of the temperature and baseline signal, which is particularly important if slow processes are investigated. Differential scanning calorimetry {#sec2-1} --------------------------------- For a detailed introduction to Differential Scanning Calorimetry (DSC), see Leharne *et al*.,\[[@CIT11]\] although an accurate description of the instrumental apparatus is reported in Privalov *et al*.\[[@CIT12]\] and Protnikov *et al*.\[[@CIT13]\] Briefly, DSC records the temperature-dependent isobaric heat capacity, *C~p~(T)*, of a sample. For first order (or weakly first-order) phase transitions, such as the bilayer gel to liquid-crystalline transition, the transition temperature, *T~m~*, is where the heat capacity, Cp, reaches its maximum value. The value of calorimetric enthalpy (∆*H~cal~*) for phase transition is determined by integrating the area under the peak $$\Delta H_{cal}\ = \ \int\ C_{p}dT$$ From these values, the entropy of the phase transition is determined: $$\Delta S\ = \ \frac{\Delta H_{cal}}{T_{m}}$$ Comparison of ∆*H~cal~*, ∆*S*, and *T~m~* shows the effect of structural modification (e.g., chain length or ion binding) on the thermodynamics of phase transition. However, unlike a simple organic compound's crystal to liquid melting transition, the phase transition in bilayers involves more than just the initial and final states. In fact, intermediate 'states' are formed during the transition, and a 'non-two-state' model is necessary for phospholipids in liposomes.\[[@CIT14]--[@CIT16]\] These intermediate states result from the formation of domains (e.g., disordered, mobile areas within the gel phase) before the phase transition temperature, and are due to the lateral movement of the phospholipids within the bilayer. The asymmetric shape of the DSC peak reflects the fact that a non-two-state transition has occurred. In order to adequately fit these data, a 'non-two-state' model is required. For any phase transition that occurs between two phases, A and B: A → B an equilibrium constant characterizes this process: $$K = \ \frac{a_{A}}{a_{B}}$$ where *a~A~* and *a~B~* represent the activities (concentration in ideal solutions) of each phase. The temperature dependence of the equilibrium constant is related to the enthalpy by the van't Hoff equation: $$\left( \frac{\partial\ln K}{\partial T} \right)_{P}\ = \ \frac{\Delta H_{vH}}{RT^{2}}$$ The van't Hoff enthalpy, ∆*H~vH~*, is equal to the amount of heat required for each cooperative unit to undergo the phase transition. The units are energy / cooperative unit. For a first-order two-state transition, the van't Hoff enthalpy is equal to the calorimetric enthalpy, ∆*H~cal~*. In other words, the heat effect for the transition A → B is the calorimetric enthalpy, which correspondingly governs the distribution between the two phases. If ∆*H~vH~* \< ∆*H~cal~* the process involves one or several intermediate stages, such as A → B → C, and is called non-two state. If ∆*H~vH~* \> ∆*H~cal~*, the process involves cooperativity, but is not 'completely cooperative' as in a first order transition. In other words, the distribution of molecules between the two phases is much more temperature-dependent than the actual heat effect of the phase transition, due to the cooperative motion of the molecules. Therefore, for a non-two-state transition or a partially cooperative transition there are two separate enthalpy parameters, ∆*H~vH~* and ∆*H~cal~*. After subtracting a baseline from the data, which negates any temperature dependence of ∆*H~cal~*, we use [equation (4)](#FD4){ref-type="disp-formula"} to obtain an expression to fit our data:\[[@CIT14][@CIT16]\] $$C_{p}\left( T \right) = \frac{K\left( T \right)\Delta H_{vH}\ \Delta H_{cal}}{\left( {l + K\ \left( T \right)} \right)^{2\ }RT^{2}}$$ where *K(T)* is just the equilibrium constant ([3](#FD3){ref-type="disp-formula"}), which is obtained as a function of temperature, by solving ([4](#FD4){ref-type="disp-formula"}) for *K(T)*: $$K\left( T \right)\ = \ \exp\left( {- \frac{\Delta H_{vH}}{RT}\left( {1 - \frac{T}{T_{m}}} \right)} \right)$$ The software of the DSC apparatus completes this fit and provides the values of ∆*H~cal~*, ∆*H~vH~*, and *T~m~*. For a more physical picture of the van't Hoff enthalpy, we note that ∆*H~vH~* can be calculated directly from the calorimetric data. First, the Cp versus T output scan from the calorimeter is integrated to form a plot of the enthalpy for the phase transition, ∆*H~cal~*. The maximum of *Cp* versus the *T* curve is *Cp* max. The van't Hoff enthalpy for the equilibrium is given by: ([8](#FD8){ref-type="disp-formula"}) $$\Delta H_{vH}\ = \ 4R{T_{m}}^{2}\frac{C_{pmax}\ \Delta H_{cal}}{\Delta H_{cal}}$$ A sharper transition results in a larger value of ∆*H~vH~*, as *Cp* max is larger. The sharpness of the transition can also be characterized by the full width at half-maximum, of the *Cp* versus *T* peak, ∆T~1/2~. Sharp transitions have a large ∆*H~vH~*, and correspondingly small ∆*T~1/2~*. As the units of ∆*H~vH~* are energy/cooperative unit, and those of ∆*H~cal~* are energy/mole, the ratio of the two (∆*H~vH~* /∆*H~cal~*) gives the value of the moles (or molecules) per cooperative unit: $$C.U.\ = \ \Delta H_{vH}\ /\ \Delta H_{cal}$$ The larger the value of C.U., the more cooperative the phase transition is. Therefore, cooperative phase transitions have larger ∆*H~vH~*. The value of ∆*T~1/2~* can be used as a qualitative measure of molecular cooperativity. Wider peaks correspond to less cooperative phase transitions. The concept of molecular cooperativity is used for proteins, to determine the number of subunits involved in a transition. The use of this concept for phospholipid bilayers is controversial, but the value of C.U. or ∆*T~1/2~* can give a relative measure of the cooperativity of the bilayer phase transition. Isothermal titration calorimetry {#sec2-2} -------------------------------- The Isothermal Titration Calorimetry (ITC) technique is based on a series of consecutive injections of a liquid sample (a few *µ*l each) from a syringe into the calorimeter cell under isothermal conditions. The heat of the reaction is measured as a function of the injection number, that is, it depends on the concentration of the injectant in the cell. The term 'reaction' describes any transition of molecules between different chemical or physical states (including those involving mass transfer inside the solution). Considering that the injection causes ∆*N^Tr^* moles of a compound to undergo a transition accompanied by a molar enthalpy change of ∆*H^Tr^*, therefore, the measured heat q is the sum of the enthalpy changes of all n processes induced by the injection: $$q - q_{dil}\ = \ \sum\limits_{n}\Delta{N_{n}}^{Tr}\Delta{H_{n}}^{Tr}$$ *q~dil~* denotes the heat of dilution that occurs due to changes in intermolecular interactions of the injectant and of the cell content. These effects are determined by blank runs injecting the titrant into the buffer inside the cell and are eliminated by subtracting the resulting heats. It is often convenient to work with normalized differential heats, Q, which are given per mole of titrant, ∆*N^Inj^*. In the simple case that only one heat-producing (or adsorbing) process occurs (*n* = 1), we find: $$Q - Q_{dil}\ = \frac{q - q_{dil}}{\Delta N^{Inj}}\ = \ \frac{\Delta N^{Tr}}{\Delta N^{Inj}}\Delta H^{Tr}\ = \frac{\Delta c^{Tr}}{\Delta c^{Inj}}\Delta H^{Tr}$$ where ∆*C^Tr^* specifies the moles per cell volume that undergo heat-producing transition, and ∆C^Tr^ denotes the change in the concentration of the injectant in the cell caused by the injection. To evaluate the ITC curves, one has to derive a model for the process under investigation that relates ∆*C^Tr^* to the known total concentrations of all compounds in the cell and a few adjustable parameters. Different types of assays can be performed, we refer to the specialized literature for a complete description of these experimental approaches.\[[@CIT8][@CIT17]--[@CIT23]\] Pressure perturbation calorimetry {#sec2-3} --------------------------------- Different calorimeters have been designed for measurement of the heat accompanying an isothermal pressure change, *dQ* / ∂*p*\|*~T~*. Such techniques have been referred to, for example, piezothermal analysis,\[[@CIT24]\] scanning transitiometry,\[[@CIT25][@CIT26]\] pressure jump calorimetry\[[@CIT27]\] or Pressure Perturbation Calorimetry (PPC).\[[@CIT7][@CIT28]\] A related, adiabatic technique has been termed as volume perturbation calorimetry.\[[@CIT29]--[@CIT32]\] PPC is mainly used to determine the temperature-dependent, isobaric volume expansion of a sample, *dV / ∂T/~p~*. This approach is based on the Maxwell relation of the reversible heat exchange on a change in pressure, *∂Q~rev~ /∂p* at constant temperature, *T*, to the temperature-induced volume change, *∂V / ∂T*, at constant pressure, *p*: $$\frac{\partial Q_{rev}}{\partial p}\left| {\frac{}{T} = - T\frac{\partial V}{\partial T}} \right|_{P}$$ Over many years, mainly bulk liquids or solutions were studied on home-built, heat flow calorimeters, mostly using high pressures. Of late, a new generation of PPC instruments have become commercially available as accessories to highly sensitive scanning calorimeters of the power compensation type. The extremely high sensitivity of the calorimeter makes it possible to study changes in the partial volume of as little as ≈1 mg of a protein using only very small pressure jumps of five bars. The first applications of the technique to lipids were studies on the kinetics of phase transitions, on the basis of the relaxation of the temperature or heat changes following a pressure variation (see section Kinetics Phenomena). Volumetric investigations were performed characterizing lipid melting\[[@CIT33]--[@CIT36]\] and domain formation in membranes.\[[@CIT37]\] Water sorption calorimetry {#sec2-4} -------------------------- Different calorimetric techniques have been applied to characterize the enthalpy and free energy of water binding to hygroscopic materials. In all the instruments a lipid film is deposited on the wall of a cell exposed to an atmosphere of varying water vapor activity (the relative humidity, RH). An increase in gas humidity gives rise to an exothermic heat that depends on the molar enthalpy of adsorption from vapor, $\mathrm{\Delta}H_{W}^{vap\rightarrow B}$, and the mole number of adsorbed water molecules, $$q = \Delta{N_{W}}^{vap\rightarrow B}\Delta{H_{W}}^{vap\rightarrow B}$$ The adsorption of vapor to the membrane is exothermic $\mathrm{\Delta}H_{W}^{vap\rightarrow B}\ < 0$ and includes: (i) the enthalpy of condensation of water, $\mathrm{\Delta}H_{W}^{liq\rightarrow B}\ < 0$ = --40.6 kJmol^-1^, (ii) a much smaller enthalpy of binding of liquid water to the bilayer, $\mathrm{\Delta}N_{W}^{vap\rightarrow liq}$ : $$\Delta{H_{W}}^{liq\rightarrow B}\ = \Delta{N_{W}}^{vap\rightarrow B}\ - \Delta{H_{W}}^{vap\rightarrow liq}$$ Thus, from ([12](#FD12){ref-type="disp-formula"}) and ([13](#FD13){ref-type="disp-formula"}) one can estimate $\mathrm{\Delta}N_{W}^{vap\rightarrow liq}$ from independent measurements of *q* and $\mathrm{\Delta}H_{W}^{liq\rightarrow B}\ < 0$. There exist different techniques to measure the heat *q* and the amount of adsorbed water Molar volumes and dilatometry {#sec2-5} ----------------------------- Direct measurement of the lipid molar volumes and / or their variation with temperature (the thermal expansion α=*V^-1^ dV/∂T~p~* coefficient at constant pressure). This relevant parameter α can be also measured by PPC as discussed in section *Pressure perturbation calorimetry*. Results of molar volume are routinely accurate to the 0.1% level with very good agreement obtained by different researchers using different instrumental approaches. Since the pioneering studies by Nagle *et al*.\[[@CIT44]\] a consistent number of studies have addressed this relevant topic. Among these studies (often performed by integrating the density measurements with other structural techniques) we mention the still debated problem of lipid-cholesterol mixtures,\[[@CIT45]\] the nature of the gel to fluid phase transition,\[[@CIT46]\] the salt effect on the membrane density,\[[@CIT47]\] and the undulated phase (ripple phase) appearing before the main melting transition,\[[@CIT48][@CIT49]\] the lipid-protein interaction,\[[@CIT50]\] just to quote a limited number of interesting issues. Static and dynamic volume compressibility {#sec2-6} ----------------------------------------- In pseudo two-dimensional systems, such as the Langmuir-Blodgett films spread at the water--air interface, lateral compressibility measurements represent the most employed tool used to investigate molecular monolayers. On the contrary, compressibility measurements have been far less used in studying lipid bilayers. Volume compressibility of lipid membranes can be measured by ultrasonic velocity techniques. Briefly, the speed of sound in lipid dispersion depends on the combined compressibility of water and lipid membranes. Thus, in ultrasonic resonators one can calculate the volume compressibility from the wave-length of a standing wave. Consider a membrane being compressed at constant temperature. This means that the heat released on compression is adsorbed and transferred by the surrounding water molecules. For lipid vesicles in an aqueous environment, such a condition is fulfilled if compression is applied very slowly (much slower than the relaxation processes within the membrane), otherwise the measured compressibility is termed as adiabatic compressibility, $$\mathbf{K}\frac{V}{S}.$$ The hydrostatic pressure change, ∆*p*, in the liquid, is proportional to the relative volume change ∆*V / V~o~* $$\Delta p = K_{v}\left( \frac{\Delta V}{V_{deg}} \right)$$ where *K~v~* is the module of compression. On the other hand, the isothermal compressibility $\mathbf{\kappa}_{T}^{V}$ is defined as: $\mathbf{\kappa}_{T}^{V}\ = \ V_{o}^{- 1}\left( {\partial V/\partial p} \right)_{T}$, thus: $\mathbf{\kappa}_{T}^{V}\ = \ 1/K_{V}$. The adiabatic compressibility $\mathbf{\kappa}_{T}^{V}$ is simply related to the measured sound velocity c by the relationship $$\mathbf{c}\ = \ \sqrt{\frac{\mathbf{l}}{\mathbf{\rho}{\mathbf{K}_{S}}^{V}}}$$ where *ρ* is the sample density. [Equation (15)](#FD16){ref-type="disp-formula"} can be easily generalized in the case of a suspension (water + vesicle), enabling one to extract the bilayer compressibility by performing experiments at different water / vesicles ratios. Thermodynamics provide a useful link between isothermal $\mathbf{\kappa}_{T}^{V}$ and adiabatic $\mathbf{\kappa}_{T}^{V}$ compressibilities $${\mathbf{K}_{S}}^{V}\ = \ {\mathbf{K}_{T}}^{V}\ - \frac{T}{VC_{P}}\left( \frac{\partial V}{\partial T} \right)_{P}$$ where *C~P~* is the specific heat at constant pressure and *dV* / *∂T*\|*~P~* is the isobaric volume expansion of a sample. A useful property of the compressibility is its relationship to volume fluctuations: $${\mathbf{K}_{T}}^{V}\ = \ \frac{< \ V^{2}\ > \ - \ < \ V\ >^{2}}{< \ V\ > \ RT}$$ where \< *V* \> is the mean volume and \< *V^2^* \> -- \< *V \>^2^* is the mean standard deviation of the volume. Volume and area (see the next section) fluctuations are very sensitive to bilayer properties, for instance, they increase on decreasing the lipid chain length.\[[@CIT51]\] However, the most intriguing effect is the divergence of the compressibility at the phase transition. This issue will be discussed a little later. Measurements of isothermal and adiabatic compressibility have been performed on model\[[@CIT52]--[@CIT54]\] and biological\[[@CIT55]\] membranes, the results will be discussed later. Area compressibility {#sec2-7} -------------------- Strongly anisotropic systems, as lipid bilayers, show a different behavior, depending on whether the force is applied perpendicularly, parallelly or isotropically. Assuming that the energy cost for compression and extension of a membrane about the minimum energy configuration are identical (harmonic approximation), the energy ∆*G* associated with the lateral expansion (compression) area variation is: $$\Delta G\ = \ \frac{1}{2}\mathbf{K}_{A}\ \left( {dA\ /\ A} \right)^{2}$$ where *K~A~* is the area compressibility modulus and *dA / A* is the relative area variation. The force associated with the energy (18) is called membrane tension, τ=*K~A~ (dA/A)*. Direct measurement of *K~A~* is not simple. Nowadays a common method is used, based on the micropipette aspiration technique, developed by Evans and his associates.\[[@CIT56]\] Typical values of *K~A~* for lipids are in the range of 100 -- 200 dyn / cm, but larger values are found in lipid / cholesterol mixtures (for a 1: 1 PC / cholesterol mixture *K~A~ ≈ 800dyn/cm)*. Another elastic constant closely related to the area compressibility modulus described by [eq. 18](#FD19){ref-type="disp-formula"} is the bending elasticity modulus *K~M~*. Indeed, on bending the external leaflet of a lipid, the bilayer expands, while the inner leaflet is compressed and for weak deformations, the contribution of both modes is additive. Theoretical and experimental correlations between the two elastic constants *K~A~* and *K~M~* have been thoroughly investigated (from the standard theory of elasticity: *K~A~* /*K~M~* =*h^-2^*, where *h* is the bilayer thickness). Analogous to the isothermal volume compressibility discussed in 2.8, the isothermal area compressibility, $\mathbf{\kappa}_{T}^{A}\ = \ 1/K_{A}$, can also be related to the lateral density fluctuations of a lipid bilayer: $${\mathbf{K}_{T}}^{A}\ = \ \frac{< \ A^{2}\ > \ - \ < \ A\ >^{2}}{< \ A\ > \ RT}$$ Such an equation is noticeable and it will be used in section *Passive Membrane permeability*, while discussing the passive transport of lipid membranes. Application to Lipid Systems {#sec1-2} ============================ Properties of lipid bilayers {#sec2-8} ---------------------------- Measurable thermodynamic parameters of membranes in their different states (gel, sub-gel, ripple, fluid) are, in particular, the isobaric heat capacity, the thermal volume expansion, and the isothermal or adiabatic compressibilities. It is interesting to compare the thermodynamic properties of lipid membranes with those of the corresponding alkanes of the same length, in order to unravel the peculiar properties induced by bilayer ordering. For instance, absolute heat capacities of different lipid bilayers were determined by Blume\[[@CIT57]\] using DSC. He found that *Cp* depended strongly on the head group and chain length and the contribution per methylene group in most lipids was larger than in alkanes. The results were discussed in terms of contributions of hydrophobic hydration of the lipid tails to Cp. Furthermore, the thermal volume expansion coefficient of fluid membranes was typically about 10^-3^ K^-1^, a value close, but a little bit larger than that typical to organic solvents. It could be measured with great accuracy by static densitometry in a carefully thermostated heat bath. However, it could also be conveniently measured by PPC as discussed in section *Pressure perturbation calorimetry*. The method determined the volume changes by applying small pressure jumps, which were applied homogeneously to the whole sample. A comparison between the different techniques had been discussed recently.\[[@CIT58]\] However, the main difference between isotropic fluids and membranes is that the reduction of the partial volume of the lipid in a bilayer induced by an increase in pressure is highly anisotropic. As more ordered chains can be packed more tightly together, a relatively small reduction in the volume is accompanied by huge lateral area condensation. Consequently, the more ordered straight chains determine an increase in membrane thickness. This means that a typical reduction of the surface area of about 20 -- 25%, on going from the fluid to the gel phase\[[@CIT59]\] is accompanied by a volume decrease as small as 3%. As both bilayer volume (by densitometry or PPC) and thickness (by X-ray or neutron scattering) are available with a great accuracy, the surface area increment is easily calculated. This is an important result, because even subtle variations of the surface area of a lipid bilayer may have a dramatic impact on the morphology of a membrane. Thus, we can define (and measure) three different kinds of compressibilities: The volume compressibility (similar to that of the isotropic liquids),The area compressibility;The thickness compressibility. Volume compressibility can be easily measured by the techniques described in section *Static and dynamic volume compressibility*. The area compressibility of a bilayer is similar to that measured by lateral pressure measurements in monolayers spread at the water--air interface; there are, however, two main differences: (a) monolayers and bilayers are related, but in different systems; (b) expansion and compression of a lipid bilayer requires comparable energy spending (Hooke law); this is not generally true for it concerns monolayers that monotonously expand against the applied external pressure. For these reasons a direct experimental determination of the lateral compressibility of a lipid bilayer is extremely useful and it can be performed by the techniques described in section *Area compressibility*. The heat accompanying an area change of the membrane can be measured by ITC experiments injecting vesicles into a hypo- or hyperosmotic solution.\[[@CIT60]\] The osmotically driven uptake of water into the interior of the vesicles induces an elastic lateral stretching of the membrane, which is endothermic, while the lateral compression of the membrane in a hyperosmotic environment is exothermic. Thermotropic phase behavior of pure lipids {#sec2-9} ------------------------------------------ Lipid-water mixtures may assume a variety of geometrical structures depending on the nature of the lipids and on the lipid / water content. At high water content the most common structure is the planar lipid bilayer, where, in order to minimize the unfavorable energy associated with water exposure, the bilayer edge assumes an edge-free arrangement: the vesicle. Bilayers can form a large variety of phase structures as a function of chemical composition (including length, branching, and unsaturation of the chains and charge distribution of the heads), temperature, pressure (see below), hydration, and so on. Typical structures at low temperature are bilayers in different subgel, gel, and ripple phases. These phases have stretched acyl chains (i.e., in all-trans conformation) giving rise to wax-like properties. At the main transition or melting temperature, *T~m~*, the ordered phase (L~β~\') is transformed into the fluid phase (L~α~). Before the L~β~\' ⇔ L~α~ takes place, a phase characterized by undulations of the bilayer surface (the ripple phase) is usually observed, within a narrow range of temperatures. The nature of the reversible L~β~\' ⇔ L~α~ transition has been debated over decades. Recent combinations of several experimental techniques, supported by computer simulations, both at the atomistic and coarse-grained levels have shared some light on the detailed mechanism of this complex event that involves several correlated steps, where the final and most important effect is the sharp correlated increase of the entropy-favored gauche conformation of the hydrocarbon tails in respect to the number of ordered trans-conformations. The sharp increase of gauche conformations, however, is not homogeneous along the membrane plane: local patches of melted domains transiently coexisting with solid-like patches appear in the course of the melting event.\[[@CIT61]--[@CIT70]\] This is the reason for the experimental observation of a divergence of bilayer compressibility values, as discussed in sections *Static and dynamic volume compressibility* and *Area compressibility*. Finally, at an even higher temperature, different kinds of lipids form inverse hexagonal phase (H~II~). Sometimes, between the lamellar and the H~II~, lipids form liquid crystalline structures with an astonishing degree of geometrical complexity: the cubic phases. Their structure consists of two mutually interpenetrating, but separate, mesh works of water channels separated by a multiply connected bilayer wall of lipid molecules, organized on a three-dimensionally periodic cubic lattice. The stability of the H~II~ phase depends on several parameters, the main factors influencing the reversible lamellar-to- H~II~ transitions are summarized in [Figure 1](#F0001){ref-type="fig"}.\[[@CIT71]--[@CIT73]\] ::: {#F0001 .fig} Figure 1 ::: {.caption} ###### Factors influencing the liquid crystalline bilayer - hexagonal phase preferences of membrane lipids. (Adapted from\[[@CIT71]\]) :::  ::: Since the pioneering studies of Chapman and others,\[[@CIT74]\] the standard technique to monitor the phase transitions described above is DSC. Pure lipids usually have very sharp melting transitions with half-widths of the order of 0.05 K. As impurities tend to broaden the transition, the width can be considered as an indicator of purity. Strong membrane curvature in small vesicles as well as undulations or shape fluctuations in large unilamellar vesicles also broaden the transition, and may shift its maximum to (generally) lower temperature. In a similar way, variation in the solvent properties, mainly due to the presence of ions in the solution, may appreciably shift the phase transition temperature (even subtle variations such as the replacement of H~2~O by D~2~O may change the thermotropic behavior,\[[@CIT75]\]). Over the years, a wealth of lipid melting data has been collected and the effects of chain length, branching, and unsaturation, head group and backbone structure, asymmetry between the two hydrocarbon chains, deuteration or fluorination of the tails, chirality of the lipid molecule, and the like, on *T~m~* and ∆*H*, have been thoroughly studied and modeled. For extensive reviews of phase transitions in different lipid classes, see Koynova and Caffrey's reviews on glycerolipids,\[[@CIT76]\] phosphatidylethanolamines,\[[@CIT77]\] sphingolipids\[[@CIT78]\] and phosphatidylcholines,\[[@CIT79]\] phosphatidic acids,\[[@CIT80]\] and the lipidat data bank.\[[@CIT81]\] As an example, in [Table 1](#T0001){ref-type="table"} we report the transition temperature of some lipid bilayers as a function of chain length and unsaturation. ::: {#T0001 .table-wrap} Table 1 ::: {.caption} ###### Transition temperature (in °C) as a function of tail length and saturation. All data are for lipids with the same headgroups and two identical tails\[[@CIT82]\] ::: Tail Length Double Bonds Transition Temperature ------------- -------------- ------------------------ 12 0 -1 14 0 23 16 0 41 18 0 55 20 0 66 22 0 75 24 0 80 18 1 1 18 2 -53 18 3 -60 ::: Thermotropic phase behavior of lipid mixtures {#sec2-10} --------------------------------------------- Lipid mixtures can show a very complex thermotropic phase behavior, including eutectic or peritectic points or compound formation.\[[@CIT1][@CIT83]\] DSC is the standard method to establish phase diagrams, by detecting the onset and completion of thermotropic phase transitions. More sophisticated studies have modeled the complete DSC peak, yielding not only transition temperatures, but also thermodynamic non-ideality parameters, describing the interactions (or the associations) in the mixture.\[[@CIT84][@CIT85]\] A very intriguing and biologically relevant system is given by sterols. Molecules such as cholesterol can split the melting transition of phospholipid membranes into a sharp and broad component, suggesting a gradual de-mixing of the membrane. Cholesterol disrupts the lateral order of the gel phase (so), tends to order the liquid phase (ld), and at a higher cholesterol content, stabilizes a new phase, the liquid-ordered phase (lo). This lo phase exhibits both rapid transverse diffusion and translational disorder of the liquid-disordered phase (ld) and relatively orders lipid chains characteristic of the solid ordered phase (so). The overall topology of the obtained phase diagram for binary lipid-cholesterol mixtures has been shown to hold for a range of PC-lipids with both saturated and monounsaturated acyl chains,\[[@CIT86]--[@CIT91]\] including palmitoyl oleoyl phosphatidyl choline (POPC)-cholesterol mixtures.\[[@CIT88]--[@CIT91]\] Other sterols as lanosterol and ergosterol have also been found to promote acyl-chain order at high concentrations.\[[@CIT92]\] Comparative studies of these three sterols have been conducted and reveal, despite their structural similarities, differences in the effect of cholesterol, lanosterol, and ergosterol on the lipid bilayer properties.\[[@CIT93][@CIT94]\] Similar results have been obtained for other side chain-modified sterols.\[[@CIT95]\] Accurate deconvolution procedures give a correct phase diagram of sterols / phospholipids mixtures.\[[@CIT96]--[@CIT99]\] The properties of mixed bilayers described earlier may also have a deeper impact on other thermodynamic parameters, such as, molar volume and compressibility. For instance, it is known that cholesterol sharply increases the compressibility modulus of phosphatidylcholine bilayers,\[[@CIT100][@CIT101]\] which is accompanied by a rigidification of the chains, as seen by the structural determination of the lipid bilayer thickness. Barotropic phase behavior {#sec2-11} ------------------------- The fact that lipid phase transitions are accompanied by substantial volume changes implies the existence of pressure-induced phase transitions.\[[@CIT102]\] Such an effect is the rationale for the well-known adaptation of the lipid membrane composition to extreme pressure conditions observed in deep sea living organisms.\[[@CIT103][@CIT104]\] In recent times, PPC has become commercially available as another tool to detect lipid melting, which is accompanied by a peak in thermal expansivity. Interestingly, the PPC and DSC peaks of lipid melting exhibit, almost perfectly, the same shape,\[[@CIT33][@CIT34][@CIT37]\] suggesting that both the enthalpy and volume of the membrane are governed by the same molecular parameter, most likely the abundance of gauche isomers in the chains. For a more sophisticated discussion of the phenomenon, see Ebel *et al*.\[[@CIT33]\] The increase in partial volume of the lipid bilayers on chain melting is of the order of 3 -- 4%\[[@CIT33][@CIT34][@CIT105]\] and the area by about 25%. Interestingly, many phospholipids with saturated chains of various lengths share the same pressure dependence of the phase transition, *dT~m~* /*dp* ≈ 20 K kbar^-1^, suggesting that this is an intrinsic property of the trans-gauche isomerization of the chains. Hence, this parameter could serve to distinguish chain melting transitions from others. *dT~m~* /*dp* can be determined from a series of DSC scans at various pressures (yielding *T~m~ (p)*), or by comparing ∆*V* and ∆*H* obtained by PPC and DSC according to the Clausius--Clapeyron equation: $$\frac{dT_{m}}{dp}\ = \ T_{m}\ \frac{\Delta V}{\Delta H}$$ [Equation (20)](#FD21){ref-type="disp-formula"} can also be used to compute ∆*V* from the pressure-dependent measurements of *T~m~* and ∆*H*, using DSC. The sensitivity of a phase transition to pressure can be quantified in terms of the pressure-induced shift of the transition temperature, *dT~m~* /*dp*, or the volume change of the transition, ∆*V.* Both parameters are related to each other according to (20). Shifted transition temperatures of lipids under external pressure have been measured by DSC using pressures ranging from 5 bar to kilobars,\[[@CIT33][@CIT34][@CIT106][@CIT107]\] yielding *T~m~(p)* and *dT~m~ /dp*. Phase changes of samples have also been induced by pressure jumps at constant temperature (PPC, pressure calorimetry), yielding ∆V of the transition. An increase in pressure can induce a transition from an inverse hexagonal to a fluid lamellar phase (*dT~hex~ /dp* ≈ 40 K kbar^-1^\[[@CIT107]\]), the freezing of the fluid-lamellar to a ripple phase (≈20 K kbar^-1^ for saturated chains\[[@CIT33][@CIT34][@CIT106]--[@CIT108]\] and ≈14 K kbar^-1^ for DOPE\[[@CIT107]\]), and the pre-transition from the ripple to the lamellar gel phase (≈10-15K kbar^-1^,\[[@CIT33][@CIT34][@CIT108]\]). In recent times, Ichimori, Kaneshina, and other authors\[[@CIT109][@CIT110]\] investigated the transition from the pressure-induced transition to the interdigitated phase of phospholipids bilayers \[[Figure 2](#F0002){ref-type="fig"}\]. ::: {#F0002 .fig} Figure 2 ::: {.caption} ###### Formation of an intergitaded phase of a two-component lipid bilayer at high pressure :::  ::: [Figure 2](#F0002){ref-type="fig"} showing that lipids with two asymmetric hydrocarbon chains or mixtures of long and short lipids easily interdigitate in order to avoid vacancies within the lipid matrix. Finally, an extensive review of the above concepts and experiments but mainly focusing on the interesting issue of the protein behavior at high pressure has been recently reported in the literature.\[[@CIT111]\] Lipid hydration and lyotropic phase behavior {#sec2-12} -------------------------------------------- The interactions of the polar and apolar parts of the lipids with water are the driving force for the formation of different phases. Several calorimetric techniques quantify the interaction of water with lipids under different conditions and allow characterizing hydration phenomena in detail. Water sorption calorimetry determines the enthalpy and entropy of water binding at a given temperature as a function of water activity. It has recently provided valuable insight into the molecular origin of the so-called hydration force, which causes a strong, short-range repulsion between two hydrated (bilayer) surfaces\[[@CIT112]--[@CIT114]\] due to interfacial water ordering. The ordering of water molecules by lipid--water and water--water interactions, as well as the entropy gains arising from fluctuations in the membrane structure, have been discussed as the basis of hydration forces. For DOPC bilayers, sorption calorimetry showed that only one or two water molecules per lipid exhibit an exothermic binding at 25°C, that is, these are bound and ordered by specific interactions. The adsorption of the remaining water molecules onto the lipid molecules is endothermic and is therefore driven exclusively by an entropy gain. Hence, water bound to the lipid increases its motional and conformational freedom, and the resulting entropy gains must also be considered on the basis of the hydration force. This important conclusion is further supported by the sorption calorimetric studies of POPC\[[@CIT38]\] and a series of saturated lipids showing three to four enthalpically bound water molecules per lipid.\[[@CIT40]\] The thermodynamics of a lyotropic gel-to-liquid crystalline transition of POPC at low relative humidity have also been discussed on the basis of sorption calorimetry.\[[@CIT38]\] The enthalpy change accompanying a lyotropic lamellar-to-hexagonal transition depends on whether the lipid forms direct hydrogen bonds or not.\[[@CIT115]\] Another approach for determining the hydration pressure of lipid phases is to record the phase transition temperatures at different, well-defined, water contents, by DSC (see, for instance the pioneering work by Cevc and Marsh\[[@CIT116]\]). For more recent studies on this issue see the study by Pfeiffer *et al*.\[[@CIT117]\] By calculating the mean pressure among the planar neutral bilayers brought close to each other on account of water ordering at the membrane surface, they found that the shift in temperature of the gel-to-liquid crystalline transition of lipid membranes behaves as $$\Delta{T_{t}}^{hyd}\ = \ \Delta{T_{t\infty}}^{hyd}\ \tanh\left( \frac{n_{w}V_{w}}{\xi S_{L}} \right)$$ where *n~w~* is the number of bound water molecules, *V~w~* the volume of one water molecule, ξ the correlation length of water polarization (a measure of the decay of water orientation on going from the membrane surface to the bulk phase), *S~L~* the lipid area, and the $\mathrm{\Delta}T_{t\infty}^{hyd}$ transition temperature shift at limiting hydration (nw → ∞), which is independent of water content. tanh(χ) is the hyperbolic tangent function (tanh → (X) as x\<\<1 and tanh(χ) → 1 when χ → ∞). These theoretical results agree well with the DSC measurements performed at a controlled water content.\[[@CIT116]\] Finally, a characteristic number of lipid-bound water molecules, called 'unfreezable water', can be deduced from the enthalpy of water freezing / melting of a sample of well-defined water content.\[[@CIT118]--[@CIT120]\] Self-asso ciation of lipids {#sec2-13} --------------------------- The critical micelle concentration (c.m.c.) and enthalpy of micelle formation, ∆*H~mic~*, can be determined by titration calorimetric experiments. From measurements at varying temperatures, the heat capacity change, ∆C~p,mic~, is also derived. Unfortunately, such an approach has limited application in the investigating lipid vesicles, because typical membrane lipids have critical association concentrations in a range that is not accessible by the ITC. Studies have, however, been performed on shorter chain analogs like diacylphosphatidylcholines and lysophosphatidylcholines.\[[@CIT121]\] The results have been discussed in terms of group contributions to enthalpy and free energy of self-association and changes in the water-accessible surface area of lipids. Furthermore, they have shown that the alignment of the acyl chains in an aggregate gives rise to a significant change in enthalpy (but not in free energy) compared to the state in bulk hydrocarbon. This finding is also important for the interpretation of enthalpies of insertion of molecules into lipid membranes. Membrane Partitioning and Binding of Additives {#sec1-3} ============================================== Modification of the membrane phase diagram by solutes {#sec2-14} ----------------------------------------------------- Small molecules, drugs, peptides, and proteins are not in general readily soluble in the solid-like phase of the lipid bilayer due to their crystalline structure. They are much more soluble in the fluid-like phase. This leads to the well-known reduction of melting points, demonstrated in the early seventies by a number of authors.\[[@CIT122]\] This effect is known as the vant'Hoff freezing point depression. For example, the solubility of NaCl is high in water and low in ice. Thus, salt lowers the freezing point of water. This effect is due to the difference in mixing entropy of the ions in water and ice. For low solute concentrations and with reasonable assumptions of perfect miscibility of lipids and solute in the fluid-like phase and immiscibility in the solid-like phase, one arrives by classical thermodynamics at the well-known relation between melting point depression and solute concentration $$\Delta T_{m}\ = \left( \frac{R{T_{m}}^{2}}{\Delta H} \right)\log\ a_{w}\ \approx \ - \left( \frac{R{T_{m}}^{2}}{\Delta H} \right)X$$ where ∆*H* is the lipid-melting enthalpy of the lipid bilayer (about 35 kJ/mol for DPPC), *R* the universal gas constant, and *T~m~* the lipid-melting temperature (314.3 K for DPPC), a~w~ is the solvent activity related to the molar fraction of the impurity inside the membrane X by the relationship: logaw=log(*γ~w~(1-X)*). In the simplest approximation, the water activity coefficient log *γ~w~* =0, but other choices can improve the analysis (see in the following text). Several DSC data can be interpreted with the aid of [eq. (22)](#FD23){ref-type="disp-formula"}. There are, however, several points to consider: The solute is soluble both in the gel and fluid phase, the greater solubility occurring in the fluid phase. When the solute has the same solubility both in solid-like and fluid phase, then ∆*T~m~* =0.There is a solute partitioning between the water and the membrane (see the next section). The net effect is a decrease of the solute concentration, which now depends on the lipid / water ratio.Solute-membrane mixing is not ideal (see section Non-ideal mixing). In this case the solvent activity coefficient is no longer zero and it does depend on solute concentration: log*γ~w~≈ A=CX^2^* +\.... The shift of the melting temperature, ∆*T~m~*, with the solute concentration *X* assumes a typical parabolic shape, often observed in DSC experiments.\[[@CIT123]\]There are no structural variations in the bilayer structure. Solutes, for instance, may induce interdigitation among the tails of the lipid leaflets\[[@CIT124]\] or other morphological phase transitions toward non-planar shapes. Despite these serious limitations, the vant'Hoff-based picture of the depression of the freezing point can be useful in studying the solute-induced variations of the membrane transition temperature, because it is conceptually simple and straightforward for practical purposes. Improvement of [eq. (22)](#FD23){ref-type="disp-formula"} can be reached either by remaining in the realm of classical thermodynamics (for instance, by introducing a partition coefficient for the impurity between the melted and un-melted lipid phases as done by Inoue\[[@CIT125]\] or by introducing a phenomenological non-ideal mixing enthalpic term\[[@CIT123]\] and discussed by us later). Alternatively, one can shift toward a microscopic modeling of the lipid-impurity interactions as pioneered by Mouritsen *et al*.\[[@CIT126]\] The physics behind the solute-induced temperature shift of phase transitions is common also to other systems. For instance, statistical mechanical theories, similar to those employed to explain the shift of the main transition in lipid bilayers, were developed by Crothers and McGhee. They allow a simple interpretation and calculation of DNA melting curves (detected by DSC techniques) in the presence of ligands or proteins.\[[@CIT127]--[@CIT129]\] The partitioning of non-ionic solutes into membranes {#sec2-15} ---------------------------------------------------- Several techniques (e.g., radio-labeling or spectroscopic techniques) can be employed to investigate a solute binding to a membrane. Among them, ITC has become a standard method for characterizing ligand binding.\[[@CIT130]\] For this assay, a solution of a compound A filled into the cell is titrated with the solution of a different compound B loaded into the syringe. By making use of a proper model equation it is easy to fit the data obtaining the binding constant, *K~o~*, the molar enthalpy change, ∆*H*, and the stoichiometry of the reaction. This model is appropriate for binding the ligands to the receptors residing in the membrane. Similarly, solute partitioning into membranes can be studied very favorably by different types of ITC assays. The process giving rise to the heat, *Q*, is the transfer of solute (S) molecules from the water (w) to the lipid bilayer (L), which is accompanied by a molar enthalpy difference, $\mathrm{\Delta}H_{S}^{w\rightarrow L}$. In a similar manner, one can measure the release of a solute from the bilayer, accompanied by an enthalpy change $\mathrm{\Delta}H_{S}^{L\rightarrow w}\ = \ - \mathrm{\Delta}H_{S}^{w\rightarrow L}$. Hence, the transferred concentration ∆*c^Tr^* in [equation (10)](#FD10){ref-type="disp-formula"} has to be replaced by the change in concentration of the bilayer-bound solute, ∆*C~S~*, derived on the basis of a lipid / water partition coefficient *K~o~*. A variety of definitions have been used for the partition coefficient; for a detailed discussion.\[[@CIT131]\] A good description of the partitioning of amphiphiles is often possible in terms of a constant mole ratio partition coefficient, K~o~, obtainable by standard thermodynamic arguments: $$K_{deg}\ = \frac{C_{S}}{C_{L}{\overset{-}{C}}_{S}} = \frac{C_{S}}{C_{L}\left( {\ {\overset{-}{C}}_{S}\ - C_{S}} \right)}$$ where the symbols *C~S~* and *C~L~* denote the molar concentrations of solute dissolved in the lipid bilayer and that of lipids, respectively, (virtually, lipids are completely located in the bilayer because of the extremely small CMC of most lipids). ${\overset{-}{C}}_{S}$ denotes the molar concentration of the solute dissolved in water and is related to its stoichiometric concentration. Most ITC partitioning assays are based on injections of lipid vesicle suspensions into the calorimeter cell. For the uptake protocol,\[[@CIT132]--[@CIT135]\] the cell contains the buffer-dissolved solute, so that every aliquot of lipid vesicles injected into the cell binds a fraction of the remaining free solute. The release protocol\[[@CIT136][@CIT137]\] is based on small injections of lipid vesicles containing solute into a large excess volume of buffer; the dilution gives rise to a release of solute from the bilayers. A model equation that allows one to fit the uptake data has been derived, resulting in $$Q = K_{deg}\frac{C_{S}}{\left( {l + KC_{L}} \right)^{2}}\Delta{H_{S}}^{W\rightarrow L}\ + Q_{dil}$$ Similar equations can be obtained for what concerns the release protocol. The above model assumes that *K~o~* and $\mathrm{\Delta}H_{S}^{w\rightarrow L}$ are independent of solute and lipid concentration. In many cases, the solute mixes non-ideally with the lipid (see section Non-ideal mixing), therefore these assumptions are not a priori warranted, more refined models allowing for composition-dependent *K~o~* and $\mathrm{\Delta}H_{S}^{w\rightarrow L}$ have been used,\[[@CIT138]--[@CIT140]\] but in most cases the experimental data do not justify the introduction of other adjustable parameters (such as a non-ideality parameter). However, it must be noted that a two-parameter model, [equation (24)](#FD25){ref-type="disp-formula"}, yields good data even if the model assumptions are not strictly fulfilled. For a more detailed discussion and partitioning data for many systems, see articles on membrane binding of peptides,\[[@CIT133][@CIT141]\] surfactants,\[[@CIT131][@CIT142][@CIT143]\] alcohols,\[[@CIT144]--[@CIT147]\] and drugs.\[[@CIT148][@CIT149]\] The knowledge of the partition coefficient enables one to calculate the apparent standard chemical potential change of a solute, on transfer from water into the lipid bilayer, $\mathrm{\Delta}u_{S}^{0,w\rightarrow L}$, which is obtained as $$\Delta{\mathbf{u}_{S}}^{0,\ w\rightarrow L}\ = - RT\log\left( {K_{{deg}\ }C_{W}} \right)$$ with the water concentration in dilute solutions, *C~W~* = 55.5 M.\[[@CIT131]\] The contribution to $\mathrm{\Delta}u_{S}^{0,w\rightarrow L}$ that arises from the hydrophobic groups that are buried in the apolar core of the membrane is similar to that obtained on self-association with micelles (≈3 kJ mol^-1^ per methylene group). It is worth mentioning that the enthalpy and heat capacity changes on membrane insertion are quite different from those of micelle formation, indicating that changes in lipid packing caused by the solute may have substantial consequences. Application of these studies is the understanding of well-known, but still elusive issues. One of them is anesthesia, a phenomenon caused by a number of small molecules that partition in the biological membrane. For more than a hundred years it is known that the effectiveness of anesthetics is proportional to their solubility in olive oil (that has the properties of the membrane interior). This rule is known as the Meyer-Overton rule.\[[@CIT150]\] It holds over several orders of magnitude ranging from laughing gas, N~2~ O, over halothane to lidocaine. Even the noble gas xenon is an anesthetic. This observation excludes any specific binding to macromolecules (e.g., proteins) if one is searching for a generic explanation of anesthesia. It has also been known for a long time that anesthetics cause a lowering of phase transition temperatures.\[[@CIT151][@CIT152]\] There are strong indications that the effect of anesthetics is related to this finding. As shown in section *Barotropic phase behavior*, phase transitions are pressure-dependent. Even as pressure increases transition temperatures, anesthetics lower them. It has in fact been found that pressure reverses the effect of anesthesia.\[[@CIT153]\] Membrane binding of small or large charged solutes {#sec2-16} -------------------------------------------------- For charged solutes one has to take into account that the aqueous concentration of the solute is in the vicinity of the membrane, which is in equilibrium with the membrane-bound solute, and differs from that in the bulk solution, ${\overset{-}{C}}_{S}$. The apparent partition coefficient, *K~app~*, strongly depends on the electrostatic potential of the membrane surface with respect to the bulk, *ψ~o~*, and the charge number of the solute, *Z~S~*: $$K_{app}\ = \frac{C_{S}}{C_{L}{\overset{-}{C}}_{S}\ } = K_{deg}\ \exp\left( {- \frac{z_{S}e\psi_{deg}}{K_{B}T}} \right)$$ with e and k~B~ denoting the elementary charge and the Boltzmann constant, respectively. The potential ψ~o~ depends, in turn, on the ionic strength and the bound solute. It can be determined on the basis of the Gouy-Chapman theory, which relates *ψ~o~* to the solution ionic strength, dielectric permittivity e of the solvent, and electrical surface density σ (number of charged lipids / lipids area), through a relationship derived from the electroneutrality condition of the whole system: *ε(∂ψo/∂z)* = σ (the derivative being performed with respect to the z-axis perpendicular to the surface). In case of the weak surface potential, the above charge density-potential relationship yields: *ψO* = *σ/εκ*, where *κ = (2e^2^ c/(ek~B~T)^1/2^* is the Debye constant, proportional to the salt concentration, c; more general complex relationships between σ and *ψ~o~* valid at high potentials can be derived as well. If the intrinsic partition coefficient, *K~o~*, which does not depend on electrostatics is known,\[[@CIT133][@CIT154][@CIT155]\] measurements of K~app~ can give information on ψ~o~ (or σ) and vice versa. The thermodynamics of ionization of a lipid on NaOH addition\[[@CIT156]\] and the ion adsorption to lipid bilayers\[[@CIT157]\] were studied by ITC. A variation in the buffer used in ITC partitioning or binding experiments can be used to reveal protonation--deprotonation effects accompanying ligand binding to membranes. This approach is based on the fact that the protons released or bound by the ligand are absorbed or provided by the buffer, respectively, so that the heat of ionization of the buffer contributes to the measured heat of titration. As the protonation heats of many buffers are known,\[[@CIT158]\] the apparent heat of binding in different buffers can be plotted versus the heat of buffer protonation, yielding the change in protonation and the intrinsic heat of binding.\[[@CIT133][@CIT154][@CIT159]--[@CIT163]\] It should be noted that the assumption of a constant average membrane surface potential (the Gouy--Chapman Theory) is an approximation leading to good results in most cases. Nevertheless, the local potential may be different, in particular for ligands that carry many charges. However, even for single-charged ligands the assumption of a constant potential fails at high surface coverages. As the surface potential ψ~o~ is proportional to the effective surface charge density *σ : σ = σ~o~* (1-ZC~S~), with σ~o~ the charge density at zero coverage, Z the ligand charge, and *C~S~* its surface concentration. We conclude that the effective binding constant defined by [eq. (26)](#FD27){ref-type="disp-formula"} does depend on concentration C~S~. This effect is stronger on increasing the ligand's net charge Z. Charged ligands such as polyelectrolytes or peripheral proteins exposing many positive charges toward the membrane surface may accumulate negatively charged lipids in a mixed membrane of anionic and zwitterionic lipids. Such effects have, for instance, been discussed in detail on the basis of ITC data on cytochrome C\[[@CIT164]\] and annexin/Ca^2+^.\[[@CIT165]\] Further indirect evidences for the accumulation of charged lipids in mixed membranes are based on DSC measurements. Some examples were described by the authors.\[[@CIT166]--[@CIT168]\] In a similar manner, the binding of DNA to membranes containing cationic lipids has been characterized by ITC, revealing the thermodynamic parameters of the entropy-driven interaction as well as critical charge ratios\[[@CIT169]\] and protonation effects.\[[@CIT163]\] The effect of the interaction is also evident on the DSC curves that are shifted and split on DNA addition.\[[@CIT169]\] For a recent comprehensive thermodynamic analysis of macroions 'decorated' lipid bilayers see the study of May.\[[@CIT170]\] The Effects of Additives on Membrane Properties {#sec1-4} =============================================== Non-ideal mixing {#sec2-17} ---------------- The free energy of mixing in fluid membranes is often close to the ideal value, as enthalpic and entropic interactions balance each other to a considerable extent (see next section on phase separation). The enthalpy of mixing is, therefore, a much more sensitive parameter for investigating the non-ideal mixing behavior of membrane constituents. We may write the mixing enthalpy *h* of a two-component system, A and B, with X denoting the mole fraction of B, as $$h\ = \ N_{A}H_{A}\ \left( 0 \right)\ + N_{B}H_{B}\left( 1 \right) + \left( {N_{A} + N_{B}} \right)H_{EXC}\ \left( x \right)$$ *H~B~* (1) and *H~A~* (0) stand for the molar enthalpies of B and A in pure systems. In the simplest approximation $$H_{EXC}\ \left( x \right) = \frac{1}{2}Z\left( {w_{AA}X^{2}\ + \ 2w_{AB}X\left( {1 - X} \right) + w_{BB}\left( {1 - X} \right)^{2}} \right)$$ where *z* is the number of contacts around each molecule and *w~ij~* represents the pair interaction energy between the i-th and j-th molecules. For ideal mixing, *h* is just a linear combination of the enthalpies of the lipid and solute, and the excess enthalpy, *H~EXC~(X)*, vanishes. Non-ideal mixing is represented either by *H~EXC~(X)\<0* if the A--B contacts are enthalpically favorable, or by *H~EXC~(X)\>0* if A--B mixing is enthalpically unfavorable. If the pure components are in only one state (we neglect, for instance, the partitioning effects or the occurrence of micellization), *H~B~* (1) and *H~A~* (0) are independent of the absolute concentration. The normalized heat, *Q*, of the injection of a pure component (A or B) into the mixture was derived in Heerklotz *et al*.\[[@CIT171]\], yielding $$Q = \left( {1 - X} \right)\frac{dH_{EXC}\left( X \right)}{dX} + H_{EXC}\ \left( x \right)$$ with *X* denoting the mole fraction of the injectant (B) in the mixture. Hence, the heats measured on titration of the solute into the lipid and those measured on titration of the lipid into the solute can be used to derive the same excess enthalpy function, *H~EXC~(X)*, by solving [equation (29)](#FD30){ref-type="disp-formula"}. This was done for a series of lipid--detergent systems.\[[@CIT171]\] As one might expect, bilayer-forming additives show small non-ideality effects in lipid bilayers, but micelle-forming solutes mix highly and non-ideally with lipids in membranes, *H~EXC~(X)*\> 0. It is important to stress the difference between the state of the system (characterized by *H~EXC~(X)*) and the heat *Q* representing the partial molar enthalpy. Positive values of *Q* do not necessarily mean that the mixing is unfavorable, but only that the addition of a compound renders the enthalpy of mixing less favorable. The enthalpy of mixing in a membrane can also be studied through a detailed analysis of the shape of the DSC curves of lipid transitions. Studies of lipid mixtures have been discussed in section *Lipid mixtures*. The thermodynamic analysis of the melting point depression with the concentration of an additive inside the membrane developed in section *Modification of the membrane phase diagram by solutes* can be extended to the case of non ideal lipid-additive mixing. A simple thermodynamic calculation gives the leading terms: $$\Delta T_{m} = - \left( \frac{R{T_{m}}^{2}}{\Delta} \right)X\left( {l - wX} \right)$$ where *w ≡ z(w~AA~* +*w~B~* - *2w~AB~*) is the non-ideal mixing parameter, while the other terms have been defined in [eq. (22)](#FD23){ref-type="disp-formula"}. The physical origin of non-ideal mixing is strictly related to the molecular structure of the binary bilayer single components. Differences in the hydrocarbon chain length, nature, and charge of the head groups and ion adsorption by a specific lipid component\[[@CIT172]\] may dramatically change the mixing behavior. Similar investigations have also been performed for lipid--protein membranes containing, for example, bacteriorhodopsin,\[[@CIT173]\] cytochrome C,\[[@CIT164][@CIT174]\] gramicidin A,\[[@CIT175]\] glycophorin,\[[@CIT176]\] and tetanus toxin.\[[@CIT177]\] Some simple cases of peptide mixtures with lipid membranes were discussed by Ivanova *et al*.\[[@CIT27][@CIT178]\] For a recent review of the DSC data on lipid-protein mixtures see, e.g., Lewis and McElhaney.\[[@CIT179]\] As a rule of the thumb, it can be shown that the influence of peptides or proteins on the specific heat profiles is mainly due to the miscibility of the peptide with the two lipid phases, gel and fluid, respectively. If, for instance, the peptide mixes well with the fluid phase (low nearest neighbor interaction energy) and does not mix well with the gel phase (high nearest neighbor interaction energy), the peptide will homogeneously distribute in the fluid phase, but aggregate in the gel state. The corresponding heat capacity profile will be shifted to lower temperatures and display an asymmetric broadening at the low temperature side of the transition.\[[@CIT27]\] Lateral phase separation: Different routes to domains formation {#sec2-18} --------------------------------------------------------------- The problem of whether molecules mix randomly or tend to form clusters of certain compositions or arrangements is governed by the excess free energy defined as the difference between the energy of the mixed state and that of the pure components: *G~EXC~* = *H~EXC~ - T∆*S*~MIX~*. *H~EXC~* has been defined by [eq. (28)](#FD29){ref-type="disp-formula"}, while the mixing entropy ∆*S~MIX~(X)* reads: ∆*S~MIX~(X) = -k(N~A~* log *N~A~* + *N~B~* log *N~B~*) = *kN (X* log *X* + (1-*X*) log (1-*X*)), with *k* being the Boltzmann constant and *N* the total number of molecules. Many lipid-additive systems showing non-ideal enthalpies of mixing can nevertheless be well described as randomly arranged mixtures, as the endothermic enthalpies of interaction are essentially balanced by the gains in mixing entropy. The fact that many additives exhibit a virtually constant mole ratio partition coefficient into lipid bilayers,\[[@CIT133]\] implies slightly unfavorable excess free energies of *G~EXC~* ≤ 0.4 kJmol^-1^. However, this non-ideality does not give rise to significant deviations from random mixing, because *G~EXC~* is small compared to the thermal energy (≈2.5 kJmol^-1^ at room temperature). Combining the explicit expression for the mixing enthalpy *h*, eqs.(27,28), with that of the mixing entropy reported earlier, we calculate the excess free energy as a function of the mixture composition *X*, its plot is given in [Figure 3](#F0003){ref-type="fig"}. ::: {#F0003 .fig} Figure 3 ::: {.caption} ###### Variation of the excess free energy against the composition *X* of a fluid binary mixture. The curves have been calculated for increasing values of the non-ideal mixing parameter *w/kT*. The first curve on the bottom corresponds to w/kT = 0 :::  ::: It can be easily seen that, depending on a single parameter *w/kT* ≡ *z(w~A~* +*w~B~--2w~AB~)/kT*, either a single minimum or two minima separated by a maximum can be observed. The position of the minima (and maxima) is calculated by imposing: ∂*G~EXC~(X)/∂X*=0, from which we get: $$\log\frac{X}{l - X} = \frac{w}{kT}\left( {l - 2X} \right)$$ (extension to mixtures of molecules differing in size is straightforward). Their locus as a function of *T* define the equilibrium curve (binodal curve, full line), which separates the one-phase and two-phase regions as reported in [Figure 4](#F0004){ref-type="fig"}, panel B. Furthermore, unstable regions of negative curvature (∂*G~EXC~(X)/∂X^2^* \<0) lie within the inflection points of the curve ∂*G~EXC~(X)/∂X^2^* =0, which are called the spinodes. Their locus as a function of temperature defines the spinodal curve reported in [Figure 4b](#F0004){ref-type="fig"} (dashed line). As we shall see shortly, the difference between the binodal and spinodal curve has a key influence on the final morphology of the phase separated system. ::: {#F0004 .fig} Figure 4 ::: {.caption} ###### Panel A : variation of the temperature scaled excess free energy against the composition X of a binary fluid mixture. Panel B : Phase diagram of a binary fluid. The continous curve is the locus of the minima of panel A (binodal curve, ∂G~EXC~(X)/∂X=0), while the dashed curve is the locus of the inflexion points (spinodal curve, ∂G~EXC~(X)/∂X^2^=0). To is the composition-dependent temperature at which phase separation takes place :::  ::: Although thermodynamics fixes the conditions for phase separation to occur in a rather transparent manner, the mechanisms leading to phase separated samples and their final morphology require a combination of purely energetic considerations, together with a dynamic picture of the whole process. The two most common pathways to phase separation are nucleation mechanisms or spinodal decomposition mechanisms. Starting from a point inside the one-phase region, a change in any physical parameter (e.g., temperature) brings the system inside the two-phase region of phase in [Figure 4](#F0004){ref-type="fig"}. In the region, enclosed between the binodal and spinodal curves, phase separation occurs through the nucleation and growth mechanism, a process controlled by the undercooling temperature. As discussed later, nucleation is a slow mechanism because it requires to overcome an undercooling-modulated energy barrier. A deeper cooling beyond the spinodal curve, brings the fluid to a phase separated structure through a mechanism that involves the increase of the composition fluctuations inside the binary fluid. The final morphology associated with the two processes is different: The nucleation mechanism leads to spheroidal isolated droplets richer in one component.Spinodal decomposition leads to inter-connected domains of different compositions. These different topological patterns may bring to noticeable different in the membrane structure and function. Consider, for instance, the lateral diffusion of a tracer in patterned A-rich and B-rich membrane domains and assume for simplicity that diffusion takes place only in the B-rich domains. If the B-rich domains are disconnected (i.e., they come from a nucleation and growth mechanism) the diffusant remains trapped inside these micro-pools and cannot reach the target. On the contrary, in a connected structure of A-rich and B-rich microdomains the diffusion is slow, but finite: the diffusant may reach the target. Domains formed by spontaneous de-mixing of lipids in a membrane have recently become a focus of interest, as such domains in biological membranes, referred to as 'lipid rafts', are believed to have important biological functions.\[[@CIT180]--[@CIT183]\] Calorimetric techniques, and in particular DSC, provides a useful tool to detect the formation of laterally heterogeneous structures within the lipid membrane. Consider a two-component lipid membrane that contains micro-domains richer in one component. If the domains are large and stable on the time scale of the main lipid transition, the response to constant heating (or cooling) will be markedly different: the melting temperature of each domain will be similar to that of the more abundant component. This fact leads to a broadening or even the splitting of the calorimetric peak, which is proportional to the composition of the domain and to the properties of the separated lipid components (temperature transitions and enthalpy of pure components). By exploiting the above -mentioned ideas, a large variety of phase-separating lipid systems have been investigated by DSC. Lipid partial immiscibility may arise from the different interactions occurring among the head groups, due to electrostatic interactions, or from different interactions among the tail due to their different length or unsaturation along the hydrocarbon chains. This effect is easily understandable and it is conceptually similar to that found in binary poorly mixable liquids. However, when mixed lipid bilayers and lipid / protein membranes are considered, a new strong, indirect force, favoring micro-domain formation, emerges. The physical basis of these forces, sometimes termed as hydrophobic mismatch, has been introduced by Mouritsen and his co-workers.\[[@CIT184]\] Several authors have confirmed and further improved the original theory of Mouritsen.\[[@CIT185]--[@CIT192]\] It has been assumed that these rafts can be isolated from the membranes by detergents. ITC studies of the enthalpy and entropy of interaction of the detergent *triton* with different lipids imply, however, that the addition of a *triton* to the membrane, changes the degree of domain formation and the composition of the domains substantially.\[[@CIT193]\] The predicted exothermic process of *triton*-induced formation or growth of domains could indeed be detected by ITC, and the stabilizing effect of the triton on these domains can also be measured by DSC and PPC.\[[@CIT193]\] Some years ago, Melchior\[[@CIT194]\] proposed a useful calorimetric trick to investigate inhomogeneous lipid membranes. The application of rapid-freezing techniques to DSC provides a new approach for understanding the organization of lipids in biomembranes. Use of quick-freeze DSC on membranes of mixed lipid composition supports the existence of nonrandom distributions of lipids (domains) in fluid bilayers. In addition to allowing investigations on the organization of lipids in the fluid bilayers, the quick-freeze technique now allows calorimetric studies to be carried out on mammalian membranes, which, because of their high cholesterol content, have not been previously amenable to the study, by DSC. Differential calorimetry experiments provide extremely useful thermodynamical parameters to characterize the temperature behavior of lipid mixtures. However, as the number of components in the lipid mixture increases, data analysis becomes very difficult. Although thermodynamic information from the DSC experiment can be extracted from the system, no detailed information about the physical characteristics of lipid lateral structure at different temperatures can be obtained using this technique. Nowadays, fluorescent techniques and scanning microscopies provide additional information not available by thermodynamic measurements. Coupling between lipid domains and membrane properties {#sec2-19} ------------------------------------------------------ An astonishingly large number of membrane properties are modulated by the formation of lipid domains. One of the most challenging topics involves the possibility that local concentrations of lipids having different shapes could couple with membrane curvature to produce a sorting mechanism of the different membrane components. This issue presents us with the following questions: How strongly curved must a membrane be to produce a sizeable sorting effect? The counterpart of this question is: How does the effect of this curvature vary with the degree of asymmetry of the lipid composition? Among the rapidly growing literature in this relevant field, we have just quoted some of the more recent articles.\[[@CIT195]--[@CIT197]\] The lateral phase separation-bilayer shape coupling plays a key role in explaining a variety of biological relevant phenomena such as hexo- and endocytosis, for space reasons we will not analyze such a broad field. Effect of curvature strain on the thermodynamic properties of membranes {#sec2-20} ----------------------------------------------------------------------- Membrane curvature effects of inclusion compounds have been found to play an important role in biological membrane function,\[[@CIT198]--[@CIT202]\] and therefore, they have been widely explored over the past decades. It has turned out that most of the membrane-ordering or disordering effects of additives can be interpreted in terms of a relaxation or induction of curvature strain. The general background of these phenomena can most easily be illustrated by Israelachvili's concept of 'effective molecular shapes' dictated by the ratio of the lipid surface area / maximum stretching length (under the hypothesis of substantial volume incompressibility of the hydrocarbon chains that are assumed to adopt a liquid-like arrangement in all the geometrical conformations of the lipid aggregate).\[[@CIT203]\] It is worth recalling that the maximum stretching length is a property of a single lipid molecule, but the lipid surface area strongly depends on the lateral collective interactions of the lipid heads and on the solvent properties. Molecules such as POPC pack together to a planar arrangement, as the surface area required by two fluid chains (≈ 2 × 27 Å^2^) agrees fairly well with the surface area occupied by the PC head group (≈ 61 - 65 Å^2^). Surfactants with a large head group, but only one acyl chain are referred to as 'inverted cone-shaped'; they pack together to form a strong, positively curved (convex), micellar surface. Molecules such as DOPE, with a small head group and a hydrophobic part requiring a relatively large surface area, tend to form curved surfaces with the hydrated heads in the center; these are called inverse or negatively curved structures. Although the preferred, 'spontaneous' curvature varies gradually, the choice of surface geometries that can be realized by stable aggregates is limited. The average real curvature of a lipid bilayer of a large vesicle is practically zero, but that of other (e.g., micellar or cubic phases) geometries differs substantially from zero. The difference between the spontaneous curvature of the constituents and the real curvature of the resulting aggregate is called a 'curvature strain'. As a rule, enthalpies of membrane insertion of additives measured by ITC have been found to be more endothermic if the curvature strain they create in a membrane is more.\[[@CIT204][@CIT205]\] Similar results have been seen when comparing the DSC data of phospholipids bilayers with different radii. Experimentally, depression and broadening of the phase transition temperature is observed for strongly curved vesicles. For vesicles smaller than \~70 nm in diameter the phase transition temperature gradually decreases with decreasing vesicle size.\[[@CIT204][@CIT206]\] Similar effects have been detected by using densitometric techniques\[[@CIT207]\] or for bilayers deposited into nanopores.\[[@CIT208]\] Additives that can relax a pre-existing curvature strain may bind exothermally.\[[@CIT204]\] These results suggest that the excess enthalpy, H~EXC~, of a bilayer (see section *Non-ideal mixing*) is governed by the curvature strain. Although non-ideal mixing and general membrane ordering are strongly related to the curvature strain, a more specific interpretation of spontaneous curvature effects is possible, considering the lamellar-to-inverse hexagonal transition of suitable model lipids (e.g., POPE), as the latter is accompanied by a real change in curvature from zero (lamellar) to negative values (inverse hexagonal). Compounds that induce positive spontaneous curvature favor the lamellar phase and increase the transition temperature, *T~hex~*, whereas, substances inducing negative spontaneous curvature promote the curved phase and decrease *T~hex~*.\[[@CIT209]--[@CIT212]\] Recent applications of these concepts have been proven to be useful in investigating cationic membrane-DNA complexes.\[[@CIT213]\] Passive membrane permeability {#sec2-21} ----------------------------- A biological relevant phenomenon is the passive permeability of lipid membranes. There is a general consensus that this phenomenon does not depend on the fine chemical structure of the diffusant and membrane, but rather depends on the collective properties of the whole membrane. Density (or volume) fluctuations are the likely cause of the temperature-dependent permeability of lipid membranes. Interestingly, the maximum permeability occurs near the main phase transition of lipid monolayers\[[@CIT214]\] and bilayers,\[[@CIT215]\] where the amplitude of the density fluctuations reach a maximum (see sections *Static and Dynamic Volume compressibility* and *Area compressibility*). Papahadjopoulos *et al*\[[@CIT216]\] were the first to demonstrate that the permeability for sodium ions (they used radiolabeled^22^ Na^+^ ions) increased by at least a factor of 100 in the phase transitions of dipalmitoyl phosphatidylglycerol (DPPG) and dipalmitoyl phosphatidylcholine (DMPC), in agreement with the phase transitions of these lipids, as measured by the fluorescence changes of embedded markers. The permeation profile for DPPC was found to be similar. It was demonstrated that cholesterol both abolishes the permeability maximum and the chain melting discontinuity. Along the same lines Mouritsen *et al*\[[@CIT217]\] and Sabra *et al*.\[[@CIT218]\] found that the permeability of dimyristoylphophatidylcholine (DMPC) membranes for Co^2+^ was drastically enhanced in the phase transition regime. These authors also demonstrated that the insecticide, lindane, changes the permeability. Jansen and collaborators\[[@CIT219]\] showed that membranes in their transition are much more permeable to water. In a recent series of articles, Heimburg and his coworkers had shown that the passive permeability P is strongly related to the area compressibility defined by [eq. (18)](#FD19){ref-type="disp-formula"} through the relationship:\[[@CIT220]\] $$P = P_{deg} + const\ \cdot \ {K_{T}}^{A}$$ where *P~o~* is the ideal permeability in the absence of fluctuations. As compressibility changes are proportional to specific heat variations, ∆*c~P~*, obtained by DSC measurements, [eq. (31)](#FD32){ref-type="disp-formula"} can be re-written as $$P = P_{deg} + const'\ \cdot \ \Delta c_{p}$$ The validity of [eq. (33)](#FD34){ref-type="disp-formula"} has been experimentally tested. Analogous arguments can be set forward for what concerns the electrical conductivity.\[[@CIT221][@CIT222]\] Finally, pore formation in membranes by the inclusion of antibiotic peptides\[[@CIT223][@CIT224]\] has also been studied by ITC. Membrane stability and solubilization {#sec2-22} ------------------------------------- Membrane stability can be directly quantified in terms of the free energy of the mixed membrane compared to the free energy of the most favorable alternative structure. For micelle-forming additives, the free energy of the alternative micellar state can be approximated by that of pure additive micelles, as the freedom of micelles to vary their size and shape renders mixing in micelles typically close to the ideal. Let *CMC* be the critical micellar concentration and *K~o~* the partition coefficient, then, the standard chemical potential difference of the solute between bilayers and micelles will be, $$\Delta{\mathbf{u}_{S}}^{0,\ b\rightarrow m} = RTln\left( {K_{deg} \cdot CMC} \right)$$ Which can be considered as an indicator for membrane destabilization by micelle-forming solutes.\[[@CIT135]\] Molecules perturbing the membrane, already at a low concentration, show $\mathrm{\Delta}u_{S}^{0,b\rightarrow m}$ \< 0, that is, *K~o~· CMC*\<1. Molecules with *K~o~· CMC*\<1 do not destabilize the membrane at low concentration, but may solubilize membranes due to cooperative effects at very high additive concentrations. Another approach to shed light on the membrane-disordering effects of additives is to investigate their effect on the melting temperature *T~m~* and other characteristics of the gel-to-liquid crystalline transition of a model lipid. As discussed in section *Modification of the membrane phase diagram by solutes*, an additive that disorders the membrane can be expected to favor the fluid phase over the gel phase so that the *T~m~* is lowered. ITC is an excellent method to study membrane solubilization, which is thought to be a surfactant-induced, lamella-to-micelle transition.\[[@CIT225]\] Similarly, ITC can be also be successfully employed to study the reconstitution of vesicles on addition of lipids to a micellar lipid-surfactant system.\[[@CIT132][@CIT142]\] This method does not detect the lamellar or micellar state *per se*, but the trend of the system to form micelles or vesicles. Below the critical concentration for solubilization, the injected surfactant micelles dissolve, and the surfactant is partially inserted into the membrane depending on the membrane-water partition coefficient of the surfactant and on the surfactant-to-lipid ratio.\[[@CIT139][@CIT193][@CIT226]\] This micelle-to-membrane transfer is typically endothermic. The appearance of the first stable mixed micelles in the system cannot be detected by structural methods, as virtually all the material is still in a lamellar phase. However, it reverses the direction of the surfactant transfer, injected surfactant micelles persist now and extract surfactant (exothermic) and lipid from the vesicles. This leads to a sudden jump (usually accompanied by a reversal in sign) of the heat of titration. The surfactant-induced lamella-to-micelle transition of lipid systems has also been studied by DSC.\[[@CIT227]\] The transition of fluid lipid bilayers to the inverse hexagonal phase can be induced by increasing temperature (monitored by DSC) or by the addition of compounds or changes in ionization, inducing negative spontaneous curvature.\[[@CIT228]\] For a recent review of the broad field of the membrane-surfactant interactions see, for example, Keller *et al*.,\[[@CIT229]\] Garidel *et al*.\[[@CIT230]\] and Heerklotz.\[[@CIT231]\] Membrane fusion {#sec2-23} --------------- Some attempts have been made to employ calorimetric techniques to investigate the fusion events among lipid vesicles. The time evolution of the DSC thermograms of a suspension of lipid vesicles is a clear indication of the occurrence of fusion events among the particles. Such an observation has been exploited in different ways. Consider, for instance two suspensions of vesicles of different lipid composition rapidly mixed at time *t* = 0. The bilayer of the two vesicles undergoes a different melting temperature, therefore, if they appreciably differ, the associated DSC thermograms show two distinct, well-separated peaks. After mixing we observe a gradual shift in the temperature transition proportional to the amount of fused vesicles. The basic requirement for applying this technique is that the fusion rate must be very slow in comparison to the time-scale of a typical DSC run. Another interesting application of DSC is the relationship between the fusion rate of the lipid vesicles and the physical state of the vesicles' lipid bilayer. It has been generally observed that the fusion rate (determined by fluorimetric techniques) rapidly increases above the gel-to-liquid crystalline phase transition (determined by DSC).\[[@CIT232][@CIT233]\] The fusion of viruses with lipid vesicles has been studied using ITC.\[[@CIT162][@CIT234]\] As the integral heats of titration do not provide any information on transient states, the heat of fusion of the bilayers *per se* is small, and the heat effects observed should be mainly attributed to interactions of viral proteins with the target membrane. For example, a partial deprotonation of a viral protein on membrane fusion was detected by ITC using the buffer variation method (see section *Isothermal titration calorimetry*).\[[@CIT162]\] Also the enthalpy of proton-induced vesicle fusion was measured by ITC.\[[@CIT228]\] DSC studies of viral proteins have yielded important information on fusogenic protein states in viruses (see the next section). Finally, the measurement of another thermodynamic parameter, the volume compressibility (performed by acoustic techniques as described in section *Static and Dynamic Volume compressibility*), has been applied to investigate the well-known phenomenon of polymer-enhanced fusion of lipid vesicles, by exploiting the relationship between compressibility and vesicle surface hydration.\[[@CIT235]\] Stability and Partitioning of Proteins in a Lipid Environment {#sec1-5} ============================================================= The fundamental issue of the insertion of a hydrophobic protein into a lipid membrane has stimulated an extremely large number of studies. On account of the complexity of the problem, several attempts to capture the main factors involving the energetics of protein insertion into the lipid core have been done. We list a typical thermodynamic description of the whole process: Hydrophobic effect {#sec2-24} ------------------ The free energy gained from the hydrophobic effect on the incorporation of proteins into a lipid bilayer can be calculated in two ways: (i) On the basis of the amino acid sequence and the free energies of transfer of individual amino acid side chains from water into the vapor phase\[[@CIT236]\] or into hydrocarbons,\[[@CIT237]\] for a recent critical analysis of the transfer energy of a protein into different solvents see, e.g., the study of Simon *et al*;\[[@CIT238]\] or (ii) from the change of the protein / water interfacial area together with a value for the free energy change per unit area.\[[@CIT239]\] From studies of hydrocarbons and hydrophobic amino acids, the free energy per area was found to be 20 -- 25 cal/(*mol* · A^2^), while the interfacial area is that area of a protein molecule that is accessible to water molecules. It can be estimated by describing the protein as a sphere or cylinder or by numerically evaluating the true surface area by standard packages, which calculate the solvent inaccessible regions for any protein geometry.\[[@CIT240]\] This energy strongly depends on whether the protein is in a helical or unfolded conformation. As the hydrophobic effect originates in the reduction of the mobility of water molecules, it is predominantly of an entropic nature. The enthalpy change is relatively small, less than a few kcal/mol, and it is often neglected. Hydrogen bonds and conformational changes {#sec2-25} ----------------------------------------- If hydrogen bonds between protein and water molecules are broken by the incorporation of the protein and not restored in the membrane, an energy of about 5.8 kcal/mol of the hydrogen bond is lost.\[[@CIT241]\] To prevent this large loss of energy, protein molecules in the membrane adopt a conformation that allows the intramolecular formation of hydrogen bonds. This is optimal in an α-helical conformation. Hence, the hydrogen bonds are the cause for the frequently observed α-helical conformation of membrane-incorporated protein segments. Considering the final conformation of the protein in the membrane as an a-helical, the change in the internal degrees of freedom depends on the protein conformation in water. If the protein in the water is also helical, the internal degrees of freedom do not contribute to the free energy change. If, however, the protein in water adopts an unfolded conformation, internal degrees of freedom become lost on incorporation in the lipid matrix. The corresponding free energy change can be estimated for helix-coil transitions and amounts to 1.2 kcal/mol of the residue.\[[@CIT242]\] Protein immobilization effect {#sec2-26} ----------------------------- The change in free energy due to the immobilization of external degrees of freedom of a protein, on incorporation, can be easily estimated. The protein in water is treated as a freely moving particle; its free energy given by that of an ideal gas. In the membrane it is treated as completely immobilized without an energy cost. The change in free energy of the translational degrees of freedom can then be calculated from standard formulas of statistical thermodynamics. In the case of a bilayer-spanning protein of 20 amino acid residues with a molecular weight of about 2,000, at *T* = 300 K, one obtains ∆*G*≈ 10 kcal/mol. Under the same assumptions, immobilization of the rotational degrees of freedom yield approximately the same value, thus, protein immobilization is found to involve an energy change of about 20 kcal/mol. The above-mentioned figures can be slightly modified by allowing a partial retention of the freedom degrees of the protein within the lipid bilayer, but the overall picture remains unchanged. Perturbation of the lipid matrix {#sec2-27} -------------------------------- Protein insertion gives rise to a significant alteration of the lipid bilayer order parameter at the protein-bilayer periphery. The effect is even more dramatic if the lipid bilayer and the protein inclusion have different thicknesses. This point is very delicate and arises from a subtle interplay of different contributions, requiring the knowledge of the order parameter of the lipid chains, near the lipid-protein interface. Neglecting for a moment, the problem of different lipid-protein heights, the lipids contacting the included proteins lose several of their internal degrees of freedom just for geometrical reasons. The lipid molecules are strongly coupled with each other, this local anomalously large order parameter slowly relaxes with the distance from the lipid-protein interface. Things are even more complex if one includes lipid-protein height variations. As stated in section *Lateral phase separation: different routes to domains formation*, the different height between a membrane protein and the lipids generate a curvature of bilayer thickness around the protein: in order to minimize this unfavorable arrangement, proteins attract each other, eventually leading to the formation of large protein clusters. The sum of the four energies described above is the clue for membrane-water partitioning and protein conformational transition. As a rule of thumb: Because the net number of hydrogen bonds is not significantly changed on going from a helix to a water-solvated coil, the aqueous helix-coil transition is approximately isoenergetic. Water-lipid partitioning of the helix is estimated to be about 30 kcal/mol in favor of the lipid by virtue of the hydrophobic effect: exposing a hydrophobic helix to water will dramatically reduce water entropy. The water-lipid partitioning of the coil is estimated to be about 40 kcal/mol in favor of the water, due to the loss of protein-water hydrogen bonds on entering the bilayer. The resulting energy estimated by the above thermodynamic cycle must be augmented in the presence of interactions between the different helices belonging to the same protein and embedded into the lipid matrix. A large number of articles dealing with the key issue of the energetics of membrane partition into a lipid bilayer can be found in the literature. A non-exhaustive list of some representative articles is as follows: Janhig,\[[@CIT243]\] Popot *et al*.,\[[@CIT244]\] Ben-Tal *et al*.,\[[@CIT245]\] White *et al*.,\[[@CIT246]\] Engelman *et al*.,\[[@CIT247]\] and Babakhani *et al*.,\[[@CIT248]\] while for some representative reviews on the simulation of lipid-protein interactions see, e.g., Biggin and Sansom\[[@CIT249]\] and Bond *et al*..\[[@CIT250]\] As stated before, membrane partitioning and helix formation are strongly coupled. On the experimental side, studies on this link are not always investigated and many studies focus on the helix stability alone. The denaturation behavior of membrane proteins has been studied by DSC in reconstituted vesicles as well as in whole viruses or cells; for a review see Shnyrov *et al*..\[[@CIT251]\] It is worth mentioning that most membrane proteins seem to exhibit smaller enthalpies of denaturation (≈14 kJ g^-1^) than typical soluble proteins (≈33 kJ g^-1^), suggesting that the membrane stabilizes some residual structure.\[[@CIT252]\] Wieprecht *et al*.\[[@CIT253][@CIT254]\] claimed that they could separate the conformational and partitioning effects by ITC experiments, by comparing all-L peptides with DD-isomers, which should show the same hydrophobicity, but are not (or are less) capable of forming a helical structure. Kinetics Phenomena {#sec1-6} ================== Heating and cooling modes DSC {#sec2-28} ----------------------------- Indirect information on the kinetics of transitions from an ordered to a disordered lipid configuration can be obtained by investigating the effect of the DSC scan rate on the apparent transition temperature and the shape of the DSC peaks.\[[@CIT11][@CIT255]\] This is a well-known general chemicophysical effect, independent of the peculiar nature of lipids and valid for any melting process of simple, point-like molecules, where molecules can be arranged in an ordered (solid) regular lattice on a disordered (fluid) structure. The extremely large number of internal degrees of freedom, typical of lipid molecules, introduce additional effects due to the coupling between positional and internal order parameters of the lipids. Let us start with the simplest picture. The process of phase transformation is studied by driving an initial phase into a region of the phase diagram where it is metastable or unstable. Hysteresis is usually observed during phase transformation. According to the two-dimensional nucleation theory,\[[@CIT256][@CIT257]\] the transformation from a disordered to an ordered phase requires the formation of a so-called critical nucleus, or gel domain when we consider lipid bilayers. The free energy ∆G to form a gel domain in the fluid phase is given by the expression $$\Delta G = \Delta un + 2\gamma\ \left( {\pi\sigma n} \right)^{1/2}$$ where ∆*µ* denotes the temperature-dependent chemical potential of a lipid in the gel with respect to the fluid phase, g the line tension between the fluid and gel phase,\[[@CIT258]\] n the amount of lipid constituting the nucleus, and s the area per lipid in the gel phase. Above the main phase transition temperature, both terms are positive and only small nuclei can form (so-called heterophase fluctuations).\[[@CIT63][@CIT259]\] Below the transition temperature, the chemical potential in the gel phase becomes lower than that of a lipid in the fluid phase, driving the transformation. However, this driving force is opposed by the line tension arising from the gel-fluid interface. There exists a critical nucleus size n\*=*πσγ^2^*/(∆*µ*)^2^ for which the free energy exhibits a 0maximum ∆G\* $$\Delta G^{*} = \frac{\pi\sigma\gamma^{2}}{\Delta u}$$ Gel nuclei with a size n \< *n*\* are unstable and will dissipate. Nuclei with n \> *n*\*, however, will grow, thereby transforming the entire system into a state of lower free energy, that is, the gel. The time *t*\* required to overcome this barrier will scale as *t*\*≈exp(∆*G*\*/*kT*). For a system quenched to a temperature much lower than the transition temperature, ∆*µ* becomes large and ∆*G*\* vanishingly small. In this case, there is almost no impediment to the phase transformation process. On the other hand, at a temperature close to the phase transition temperature, both the critical cluster size and the time required to form the critical cluster diverge. If cluster growth results from the (reversible) addition of single lipids to the cluster boundary, the speed of gel phase propagation is given by:\[[@CIT63]\] $$u = u_{MAX}\ \left( {1 - \exp\left( {\Delta u/kT} \right)} \right)$$ where u~MAX~ denotes the maximum achievable speed when the probability of the reverse process can be neglected. Therefore, if the cooling rate of a typical DSC experiment is fast, only a limited number of solid nuclei begins to form and grow inside the membrane, which remains in a fluid undercooled state. This has a deep influence on the position intensity and shape of the calorimetric peak detected by the DSC measurement. Direct experimental evidence for the nucleation and growth mechanism in lipid bilayers is difficult to obtain. Within the framework of heterophase fluctuations, Kharakoz *et al*,\[[@CIT63]\] were able to derive a kinetic model explaining the ultrasonic anomalies observed in experiments on multilamellar vesicles. By fitting to the kinetic model, estimates of the line tension and the thermodynamic driving force can be obtained. Direct visualization of the initial stages of cluster nucleation and growth has thus far only been achieved for two-dimensional colloidal systems, by using fluorescent probes more soluble in the fluid phase. This picture has been recently confirmed by detailed Molecular Dynamics simulation.\[[@CIT260]\] Very good agreement with the classical two-dimensional nucleation theory has recently been reported for colloidal nucleation driven by an electric field, which allows precise control over the thermodynamic driving force.\[[@CIT261]\] The different transition temperature values observed in the heating and cooling modes and described earlier, is a reversible phenomenon observed in most of the lipid systems. A limited number of lipid systems show irreversible effects observed only in the first temperature run. Sometimes people observe a temperature shift, or even the appearance of a peak, just in the first DSC run; following that the peak remains constant in all the subsequent scans. A typical example is given by glycolipids, and especially from a glycolipid subclass: the gangliosides. In these lipids the head is bulky (5 -- 7 sugar units) and has a size comparable to that of the tails. Tightly packed head groups may show cooperative effects similar to those observed for the tails. Therefore, the system exhibits a richer phase behavior as extensively studied by Corti and coworkers by DSC and structural techniques.\[[@CIT262][@CIT263]\] Kinetics of phase transitions {#sec2-29} ----------------------------- In the previous section we have investigated non-equilibrium phenomena in lipid bilayers through DSC measurements performed at different scan rates. A complete and more direct approach, made possible by the progress in calorimetric instrumentation, exploits the response of the system to a sudden perturbation. Detailed studies of the kinetics of lipid phase transitions, in the absence and presence of additives, have been performed by measuring the time-dependent thermal response of lipid samples to periodic pressure modulations\[[@CIT32]\] and pressure jumps.\[[@CIT27][@CIT35]\] Experimental results evidence a good relationship between the temperature-dependent relaxation times of chain melting and heat capacity. Small amounts (1 mol%) of cholesterol added to DPPC reduce the relaxation time, τ, by a factor of 4,\[[@CIT27]\] whereas, about 1 mol% of the anesthetic dibucaine increases τ two-fold.\[[@CIT31]\] Often these effects have been interpreted as being related to the size of the cooperatively melting clusters in the membranes. In a recent series of articles, however, Heimburg and his coworkers suggested a different interpretation. Starting from the theory of thermodynamic fluctuations and the Landau ansatz for the relaxation rate of out-of-equilibrium phenomena, Heimburg *et al*. derived a linear relationship between the relaxation time τ and the specific heat *c~P~*.\[[@CIT27][@CIT264]\] They used the standard Landau assumption that a relaxation of any 'order parameter' *S* is proportional to its distance from the equilibrium value *S~eq~*: $$\frac{\partial S}{\partial t} = \ - \Lambda\frac{\partial G\left( {S - S_{eq}} \right)}{\partial S}$$ Where *G(S -- S~eq~)* is the free energy of the system expressed as a function of the deviation from the equilibrium *S -- S~eq~* and Λ is a viscosity-related mobility factor. Thus --∂G(S --S~eq~)/∂S plays the role of a thermodynamic force, driving the system to a new equilibrium in response to an external perturbation. A calculation gives an exponential time decay: *S --S~eq~* ≈ exp(--*t/τ*), where $$\tau \approx \frac{RT^{3}}{L}c_{p}$$ L being a phenomenological constant. The linear relationship between τ and *c~P~* foreseen by [eq. (39)](#FD40){ref-type="disp-formula"} has been indeed experimentally tested. Some studies have also been performed on the kinetics of the transition from the ordered gel phase to the undulated (ripple) phase in phospholipid vesicles.\[[@CIT49]\] Kinetics of solute sorption and exchange {#sec2-30} ---------------------------------------- Isothermal titration calorimetry provides information on the kinetics of re-equilibration after injections.\[[@CIT265]\] Membrane partitioning of solutes is often fast compared to the time constant of fast calorimeters (≈ 15 s). If the penetration of the solute to the inner monolayer occurs within a few minutes, the heat peaks may exhibit a biphasic behavior.\[[@CIT137]\] Water sorption calorimetry with small sudden changes in RH (the relative humidity) reveals the swelling to occur within about ten minutes if the film is thin enough and the gas flow is sufficiently fast; an interpretation in terms of system kinetics is hardly possible. Even if kinetic constants are not of particular interest, it must be guaranteed for a thermodynamic evaluation of the data that the system reaches an equilibrium during the experiment. For example, the next injection of an ITC run should only be made after a sufficient time for the heat response to reach the baseline level. It must, however, be stressed that this is necessary, but not a sufficient criterion for having reached the equilibrium, as the re-equilibration of the system after an injection may exhibit complex kinetics involving processes with different time scales. In most cases, there is a simple, but very effective means, to rule out problems arising from slow processes: to combine up- and down-scans or scans with different speeds. This is a routine in the DSC of lipids. In ITC, it is advisable to combine, for example, the uptake and release or solubilization and reconstitution experiments, to rule out incomplete equilibration. In sorption calorimetry, it is useful to compare up- and down-scans in RH. PPC performs up- and down-jumps in pressure routinely, thus allowing one to recognize the irreversible effects and metastable states occurring in a transition.\[[@CIT266]\] In a recent series of articles\[[@CIT267]--[@CIT269]\] we investigated by DSC, the transient variation of the calorimetric peak associated with the lipid main transition of multilamellar one-component vesicles, incubated at different times with a diffusant impurity dissolved / dispersed in the buffer solution. In the early stages of the sorption process the DSC scan showed a single narrow calorimetric peak, typical of a pure lipid bilayer. At longer incubation times the peak broadened and shifted in temperature, and finally, on approaching the equilibrium distribution of the impurity between the lipid and water the peak became narrow again, but the transition temperature shifted to a new position. This effect was due to the unequal distribution of the drug between the outer and inner bilayers of the multilamellar vesicles during the partition / permeation kinetics. As discussed in section *Modification of the membrane phase diagram by solutes*, impurities shift the transition temperature of a bilayer in a way dependent on their local concentration. What we observe at intermediate times is just the convolution of signals coming from regions with different concentrations of the impurity. At equilibrium the two-peak structure merges into a unique peak because the impurity is evenly distributed over the entire multi-lamellar structure of the liposome. This finding may provide useful information about the lipid bilayer permeability and partition coefficient in model membranes. These parameters could be quantitatively measured in a series of DSC measurements performed at different times, provided a proper diffusion / partition interpretative model is developed. The obvious limitation of this technique is that it applies to slow the permeation kinetics. Studies are in progress in this field. Financial support from the Italian National Science Foundation is gratefully acknowledged. **Source of Support:** Italian National Science Foundation **Conflict of Interest:** None declared.

PubMed Central

2024-06-05T04:04:19.948215

2011-01-01

PMC3053514

Eclipse was observed in India on 15^th^ January, 2010. The solar eclipse of this day was an annular eclipse of the sun with a magnitude of 0.9190. A solar eclipse occurs when the moon passes between earth and the sun, thereby totally or partially obscuring earth's view of the sun. An annular solar eclipse occurs when the moon's apparent diameter is smaller than the sun, causing the sun to look like an annulus/ring, blocking most of the sun's light. An annular eclipse appears as a partial eclipse over a region thousands of kilometers wide. At approximately 13:20 IST, the annular solar eclipse entered India at Thiruvananthapuram, Kerala, and exited India at Rameswaram, Tamil Nadu. The eclipse was viewable for 10.4 min in India. It was a total eclipse in some parts of the country, while it was partial in other parts. This occurs near a new moon when the moon is between sun in some parts of the world. During this period, the direct sun rays do not fall on the earth; a significant trait of sun rays only falls which have an impact on the nature. To understand this, a primitive study is carried on a selected group of bacteria and *Candida albicans*. The aim of this study was to scientifically prove life supportive influence of solar eclipse on prokaryotes and eukaryotes. To study the effect of solar eclipse, we selected bacteria for prokaryotes and *C.albicans* for eukaryotes. This subject is important to understand the concept that confluent growth on culture plate, morphological change, antigenic variation, decreased antibiotic sensitivity, and increased virulence mechanism is exhibited on exposure to solar eclipse by bacterial and yeast cultures. Phenotypic transitional variation is considered as a potential virulence character. This common feature observed in fungus refers to phenotypic switching which may assist transition from commensalisms to pathogenic phase of microorganisms.\[[@CIT1]\] The filament-inducing property of *C. albicans* was also studied by other workers as documented in our study. Microbes invade the host defense mechanism. A phase transition in microbes explains drift and shift in the antigenicity of the microbes.\[[@CIT2]\] This change results in the incompatibility of the host defense mechanism, further leading to impairment of immunological status in altered occasions. With regard to C. *albicans* which form the germ tube, transitional stage of hyphal development is analogy to bacterial phase transition phenomenon. This factor has an influence on immunoregulatory of fungal surface components.\[[@CIT3]--[@CIT5]\] There was a markable metabolic activity observed before the emergence of germ tube. When the cell-wall deposit becomes polarized, the germ tube forms.\[[@CIT6]\] Phospholipases concentrated at the hyphal tip enhances virulence. Hyphae being larger than the yeast is more resistant to phagocytosis.\[[@CIT7]\] Materials and Methods {#sec1-1} ===================== The study was carried out in A.J. Institute of Medical Sciences, Department of Microbiology, Mangalore, Karnataka, India. Study of the effect of solar eclipse on *Staphylococcus aureus, Klebsiella* species, and *Escherichia coli* (cultures) and C. *albicans* culture were included in the study. The criterion adopted in this study was the demonstration of morphological changes observed during exposure to normal sunlight and eclipse phase. A detailed cultural characteristics and biochemical reactions was also compared. We alsoevaluated differences in antibiotic sensitivity of microorganisms in comparison with bacterialcultures at the time of exposure to solar eclipse and normal sunlight. Bacterial and yeast colonies were inoculated onto brain heart infusion broth from Hi-Media Laboratories Pvt. Limited, Mumbai. These bacterial suspensionswere assayed during the time period from 11.15 am to 3.15 pm (total duration = 4 h).. Bacterial and yeast colonies were studied on nutrient agar, Nutrient broth and Sabourauds dextrose agar and broth, respectively. Nutrient broth : Beef extract 1.50 g/l, yeast extract 1.50 g/l, peptic digest of animal tissue 5.0 g/l, sodium chloride 5.0 g/l, distilled water 1000 ml, pH to 7.5-7.6 from Hi-Media Laboratories Pvt. Limited, Mumbai. Nutrient agar : To the ingredients as in nutrient broth (beef extract 1.50 g/l, yeast extract 1.50 g/l, peptic digest of animal tissue 5.0 g/l, sodium chloride 5.0 g/l, distilled water 1000 ml, pH to 7.5-7.6), add 15 g agar per liter. Dissolve the agar in nutrient broth and sterilize by autoclaving at 121° C for 15 min. Blood agar: Nutrient agar 100 ml, sheep blood (defibrinated) 10 ml from Hi-Media Laboratories Pvt. Limited, Mumbai. Mac Conkey's agar : Peptic digest of animal tissue 20 g/l, lactose 10 g/l, sodium taurocholate 5g/l, neutral red 0.04 g/l, agar 20 g/l, Ph = 7.4±0.2 from Hi-Media Laboratories Pvt. Limited, Mumbai. Sabourauds dextrose agar: Mycological peptone 10 g/l, dextrose 40 g/l, agar 15 g/l from Hi-Media Laboratories Pvt. Limited, Mumbai. Muller Hinton agar(MHA): Beef infusion form 300 g/l, casein acid hydrolysate 17.50 g/l, starch 1.5 g/l, agar 17.0 g/l from Hi-Media Laboratories Pvt. Limited, Mumbai. Hi-CHROME *Candida* differential agar: Peptone 15 g/l, yeast extract 4 g/l, dipotassium hydrogen phosphate 1 g/l, chromogenic mixture 7.22 g/l, chloramphenicol 0.50 g/l, agar 15 g/l, Ph 6.3±0.2 at 25 c from Hi-Media Laboratories Pvt. Limited, Mumbai. *S. aureus, Klebsiella* species, and *E. coli* and C. *albicans* were studied after incubation for 24 h at 37 °C. It was further streaked on to blood agar and CHROM agar and incubated for 24 h. An isolated colony was picked aseptically fromthe agar plate and inoculated onto Muller Hinton broth and peptone water broth for further biochemical reactions and to perform antibiotic sensitivity testing. Antibiotic discs from Hi-Media Laboratories Pvt. Limited, Mumbai, were used for antibiotic sensitivity testing. *S. aureus, Klebsiella* species, and *E. coli*, and *C. albicans* were studied after incubation for 24 h at 37°C. Results {#sec1-2} ======= The study was performed during the time period from 11.15 am to 3.15 pm on 15 ^th^ January, 2010, at A.J. Institute of Medical Sciences and Research Centre in Mangalore, Dakshina Kannada District, Karnataka State. Number of cultures tested were cultures of *S. aureus, Klebsiella* species, *E. coli*,and C. *albicans* from clinical isolates. Interpretation of results was based on definite changes in their morphology on smear examination, cultural characteristics on blood agar, Mac Conkeys agar and nutrient agar, biochemical reactions: catalase test, indole test, coagulase test, methyl red test, vogues Prausker test, triple sugar iron agar, urease test, citrate utilization test, and antibiotic sensitivity pattern on Muller Hinton Agar by Kirby Bauer's disc diffusion method. Effect of normal sunlight and solar eclipse on bacterial and yeast colonies {#sec2-1} --------------------------------------------------------------------------- There was an appreciable difference in the morphology observed on gram-stained smears of microorganisms - *S. aureus, Klebsiella* species, and *E. coli*, and C. *albicans* cultures both in broth and agar media. While *S. aureus, Klebsiella* species, and *E. coli* were stained intensely/dark stained from those cultures which were exposed to normal sunlight. Uneven staining was observed in cultures exposed during eclipse in bacteria and *C. albicans* culture. When stained showed abundant formation of germ tubes with reduced cell size and densely stained nucleus and clear cytoplasm.\[[@CIT2]--[@CIT4][@CIT8]\] The number of germ tube formation in *C. albicans* remarkably increased suggesting active multiplication with acquisition of greater virulence.\[[@CIT1]\] The effect of solar eclipse on the growth of *S. aureus, Klebsiella* species, and *E. coli* cultures has resulted in phenotypic change \[Tables [1](#T0001){ref-type="table"}--[3](#T0003){ref-type="table"}\]. No changes were observed in biochemical reactions, colony characteristics, or capsule formation in case of *Klebsiella* species. ::: {#T0001 .table-wrap} Table 1 ::: {.caption} ###### Comparison of antibiotic susceptibility pattern on MHA : *Staphylococcus aureus* ::: Antibiotics Normal sunlight (mm) Solar eclipse phase (mm) Standard sensitive zone size (mm) ---------------------------------------------- ---------------------- -------------------------- ----------------------------------- -- Penicillin (10 units) 25 24 29 Cefazolin (30 μg) 27 26 1 Vancomycin (30 μg) 17 16 15 Teicoplanin (30 μg) 11 11 14 Clindamycin (2 μg) 25 24 21 Amoxycillin (20 μg)+ clavulanic acid (10 μg) 27 18 20 ::: ::: {#T0002 .table-wrap} Table 2 ::: {.caption} ###### Comparison of antibiotic susceptibility pattern on MHA :*Escherichia coli* ::: Antibiotics Normal sunlight (mm) Solar eclipse phase (mm) Standard sensitive zone size (mm) ----------------------------------------------------------- ---------------------- -------------------------- ----------------------------------- Aztreonam (30 μg) 24 22 22 Amikacin (30 μg) 24 22 17 Piperacillin (100 μg) 24 21 21 Ceftazidime (30 μg) 25 25 18 Ceftazidime (30 μg)+clavulinic acid (10 μg) (ESBL screen) 25 25 25 Ciprofloxacin (5 μg) 40 37 21 ::: ::: {#T0003 .table-wrap} Table 3 ::: {.caption} ###### Comparison of antibacterial susceptibility pattern on MHA : *Klebsiella species* ::: Antibiotics Normal sunlight (mm) Solar eclipse phase (mm) Standard sensitive zone size (mm) ----------------------- ---------------------- -------------------------- ----------------------------------- -- Cefotaxime (30 μg) 31 28 23 Meropenem (10 μg) 27 25 16 Ciprofloxacin (5 μg) 27 22 21 Piperacillin (100 μg) 18 17 21 Ceftazidime (30 μg) 28 25 18 Amikacin (30 μg) 23 21 17 ::: To observe any mutation on exposure to solar eclipse, a comet assay was performed. The comet assay methodology was performed by layering of normal agarose on slides. Cells are mixed with low-melting agarose and spread over glass microscope slides. Following the lysis of cell membrane and proteins, the unwound DNA subjected to electrophoresis, stained, and image analysis performed. There was no significant change observed in the yeast formed by this assay, which indicates that there may be no major impact of solar eclipse on human beings. Discussion {#sec1-3} ========== This is an unpublished data stating that solar eclipse does no harm to prokaryotes and eukaryotes. Whatever genotypic and phenotypic variation occurs contributes for the better survival of microbes and man. Further elucidation of impact of solar eclipse studies is to understand the complex nature of the composition of short waves, quantum waves emitted and absorbed in the environment, and the duration of exposure, which decides the phenotypic and genotypic changes in microbes and man.\[[@CIT9][@CIT10]\] Thus morphologic change contributes to increased pathogenic potential of fungus. This transitional change has also been recorded by molecular studies.\[[@CIT11]\] There seems to be various environmental stimuli which may enhance a switch between yeast and filamentous growth patterns.\[[@CIT12]\] The capacity to switch the pattern of growth in response to an environmental change is termed as dimorphism.\[[@CIT7][@CIT12][@CIT13]\] Since solar eclipse in India is considered to cause adverse effects on humans, may be to some extent on eyes or skin, the effect of the solar eclipse on the microorganism isolated from clinical isolates was investigated. We also hadevaluated differences in antibiotic sensitivity of microorganisms in comparison with morphological changes within bacterialcells at the time of eclipse and normal sunlight on 15thJanuary, 2010, from 11.15 am to 3.15 pm (total duration = 4 h) to ascertain possibility of mutation as the cause for phenotypic and genotypic change. The phenotypic changes were observed in Gram-stained smear. Conclusions {#sec1-4} =========== Our observations have important implications of solar eclipse and their induced variations. To our knowledge, such studies have not been reported previously for *S. aureus, Klebsiella* species, *E. coli*, and *C. albicans* in clinical isolates. However from our studies, we could infer that exposure to solar eclipse does help in the favor of microbes and man.\[[@CIT14]\] Further study is needed in this area to use solar eclipse therapeutically to enhance the immune system and targeting the destruction of genesis of malignant cells thus, to alleviate the suffering of mankind. **Source of Support:**Nil **Conflict of Interest:** None declared

PubMed Central

2024-06-05T04:04:19.959101

2011-01-01

PMC3053515

The incidence of yeast infections has increased in recentdecades,\[[@CIT1]\] while invasive infections by opportunistic *Candida* spp. have also been reported to have significant impact on human morbidity and mortality.\[[@CIT2]\] *Candida*, once considered as a minor pathogen,is now among the most commonly cultured pathogenic microorganisms, even in intensivecare units (ICU),\[[@CIT3]--[@CIT6]\] while vulvovaginal candidosis, which affects all strata of the society, has remained a common problem worldwide.\[[@CIT7]\] However, in a developing country like Nigeria, apart from the addresses on the packages of clinical drugs in pharmacies, the sources of most of the drugs cannot be fully authenticated or ascertained. Even, in spite of the massive activities by the National Agency for Food, Drugs Administration and Control (NAFDAC) against production and importation of adulterated and substandard drugs into the country, fake drugs are still reported on regular basis. A counterfeit formulation is one that is deliberately and fraudulently mislabeled with respect to identification and/or source. Counterfeiting can apply to both branded and generic products and counterfeits may include products with the incorrect ingredients or with the wrong ingredients, without active ingredients, with insufficient active ingredient, or with fake packaging,\[[@CIT8]\] and it is known that drug quality in public and private outlets may be problematic. A previous study in Nigeria, which assessed the quality of drugs from retail outlets and pharmacies, attributd the problems of counterfeit drugs to lack of quality control in manufacture, as well as degradation during storage.\[[@CIT9]\] There is also a little existing knowledge about actual quality of drugs provided by different providers in Nigeria and in many sub-Saharan African (SSA) countries. A search of the medical literature yielded only 43 primary published research reports concerning counterfeit drugs in the world,\[[@CIT10]\] while failing products more often originated or were claimed to originate from poorer parts of the world with weaker regulatory systems.\[[@CIT11]\] Over the past decade, the massive public health problem of counterfeit and substandard drugs has become more manifest, leading to serious clinical consequences on patients, such as increased morbidity, mortality, and drug resistance, which contributes to spurious reporting of resistance and toxicity, as well as loss of confidence in the healthcare systems.\[[@CIT10]\] Other studies looking at a broader range of diseases in Nigeria found widespread inappropriate drug use, low quality of treatment and ineffective regulations.\[[@CIT12]--[@CIT14]\] Quick results of *in vitro* susceptibility testing of *Candida* spp. to the common antifungal agents are desirable,\[[@CIT15]\] but usage of inconsistent batches of antimycotics, which can give varying results during treatments, calls for general concerns. The aim of this study is, therefore, to compare the susceptibility patterns of vulvovaginal candidiasis-associated *Candida* strains isolated from ECS and HVS clinical specimens to two different batches of the most-available antifungal agents in the country. Materials and Methods {#sec1-1} ===================== Identification of yeast isolates {#sec2-1} -------------------------------- A total of 36 strains of *Candida* isolated from high vaginal swabs (HVS) and endocervical swabs (ECS), which were obtained by clinical routine from patients who presented for candidosis and who had not been on antimycotic therapy in about 6 months prior to time of collection, were obtained from the culture collections of the Department of Medical Microbiology and Parasitology, University College Hospital, Ibadan, Nigeria. The *Candida* strains were sub-cultured by streaking on Sabouraud dextrose agar (SDA) (Lab M, England) plates and incubated at 32°C for 24-48 hours until assure purity, and characterized according to their colonial characteristics on CHROM-agar, microscopic morphology, as well as biochemical tests, including assimilation of sugars- cellobiose, dextrose, dulcitol, fructose, galactose, inositol, lactose, maltose, mannitol, mellibiose, rhamnose, saccharose, sorbitol, sucrose, xylose. The identification of the *Candida* strains was based on standard phenotypic taxonomic tools and clinical practices as previously described.\[[@CIT16][@CIT17]\] In addition, fresh wet mount examinations (wet preparations) and germinal tube assay were also performed on the yeast strains, and pure, identified strains were kept in triplicates on SDA agar slants at 4°C as bench and stock cultures. *In vitro* antimycotic susceptibility testing . *In vitro* susceptibilities / resistance to commonly available antimycotic agents in Nigeria \[the imidazoles-mycoten tablets/cream, canesten tablets/cream (i.e., clotrimazoles); tetradox (doxycycline); the polyenes-mycostatine (nystatin), and the metronidazole- flagyl\] were determined on SDA at 35°C after 24 and 48 hours ofincubation, using the modified method\[[@CIT18]\] of Tagg *et al.*\[[@CIT19]\] The concentration of the inoculum suspensions of the test *Candida* isolates were between 1.6 and 2.4 × 10^3^ cells ml^-1^. Statistical analysis {#sec2-2} -------------------- Tests of hypothesis using chi-square and ANOVA were carried out to show if there exists a significant difference between the two batches of antimycotic agents (B1 and B2).\[[@CIT20][@CIT21]\] Results {#sec1-2} ======= The *Candida* spp. isolated from clinical specimens (CV/HVS) were characterized in this study as *C. albicans* (19.4%), *C. glabrata* (30.6%), *C. tropicalis* (33.3%), and *C. pseudotropicalis* (16.7%). None of the *Candida* strains had 100% susceptibility profiles toward all the antimycotic agents in both batches. Varying multiple antifungal susceptibility (MAS) rates of 14.3-100%/85.7-100%; 28.6-100%/28.6-100%; 28.6-100%/14.3-100%, and 42.9-85.7%/14.3-100% were recorded in batches 1 and 2 among the *C. albicans*, *C. glabrata*, *C. tropicalis*, and *C. pseudotropicalis* strains, respectively, but wider zones of inhibition were recorded in batch 2 antifungal drugs \[Tables [1](#T0001){ref-type="table"}--[4](#T0004){ref-type="table"}\]. ::: {#T0001 .table-wrap} Table 1 ::: {.caption} ###### Phenotypic antimycotic susceptibility/resistance profiles of *Candida albicans* strains associated with candidiasis to two batches of same antimycotic drugs ::: Candida strains Antimycotic agents (μg ml^-1^) ------------------------------------ -------------------------------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- *C. albicans* C2 B1 R R R R R 20.0 R 1 (14.3) B2 20.0 25.0 18.0 10.0 24.0 30.0 24.0 7 (100) *C. albicans* C29 B1 10.0 10.0 20.0 10.0 10.0 20.0 R 6 (85.7) B2 25.0 26.0 28.0 24.0 30.0 32.0 26.0 7 (100) *C. albicans* C51 B1 10.0 R 10.0 R 20.0 20.0 10.0 5 (71.4) B2 R 28.0 20.0 25.0 36.0 30.0 28.0 6 (85.7) *C. albicans* C23 B1 10.0 20.0 10.0 20.0 10.0 20.0 R 6 (85.7) B2 14.0 18.0 25.0 14.0 R 20.0 30.0 6 (85.7) *C. albicans* 2C2 B1 R 20.0 20.0 R 20.0 25.0 20.0 5 (71.4) B2 28.0 26.0 30.0 26.0 32.0 28.0 28.0 7 (100) *C. albicans* GC2 B1 10.0 10.0 20.0 10.0 20.0 20.0 R 6 (85.7) B2 30.0 24.0 30.0 28.0 30.0 35.0 16.0 7 (100) *C. albicans* 6C1 B1 10.0 10.0 10.0 10.0 10.0 10.0 10.0 7 (100) B2 30.0 24.0 28.0 24.0 28.0 35.0 20.0 7 (100) Total/(%) susceptibility B1 5 (71.4) 5 (71.4) 6 (85.7) 4 (57.1) 6 (85.7) 7 (100) 3 (42.9) B2 6 (85.7) 7 (100) 7 (100) 7 (100) 6 (85.7) 7 (100) 7 (100) S/R 3 (42.9) 2 (28.6) 1 (14.3) 3 (42.9) 2 (28.6) \- (0.0) 4 (57.1) ≤ 5.0 mm 1 (14.3) 1 (14.3) \- (0.0) \- (0.0) \- (0.0) 1 (14.3) \- (0.0) ≥10.0 mm 3 (42.9) 3 (42.9) 5 (71.4) 3 (42.9) 5 (71.4) 5 (71.4) 2 (28.6) [\*](#T000F1){ref-type="table-fn"} \- (0.0) \- (0.0) \- (0.0) \- (0.0) \- (0.0) 1 (14.3) \- (0.0) \* same values in corresponding antimycotics of both batches. Values of zones of inhibition are means of duplicates *P* = 0.016646. B1, Batch 1; B2, Batch 2; AF1/AM2, mycoten tablets; AF2/AM9, mycoten cream; AF3/AM5, canesten tablets (clotrimoxazole); AF4/AM8, canesten cream (clotrimoxazole); AF6/AM1, tetradox (doxycycline); AF8/AM4, mycostatine (nystatin); AF9 AM7, flagyl (metronidazole). S/R, corresponding antimycotics susceptible in one batch but resistant in the other batch; ≤5.0 mm, corresponding antimycotics having zones of inhibition differences of ≤5.0 mm in diameter; ≥10.0 mm, corresponding antimycotics having zones of inhibition differences of ≥10.0 mm in diameter; ::: ::: {#T0002 .table-wrap} Table 2 ::: {.caption} ###### Phenotypic antimycotic susceptibility/resistance profiles of *Candida glabrata* strains associated with candidiasis to two batches of same antimycotic drugs ::: *Candida* strains Antimycotic agents (μg ml^-1^) ------------------------------------ -------------------------------- ---------- ---------- ---------- ----------- ---------- ---------- ---------- ---------- *C. glabrata* C3 B1 R R R R 24.0 20.0 R 2 (28.6) B2 R R R R 20.0 32.0 R 2 (28.6) *C. glabrata* C6 B1 10.0 10.0 20.0 10.0 15.0 10.0 R 6 (85.7) B2 35.0 R R 24.0 32.0 36.0 28.0 5 (71.4) *C. glabrata* C12 B1 R 20.0 R 10.0 R 20.0 R 3 (42.9) B2 R R R R 15.0 28.0 R 2 (28.6) *C. glabrata* C27 B1 R R 20.0 20.0 20.0 25.0 20.0 5 (71.4) B2 29.0 32.0 28.0 30.0 35.0 30.0 28.0 7 (100) *C. glabrata* 34 B1 25.0 25.0 20.0 25.0 20.0 20.0 20.0 7 (100) B2 28.0 20.0 24.0 30.0 30.0 35.0 28.0 7 (100) *C. glabrata* 42 B1 20.0 20.0 20.0 20.0 20.0 20.0 R 6 (85.7) B2 R R R R 18.0 30.0 R 2 (28.6) *C. glabrata* 43 B1 15.0 15.0 15.0 15.0 25.0 20.0 15.0 7 (100) B2 28.0 34.0 28.0 30.0 35.0 24.0 38.0 7 (100) *C. glabrata* 61 B1 10.0 10.0 10.0 10.0 R 20.0 R 5 (71.4) B2 26.0 27.0 30.0 26.0 35.0 35.0 24.0 7 (100) *C. glabrata* 1TC B1 20.0 25.0 20.0 20.0 20.0 20.0 R 6 (85.7) B2 26.0 24.0 28.0 22.0 30.0 34.0 26.0 7 (100) *C. glabrata* BC2 B1 20.0 R 10.0 10.0 10.0 20.0 R 5 (71.4) B2 30.0 27.0 30.0 28.0 34.0 30.0 24.0 7 (100) *C. glabrata* 4C1 B1 R 10.0 R 10.0 R 20.0 R 3 (42.9) B2 32.0 28.0 26.0 26.0 30.0 32.0 24.0 7 (100) Total/(%) Susceptibility B1 7 (63.6) 8 (72.7) 8 (72.7) 10 (90.9) 8 (72.7) 11 (100) 3 (27.3) B2 8 (72.7) 7 (63.6) 7 (63.6) 8 (72.7) 11 (100) 11 (100) 8 (72.7) S/R 3 (27.3) 5 (45.5) 3 (27.3) 2 (18.1) 3 (27.3) \- (0.0) 5 (45.5) ≤5.0 mm 1 (9.1) 2 (18.2) 1 (9.1) 2 (18.1) 2 (18.1) 2 (18.1) \- (0.0) ≥10.0 mm 4 (36.4) 3 (27.3) 3 (27.3) 6 (54.5) 6 (54.5) 8 (72.7) 1 (9.1) [\*](#T000F2){ref-type="table-fn"} 2 (18.1) 1 (9.1) 2 (18.1) 1 (9.1) \- (0.0) 1 (14.1) 3 (27.3) \* same values in corresponding antimycotics of both batches. Values of zones of inhibition are means of duplicates *P* = 0.238954. B1, Batch 1; B2, Batch 2; AF1/AM2, mycoten tablets; AF2/AM9, mycoten cream; AF3/AM5, canesten tablets (clotrimoxazole); AF4/AM8, canesten cream (clotrimoxazole); AF6/AM1, tetradox (doxycycline); AF8/AM4, mycostatine (nystatin); AF9 AM7, flagyl (metronidazole). S/R, corresponding antimycotics susceptible in one batch but resistant in the other batch; ≤5.0 mm, corresponding antimycotics having zones of inhibition differences of ≤5.0 mm in diameter; ≥ 10.0 mm, corresponding antimycotics having zones of inhibition differences of ≥ 10.0 mm in diameter; ::: ::: {#T0003 .table-wrap} Table 3 ::: {.caption} ###### Phenotypic antimycotic susceptibility/resistance profiles of *Candida pseudotropicalis* strains associated with candidiasis to two batches of same antimycotic drugs ::: *Candida* strains Antimycotic agents (μg ml^-1^) ------------------------------------ -------------------------------- ---------- ----------- ---------- ----------- ---------- ---------- ---------- ----------- *C. pseudotropicalis* 16 B1 25.0 30.0 25.0 30.0 15.0 15.0 R 6 (85.7) B2 34.0 32.0 30.0 24.0 28.0 34.0 30.0 7 (100) *C. pseudotropicalis* 25 B1 10.0 20.0 10.0 20.0 10.0 20.0 R 6 (85.7) B2 28.0 24.0 30.0 26.0 26.0 32.0 30.0 7 (100) *C. pseudotropicalis* 48 B1 R 20.0 20.0 R R R 10.0 3 (42.9) B2 R R R R R 22.0 R 1 (14.3) *C. pseudotropicalis* 65 B1 10.0 20.0 10.0 20.0 10.0 20.0 R 6 (85.7) B2 20.0 25.0 R R 15.0 25.0 R 4 (57.1) *C. pseudotropicalis* X7C B1 10.0 10.0 10.0 10.0 10.0 20.0 R 6 (85.7) B2 R R R R 32.0 30.0 R 2 (.27.6) *C. pseudotropicalis* 6C2 B1 R R R R R R R \- (0.0) B2 26.0 28.0 24.0 28.0 34.0 32.0 28.0 7 (100) Total/(%) Susceptibility B1 4 (66.6) 5 (83.3) 5 (83.3) 4 (66.6) 4 (66.6) 4 (66.6) 1 (16.7) B2 4 (66.6) 4 (66.6) 3 (50.0) 3 (50.0) 5 (83.3) 6 (100) 3 (50.0) S/R 2 (33.3) 3 (50.0) 4 (66.7) 3 (50.0) 1 (16.7) 2 (33.3) 4 (66.7) ≤5.0 mm \- (0.0) 3 (50.0) 1 (16.7) \- (0.0) 1 (16.7) 1 (16.7) \- (0.0) ≥10.0 mm 2 (33.3) \- (00.0) 1 (16.7) \- (00.0) 3 (50.0) 3 (50.0) \- (0.0) [\*](#T000F3){ref-type="table-fn"} 1 (16.7) \- (0.0) \- (0.0) 1 (16.7) 1 (16.7) 1 (16.7) 2 (33.3) \* same values in corresponding antimycotics of both batches. Values of zones of inhibition are means of duplicates *P* =0.372246. B1, Batch 1; B2, Batch 2; AF1/AM2, mycoten tablets; AF2/AM9, mycoten cream; AF3/AM5, canesten tablets (clotrimoxazole); AF4/AM8, canesten cream (clotrimoxazole); AF6/AM1, tetradox (doxycycline); AF8/AM4, mycostatine (nystatin); AF9 AM7, flagyl (metronidazole). S/R, corresponding antimycotics susceptible in one batch but resistant in the other batch; ≤5.0 mm, corresponding antimycotics having zones of inhibition differences of ≤5.0 mm in diameter; ≥10.0 mm, corresponding antimycotics having zones of inhibition differences of ≥10.0 mm in diameter; ::: ::: {#T0004 .table-wrap} Table 4 ::: {.caption} ###### Phenotypic antimycotic susceptibility/resistance profiles of Candida tropicalis strains associated with candidiasis to two batches of same antimycotic drugs ::: *Candida* strains Antimycotic agents (μg ml^-1^) ------------------------------------ -------------------------------- ---------- ---------- ---------- ----------- ----------- ---------- ---------- ---------- *C. tropicalis* C8 B1 R R R 20.0 R 20.0 R 2 (28.6) B2 R R R R 28.0 35.0 30.0 3 (42.9) *C. tropicalis* C9 B1 R R 20.0 R 20.0 10.0 R 3 (42.9) B2 28.0 30.0 28.0 24.0 28.0 34.0 26.0 7 (100) *C. tropicalis* C14 B1 25.0 20.0 R 25.0 10.0 10.0 R 5 (71.4) B2 30.0 28.0 30.0 26.0 30.0 30.0 28.0 7 (100) *C. tropicalis* C20 B1 R 20.0 R R R 20.0 R 2 (28.6) B2 R R R R 30.0 23.0 20.0 3 (42.9) *C. tropicalis* 26 B1 10.0 R 10.0 20.0 R 20.0 R 4 (57.1) B2 30.0 28.0 26.0 28.0 32.0 38.0 18.0 7 (100) *C. tropicalis* 40 B1 30.0 30.0 30.0 30.0 30.0 30.0 30.0 7 (100) B2 26.0 28.0 30.0 22.0 30.0 34.0 24.0 7 (100) *C. tropicalis* 53 B1 10.0 10.0 20.0 10.0 10.0 25.0 R 6 (85.7) B2 R R 15.0 R 20.0 30.0 R 3 (42.9) *C. tropicalis* 10C B1 10.0 10.0 10.0 20.0 10.0 20.0 R 6 (85.7) B2 R R R R R 25.0 R 1 (14.3) *C. tropicalis* 2TC B1 10.0 R 10.0 10.0 10.0 20.0 R 5 (71.4) B2 R 28.0 R R R 30.0 R 2 (28.6) *C. tropicalis* HC B1 10.0 10.0 10.0 10.0 10.0 20.0 10.0 7 (100) B2 28.0 26.0 26.0 28.0 30.0 32.0 18.0 7 (100) *C. tropicalis* 6C B1 10.0 10.0 10.0 10.0 10.0 20.0 R 6 (85.7) B2 30.0 24.0 28.0 28.0 32.0 31.0 27.0 7 (100) *C. tropicalis* 9C B1 R 10.0 10.0 15.0 R 20.0 10.0 5 (71.4) B2 30.0 24.0 26.0 28.0 32.0 30.0 26.0 7 (100) Total/(%) Susceptibility B1 8 (66.6) 8 (66.6) 4 (66.6) 10 (83.3) 8 (66.6) 12 (100) 3 (25.0) B2 7 (58.4) 8 (66.6) 8 (66.6) 7 (58.4) 10 (83.3) 12 (100) 9 (75.0) R/S 5 (41.6) 6 (50.0) 3 (25.0) 5 (41.6) 6 (50.0) \- (0.0) 6 (50.0) ≤5.0 mm 2 (16.7) 1 (8.3) 2 (16.7) 1 (8.3) \- (0.0) 4 (33.3) \- (0.0) ≥10.0 mm 3 (25.0) 3 (25.0) 4 (33.3) 3 (25.0) 4 (33.3) 8 (66.7) 1 (8.3) [\*](#T000F4){ref-type="table-fn"} 2 (16.7) 1 (8.3) 3 (25.0) 1 (8.3) 1 (8.3) \- (0.0) 3 (25.0) \* same values in corresponding antimycotics of both batches. Values of zones of inhibition are means of duplicates *P* = 0.409089. B1, Batch 1; B2, Batch 2; AF1/AM2, mycoten tablets; AF2/AM9, mycoten cream; AF3/AM5, canesten tablets (clotrimoxazole); AF4/AM8, canesten cream (clotrimoxazole); AF6/AM1, tetradox (doxycycline); AF8/AM4, mycostatine (nystatin); AF9 AM7, flagyl (metronidazole). S/R, corresponding antimycotics susceptible in one batch but resistant in the other batch; ≤5.0 mm, corresponding antimycotics having zones of inhibition differences of ≤5.0 mm in diameter; ≥10.0 mm, corresponding antimycotics having zones of inhibition differences of ≥10.0 mm in diameter; ::: Among the *C. albicans*, just 14.3% of the strains had same susceptibility/resistance profiles toward the same test antifungal agents in both batches 1 and 2, while up to 57.1% of the strains were susceptible in the first batch but resistant to corresponding antifungals in the second batch. Between 28.6% and 71.4% of the *C. albicans* had difference of ≥10.0 mm (zones of inhibition) as the recorded values in corresponding antimycotic agents \[[Table 1](#T0001){ref-type="table"}\]. Only a maximum of 27.3% of *C. glabrata* strains had same susceptibility/resistance profiles, with as high as about 73.0% of the strains having differences of ≥10.0 mm (zones of inhibition) among corresponding antimycotic agents, while between 18.1% and 45.5% were susceptible in a batch and resistant to corresponding antifungals in the second batch \[[Table 2](#T0002){ref-type="table"}\]. As shown in [Table 3](#T0003){ref-type="table"}, as low as 16.7-33.3% of the *C. pseudotropicalis* strains had same susceptibility/resistance profiles, while about 50.0% of the strains had differences of ≥10.0 mm (zones of inhibition) among corresponding antimycotic agents, with as high as 66.7% of the strains being susceptible in a batch but resistant to corresponding antifungals in the second batch. *C. pseudotropicalis* 6C2 was resistant against all test antifungal agents in the batch 1 but susceptible towards all the test antifungals in batch 2. [Table 4](#T0004){ref-type="table"} shows that between 8.3% and 25.0% of the *C. tropicalis* strains had same susceptibility/resistance profiles toward the test antifungal agents in both batches, while up to 50.0% of the strains were susceptible in one batch but resistant to corresponding antifungals in the second batch. As high as 66.74% of the *Candida* strains had differences of ≥10.0 mm (zones of inhibition) toward the corresponding antimycotic agents in the other batch. Raw nonstatistical data indicated that most of the *Candida* strains were different in their susceptibility/resistance profiles toward the same antimycotic agents in the two batches, i.e., the *in vitro* susceptibility tests on the *Candida* strains revealed that the inhibitory activities of the two batches of antimycotic agents were significantly different from each other. Some of the *Candida* strains like *C. glabrata* C27, C43, C61, 1TC, BC2, 4C1; *C. tropicalis* C9, C14, C26, C53, 10C, 2TC, HC, 6C; *C. pseudotropicalis* X7C, 6C2 were found to have well-defined differences in their susceptibility profiles toward the two batches of same antifungal agents, meaning that the two antimycotic agents in batches B1 and B2 had different inhibitory effects or potency on the *Candida* strains. Relatively higher susceptibility rates were recorded among the antifungals in batch B2 compared to batch B1 - *C. albicans* (95.9%; 73.5%), *C. glabrata* (77.9%; 71.4%), *C. tropicalis* (66.7%; 64.3%), and *C. pseudotropicalis* (72.6%; 63.1%); while the statistical results indicated the recorded susceptibility values as *C. albicans* (p=0.016646), *C. glabrata* (0.238954), *C. tropicalis* (0.372246), and *C. pseudotropicalis* (0.409089), respectively \[Tables [1](#T0001){ref-type="table"}--[4](#T0004){ref-type="table"}\]. Discussion {#sec1-3} ========== Vaginal discharge is the symptom that most often prompts a woman to consult a physician in order to determine the presence of an infection, while diagnosis is usually based on evaluation of the vaginal ecosystem and demonstration of the presence of the suspected microorganisms.\[[@CIT22]\] In the study of Wathne *et al*\[[@CIT23]\] and the review of Sobel,\[[@CIT24]\] on the epidemiology, diagnosis, and therapy of vaginitis, it was reported that vulvovaginal symptoms are extremely common and can cause extreme distress for some patients, especially those with recurrent symptoms.\[[@CIT25]\] Women, therefore, often seek medical care for vaginal complaints.\[[@CIT26]\] *Candida* infectious complications in pregnancy and delivery are still very serious problems in obstetrical, gynecological, and neonatological practices, and the presence of vaginal infections during pregnancy has also been linked to low birth weight and obstetric disorders.\[[@CIT22][@CIT27]\] Similarly, *C. albicans*, *C. glabrata*, *C. tropicalis*, and *C. pseudotropicalis*, which are among the most implicated species in vulvovaginal *Candida*sis were also recovered from symptomatic females in this study. It is, therefore, very important that vulvovaginal *Candida*sis must be promptly treated. Several antifungal agents are available for the treatment of Candidasis,\[[@CIT28]\] but there have been reports of antagonism between antifungal compounds and isolates of *Candida* spp.\[[@CIT29]--[@CIT31]\] The *in vitro* activities of antifungal agents, however, varied among various studies,\[[@CIT32]--[@CIT35]\] with differing spectra of activities against antifungal agents, while *in vitro* testing has similarly revealed that there are clear differences among the various non-albicans *Candida* (NAC) and *C. albicans* in their susceptibility to specific antifungal drugs. It is also generally believed that there is a significant increase in the resistance of *Candida* spp. toward antifungal agents in recent times.\[[@CIT35][@CIT36]\] It was, however, observed in the current study that according to the overall results obtained, most of the *Candida* strains were susceptible to the test antifungal agents, especially mycostatine, tetradox, canesten cream, and mycoten tablet, which is in accordance with some previous studies that recorded relatively higher susceptibility rates toward certain antifungal agents by some *Candida* strains implicated in vulvovaginal Candidasis.\[[@CIT35]--[@CIT39]\] The relatively high differences in the susceptibility/resistance result patterns obtained from the two batches of corresponding antifungal drugs in this study are of serious significance and also corroborated the hazardous effect of inconsistent drug production under different production batches, which must be taken into consideration when screening and choosingantifungal agents for fungal therapy. In bioequivalence studies, the goal of testing is to determine if the drugs are functionally equivalent, due to the fact that a drug may be chemically equivalent but not clinically equivalent.\[[@CIT40]\] As an example, routine antibiotic susceptibility testing has been advocated as an essential selection criterion for potential probiotic Candidates but in a previous study,.\[[@CIT41]\] while determining the phenotypic antibiotic susceptibility of 54 potential probiotic Candidates to the same antibiotics of different production batches, it was found that the overall percentage differences among the probiotic Candidates to the same test antibiotics of different production batches, manufactured by the same company, were between 53.9% and 76.5%. The implication is that if one batch of antibiotics had been used, some potential probiotic Candidates would have been eliminated by the resistance selection criterion. Two drugs are considered pharmaceutical equivalents when they contain the same chemically active ingredient(s) and are identical in dosage form and strength,\[[@CIT42]\] but pharmaceutical equivalence may be affected by variations in inert ingredients, such as production of ingredients that vary in quality, and by batch and manufacturing methods.\[[@CIT43]\] Another factor which affects generic quality is the international buyouts and diversification, which allows the combination of questionable ingredients into generic production.\[[@CIT44]\] Most of the times, once a drug has been approved by the regulating bodies like FDA or NAFDAC, manufacturers sometimes make changes to the formulations, which were originally submitted for screening.\[[@CIT43][@CIT44]\] Although drug quality is currently receiving renewed international attention\[[@CIT45]\] but in spite of an increase in public awareness of the existence of counterfeit and substandard drugs over the past decade,\[[@CIT46]\] it is quite unfortunate that the menace of counterfeit and substandard drugs is being increasingly reported in developing countrieslike Nigeria due to ineffective drug regulations.\[[@CIT10][@CIT47]\] There is growing universal concern regarding counterfeit medications, and in particular, counterfeit antimicrobial drugs are a threatto public health with many devastating consequences for patients, such as, increased mortality and morbidity, and emergence of drug resistance. In addition, physicians treating these patients lose their confidencein the medications used due to report of high levels of resistance.\[[@CIT48]\] Usually, the way products are manufactured depends on the quantity required but the inconsistencies in activities associated with batch production of clinical drugs may be due to the fact that it is not a continuous production, since there is in-between stoppage and reconfiguration of equipment during production batches, especially as regards the downtime (idle time between batches) and cycle time (time between consecutive batches).\[[@CIT49]\] In the study of Khabriev and Yagudina,\[[@CIT50]\] while assessing the general state-of-the-art in the quality of domestic drugs on the Russian market, it was established that about 16.5 thousand of the drug batches rejected were recalled from the market over the period from 1994 to 2002 with the total number of rejected batches increasing from 660 in 1994 to 1107 in 2002. This is not usually the case in Nigerian situation; therefore, it is very difficult to regulate drug batches that do not meet the standard criteria. Assessment of clinical drugs and recall of low-quality or adulterated drugs in Nigeria is minimal and not regular due to some faults in logistics, such as consideration of the production batches of drugs prior to registration by the regulating bodies. The fact that none of the *Candida* strains had entirely the same (100%) susceptibility profiles in just two batches of corresponding antimycotic drugs, while as low as 8.3-33.3% of the *Candida* strains had similar susceptibility/resistance profiles toward the test antifungal agents in both batches, confirms that there is serious clinical and health implications as regards the inconsistency in different production batches of such antimycotic drugs. Conflicting inhibitory activities of corresponding antimycotics could be a threatto public health with consequences for patients, since prescription could be made based on the assumption that inhibition by an antimycotic drug in a batch would have the same effect by corresponding drug(s) in other batch(es). Similarly, clinical implication can be deduced, in that reports of resistance/susceptibility are mostly not the same in corresponding drugs of different batches, which will ultimately lead to errors in documented findings. It is, therefore, very important to assess the consistency of different batches of drugs with regards to potency and when understudying the susceptibility or resistance patterns of the pathogens, especially in developing countries, where most drugs in circulation are adulterated. Similarly, it is of adequate importance that every production batch of drugs in Nigeria be consistently screened by regulating bodies like NAFDAC before approval for sales and administration of such drugs. There must be complete documented investigations into the failure of drug batches, which do not meet the expected specifications. It is also important that policies are put in place to ascertain that clinical drugs are properly screened with adequate investigations into causes of manufacturing problems. **Source of Support:** Nil, **Conflict of Interest:** None declared

PubMed Central

2024-06-05T04:04:19.960731

2011-01-01

PMC3053516

Good prescribing practice is an essential part of rational drug use.\[[@CIT1][@CIT2]\] A prescription audit, therefore, is a useful method to assess the doctors' contribution to rational use of drugs in a country. Different aspects of prescribing patterns in many institutions, in different countries like India,\[[@CIT3]--[@CIT7]\] Pakistan,\[[@CIT8][@CIT9]\] Nepal,\[[@CIT10]\] and Sri Lanka\[[@CIT11][@CIT12]\] have been studied. We found that there was limited data available about the inappropriate prescription practices such as polypharmacy and over-usage of antibiotics and injections in the government hospitals in Sri Lanka. In the absence of more data in this field, we decided to study the degree of a healthcare worker's adherence to principles of rational use of medicine in the public sector initially in Galle, Sri Lanka. We planned to measure the prescribing indicators including polypharmacy, prescription of generics and essential drugs, injections / antibiotic usage, patient care indicators, and the average consultation time, to assess the adherence to the rational drug policy. We decided to use the specific parameters like average consultation time (ACT), average number of drugs per encounter (ANDE), percentage of drugs prescribed by generic name (PDPG), percentage of encounters with an antibiotic prescribed (PAP), percentage of encounters with an injection prescribed (PIP), and percentage of drugs prescribed from the essential drugs list or formulary (PEDL for our analysis because these were the standard parameters recommended by the World Health Organization (WHO) for analysis of drug-use patterns (13). It was determined to address the recognized issues at the local and national level by giving our recommendations for improvement. Materials and Methods {#sec1-1} ===================== It was a study carried out at the TH, Karapitya, GH in Balapitiya and DH in Akuressa for six months. All patients attending the Outpatient Departments in the morning clinics in theseh were considered and included in our study. The ethics and review committee of the institution approved the study. Five hundred and ninety encounters were collected from patients who attended the OPD, by the trained medical students and medical officers. Data collectors were pre-trained by a principle investigator, in an effort to ensure uniformity in data collection. This prescriber care assessment was done by doing exist-patient interview using a pretested structured observations questionnaire. The following measuring tools were used to assess the degree of prescriber care and the errors in our health facilities. ACT, ANDE, PDPG, PAP, PIP, and PEDL were calculated for three different hospitals according to the WHO criteria.\[[@CIT13]\] Results {#sec1-2} ======= Patient diagnosis and prescriber identity were absent in all the prescriptions, although the signature was present in almost all. Age was not mentioned only in 0.61% patients, but sex was not mentioned in any of the 327 prescriptions studied. Duration of treatment and frequency of drug administration were also not mentioned in 0.61% of the prescriptions \[[Table 1](#T0001){ref-type="table"}\]. ::: {#T0001 .table-wrap} Table 1 ::: {.caption} ###### Common problems noted in the prescriptions analyzed in TH, GH and DH in Galle district ::: Problems encountered Percentage of prescriptions ---------------------------------- ----------------------------- Diagnosis is absent 100 Doctors identity is absent 100 Age is not mentioned 0.61 Sex is not mentioned 100 Duration of the treatment 0.61 Frequency of drug administration 0.61 ::: Significant differences in average consultation time, in different hospital set-ups, in the Galle district {#sec2-1} ---------------------------------------------------------------------------------------------------------- We analyzed 263 prescribers in government hospitals using the pretested structured observations questionnaire used to assess patient care by WHO. The average consultation time (ACT) of three different hospitals is given in [Figure 1](#F0001){ref-type="fig"}. ::: {#F0001 .fig} Figure 1 ::: {.caption} ###### Average consultation time (ACT) in different hospitals using ANOVA. Shows the changes in ACT in different hospitals and we observed the ACT value to be 2.31±0.16 in TH, 2.17±0.08 in GH, and 0.83±0.05 in DH, respectively, M±SEM. The three means are significantly different (*P*≤0.05) :::  ::: Average number of drugs per encounter {#sec2-2} ------------------------------------- Mean ± SEM of average number of drugs per encounter (ANDE) is given in [Figure 2](#F0002){ref-type="fig"}. According to the WHO, 2008, recommended figures (1.6 -- 1.8), our ANDE was very high. We further noted that ANDE was high in our DH and TH and relatively low in GH. ::: {#F0002 .fig} Figure 2 ::: {.caption} ###### Average number of drugs per encounter (ANDE) in different hospitals using ANOVA. Shows that the changes in ANDE in different hospitals and the ANDE value are 3.24±0.08 in TH, 2.88±0.10 in GH, and 3.26±0.17 in DH, respectively, M±SEM. The three means are significantly different (*P*≤0.05) :::  ::: Percentage of drugs prescribed by generic name {#sec2-3} ---------------------------------------------- [Table 2](#T0002){ref-type="table"} shows average number of prescribed generics per prescription in different hospitals. Our results showed that percentage of drugs prescribed by generic name (PDPG) was lower than the WHO recommended values and use of generic names in the government prescriptions was significantly high. It was also noted that PDPG was high in TH and GH \[[Figure 3](#F0003){ref-type="fig"}\]. It was low in DH where only the basic facilities were available. We also found that there was no significant difference in the average number of drugs prescribed by generic name among the three groups of hospitals (*P*≤0.05). ::: {#F0003 .fig} Figure 3 ::: {.caption} ###### Percentage of drugs prescribed by generic name (PDPG) in different hospitals using ANOVA. [Figure 3](#F0003){ref-type="fig"} shows the changes in PDPG in different hospitals and we observed the PDPG value as 78% in TH, 78% in GH, and 71% in DH, respectively :::  ::: ::: {#T0002 .table-wrap} Table 2 ::: {.caption} ###### Average number of generics per prescription ::: Hospital Mean ± SE ------------------------ ------------- TH (Teaching Hospital) 2.54 ± 0.08 GH (General Hospital) 2.26 ± 0.10 DH (District Hospital) 2.31 ± 0.14 ::: Percentage of encounters with an antibiotic prescribed {#sec2-4} ------------------------------------------------------ Different values of Percentage of encounters with an antibiotic prescribed (PAP) in TH, GH, and DH are given in [Figure 4](#F0004){ref-type="fig"}. We found that the percentage of antibiotics was very much higher than the WHO recommended values (20 − 26.8%). It was observed that PAP was comparatively low (47%) in TH and high (80%) in DH. ::: {#F0004 .fig} Figure 4 ::: {.caption} ###### Percentage of encounters with an antibiotic prescribed (PAP) in different hospitals using ANOVA. It shows the changes in PAP in different hospitals and we observed the PAP values as 47% in TH, 46%in GH and 80% in DH, respectively :::  ::: In addition to the percentages of antibiotic usage, three means of the average number of antibiotics per encounter in TH (47 + 0.04), GH (46 + 0.07), and DH (80 + 0.09) were also compared. We found that the DH mean was significantly different from GH and TH (*P*≤0.05, [Table 3](#T0003){ref-type="table"}). ::: {#T0003 .table-wrap} Table 3 ::: {.caption} ###### Average number of antibiotics per prescription ::: Hospital Mean ± SE ------------------------ ------------- TH (Teaching Hospital) 0.55 ± 0.05 GH (General Hospital) 0.58 ± 0.07 DH (District Hospital) 1.11 ± 0.12 ::: Percentage of encounters with an injection prescribed (PIP) {#sec2-5} ----------------------------------------------------------- We found that percentage of injections was lower in our study than the WHO recommended values (13.4 −24.1%). Use of injections in government prescriptions was significantly low. In addition to percentages, we further analyzed the average number of injections per encounter in these hospitals \[[Figure 5](#F0005){ref-type="fig"}\], and the study showed that there was no significant difference in the average number of injections per prescription among the three groups of hospitals (*P*≤0.05, [Table 4](#T0004){ref-type="table"}). ::: {#T0004 .table-wrap} Table 4 ::: {.caption} ###### Average number of injections per a prescription ::: Hospital Mean SE ------------------------ ------------- TH (Teaching Hospital) 0.03 ± 0.01 GH (General Hospital) 0.04 ± 0.02 DH (District Hospital) 0.06 ± 0.04 ::: ::: {#F0005 .fig} Figure 5 ::: {.caption} ###### Percentage of encounters with an injection prescribed (PIP) in different hospitals using ANOVA. [Figure 5](#F0005){ref-type="fig"} shows the changes in PIP in different hospitals and we observed the PIP values to be 3% in TH, 4% in GH, and 6% in DH, respectively :::  ::: Percentage of drugs prescribed from essential drugs list or formulary {#sec2-6} --------------------------------------------------------------------- [Table 5](#T0005){ref-type="table"} shows average number of drugs prescribed from essential drug list three different hospitals. [Figure 6](#F0006){ref-type="fig"} shows the percentages of Percentage of drugs prescribed from essential drugs list or formulary (PEDL) in TH, GH, and DH, and we found that the percentages of drugs prescribed from the essential drugs list or formulary were compatible with the WHO values (100%). Our PDEL in the government prescriptions was significantly high. ::: {#T0005 .table-wrap} Table 5 ::: {.caption} ###### Average number of drugs written from EDL or formulary per a prescription ::: Hospital Mean ± SE ------------------------ ------------- TH (Teaching Hospital) 3.14 ± 0.08 GH (General Hospital) 2.87 ± 0.10 DH (District Hospital) 3.22 ± 0.17 ::: ::: {#F0006 .fig} Figure 6 ::: {.caption} ###### Percentage of drugs prescribed from essential drugs list or formulary (PEDL) in different hospitals using ANOVA. [Figure 6](#F0006){ref-type="fig"} shows the changes in PEDL in different hospitals and we observed the PEDL value to be 97% in TH, 100% in GH, and 99% in DH, respectively :::  ::: In addition to the percentage of essential drugs prescribed from the essential drug list or formulary, the average number of drugs used by the EDL or Formulary per encounter in these hospitals was also studied. We also found that there was no significant difference in the average number of drugs prescribed from the EDL or Formulary among the three groups of hospitals (*P*≤ 0.05). Discussion {#sec1-3} ========== The drug prescribing pattern needs to be evaluated from time to time.\[[@CIT14]\] Our health system mainly consists of government hospitals in all places, and treatment and other health services are given free to the population. There is a minor population who seeks private sector service, mainly in the city limits. Therefore the healthcare system and services provided by the government system bear a majority of the weight in the population when compared to other countries, where both private and public sectors have an equal burden. We found that the consultation time of a prescriber in government hospitals is very much shorter when compared with that in private practice.\[[@CIT12][@CIT15]\] ACT in Nepal was higher than in our study (6.03 ± 3.34 minutes). The short ACT can be explained by inadequate patient doctor ratio in our public sector free health service. We also found various prescriber problems noted on the 327 prescriptions. These prescriber problems were identified and listed in [Table 1](#T0001){ref-type="table"}. Although the diagnosis was an essential component in a prescription, it was not written in 100% of the prescriptions, in all the screened hospitals. This was one of the common errors seen in the government hospitals in Sri Lanka.. This finding again could be justified with the large number of patients alloted to a single medical officer. These issues should be critically addressed by the government to reduce all the complications, like adverse reactions and high health cost, and the morbidity and mortality rate in the country. Our results showed that both frequency of drug administration and duration of treatment were indicated satisfactorily, 100, 100, and 94.29% of the prescriptions in TH, GH, and DH, respectively. We also found that the duration of the drug treatment was also mentioned in 100, 100, and 94.29% of the prescriptions in TH, GH, and DH, respectively. We analyzed the degree of adherence to the use of EDL and STG in government hospitals by using the WHO standard in the recommended evaluation techniques. Our study showed that the use of essential drugs for prescriptions were parallel with the WHO standards. in the studied hospital categories. This practice could be due to the government health policy of EDL, and it was helpful to reduce the national health budget of the country and hence improve the cost-effective strategies. This study further showed evidence for a great need to improve prescription writing, as evidenced by our data, where in some of the prescriptions, duration of treatment, and frequency of drug administration were not given. This problem coupled with polypharmacy could result in less safe, more expensive. and irrational prescribing.\[[@CIT16]\] We have noticed that doctor identity is absent in all the prescriptions and it will be a serious dilemma in incidences where the prescriber is needed for an emergency. We feel that this is because of complete negligence. Despite the work load, the importance of this should be stressed, as both the patient's and doctor's identities are vital on a prescription in Sri Lanka. We further noted that prescription of injections at the OPDs set up in the government hospitals was very low, but parallel to the WHO standard\[[@CIT17]\] and could be explained by the high turnover of patients in the OPD setup. It was a similar case with the establishment of emergency treatment units at the OPD. In contrast to our results, in the research conducted in the Manipal teaching hospital, the percentage of injections was 5.21% of the encounters.\[[@CIT3]\] According to them, the percentage is more than that in the Pakistani study.\[[@CIT3]\] In contrast to that, the prescription for antibiotic use in our hospitals is very high, in all hospitals types, when compared to the WHO values, and this should be addressed as early as possible, to prevent adverse reactions due to antibiotic use. The average number of drugs per prescription is an important index to review, to plan an educational intervention in prescribing practices. We found that the number of drugs per encounter in all hospital categories was higher than the WHO recommended values (1.6 -- 1.8). Similar results were found in Cambodia, Ethiopia, Morocco, Tanzania, Zimbabwe, and Nepal.\[[@CIT14]\] Other hospital-based studies\[[@CIT10]\] in India also reported figures of three to five drugs per prescription, which was quite similar to ours.\[[@CIT11]\] Various reasons can account for this deviation from the recommended WHO values. It can be due to unrealistic expectations, quick relief from patients, common practice of irrational drug combinations, unnecessary use of vitamins, and aggressive medicine promotions. This concept, which is called as polypharmacy, can lead to increased risk of drug interactions,\[[@CIT16]\] increased hospital cost, and errors in prescribing.\[[@CIT18]--[@CIT21]\] Our study had some limitations. We did not plan to interview patients for their knowledge of taking the correct doses. This is important, because the absence of the knowledge on correct drug administration will lead to poor achievements of rational practice. This study needs to be followed up by prescriber education on rational drug use, by means of short-term training sessions, including a briefing on proper prescription writing. We further plan to do a reaudit, to measure the impact of intervention. We are happy to see that a maximum number of drugs from the EDL has been used in all three hospitals in the Galle district and we understand that it is a common practice in most of the government hospitals in Sri Lanka. This is probably because procurement has to be done according to successful implementing plans and procurement policies. One research shows only 7.61% of the prescribed generic drugs in a private hospital in their study. We also feel that usage of EDL might be low in the private sector in Sri Lanka, however, this needs to be investigated. To achieve a good goal of rational use it is not sufficient to choose the right medicine --- health workers must be employed in the most appropriate proportion. We have noticed that most of the errors found in our study are related to very low doctor / patient ratio even in the tertiary care hospital. Therefore, to minimize the prescribing errors, we recommend the administrative authority in the government to consider increasing the doctor/ patient ratio. To reduce the complications of polypharmacy and improve the rational practice we recommend that the prescribers keep the number of medicines to the lowest and prescribe only those that are necessary. Please include this message. Conflict of interest {#sec2-7} -------------------- We hereby disclose that the study which we carried out under this drug-use pattern in the Galle District does not have any type of financial or personal relationship with other people or organizations that could inappropriately influence (bias) our study. We received a small grant from the university, under the Improving Relevance and Quality of Undergraduate Education (IRQUE) project, to buy stationery and for day-to day expenditure. **Source of Support:** Nil **Conflict of Interest:** None declared.

PubMed Central

2024-06-05T04:04:19.963933

2011-01-01

PMC3053517

Sir, Periodontitis is a destructive inflammatory disease of the supporting tissues of the teeth and is caused either by specific microorganisms or by a group of specific microorganisms, resulting in progressive destruction of periodontal ligament and alveolar bone with periodontal pocket formation, gingival recession, or both.\[[@CIT1]\] Diet and nutrition impact on many oral diseases, in particular gingival and periodontal diseases. A person's diet can exert a topical or a systemic effect on the body and its tissues. Before tooth eruption, foods provide a nutritional or systemic effect during tooth development and in the maturation of dentine and enamel. After the tooth erupts, foods play a topical or dietary role in the maintenance of tooth structure. It is well known that the caries process can be modified through dietary (food selection and eating habit changes) rather than nutritional changes.\[[@CIT2]\] Vitamins are organic compounds required as nutrients in tiny amounts by the organism and are required for the body to maintain appropriate metabolic reactions. Vitamins can be grouped as fat-soluble or water-soluble. By convention, the term vitamin does not include other essential nutrients such as dietary minerals, essential fatty acids, or essential amino acids (which are needed in larger amounts than vitamins). Vitamins have diverse biochemical functions. Some have hormone-like functions as regulators of mineral metabolism, or regulators of cell and tissue growth and differentiation, and others function as antioxidants. A largest number of vitamins function as precursors for enzyme cofactors that help enzymes in their work as catalysts in metabolism. Thirteen vitamins are presently universally recognized and are classified by their biological and chemical activity. Vitamins A, D, E, and K are fat-soluble, whereas vitamins B and C are water-soluble. Water-soluble vitamins dissolve easily in water and in general are readily excreted from the body, so consistent daily intake is important. Fat-soluble vitamins are absorbed through the intestinal tract with the help of lipids (fats). Greater possibilities to accumulate in the body, fat-soluble vitamins are more likely to lead to hypervitaminosis than are water-soluble vitamins. The first signs of deficiency of some micronutrients are seen first in the mouth, such as glossitis, cheilitis, and gingivitis. Undernutrition exacerbates the severity of oral infections and is a contributing factor to life-threatening diseases such as noma. Periodontal disease is associated with an increased production of reactive oxygen species which, if not buffered sufficiently, cause damage to the host cells and tissues. Antioxidant nutrients, for example, ascorbic acid, beta-carotene, and alpha-tocopherol, are important buffers of reactive oxygen species and are found in many fruits, vegetables, grains, and seeds. Current research is investigating the potential protective role of antioxidant nutrients in periodontal disease, and the most prudent approach is to recommend a daily intake of fruits and vegetables as a likely source of essential nutrients.

PubMed Central

2024-06-05T04:04:19.965904

2011-01-01

PMC3053518

Sir, This clinical report describes the prosthetic rehabilitation of an edentulous patient,who was dissatisfied from her 8-year-old denture. To give her a better fit, we opted Biofunctional Prosthetic System (BPS) for the new prosthesis. BPS is the system designed to work with the body in a biologically harmonious way, maximizing function, and giving comfort and natural appearance to the patient. The functional impression technique and simulation of the jaw movements by the Stratos 200 articulator in BPS ensure that BPS denture meets most exacting requirements.\[[@CIT1]\] BPS denture meets the esthetic demand of patients with its unique Ivoclear teeth, which replicate anatomy of the natural tooth Ivoclear teeth are made up of 3 layers of cross-linked acrylic resins that contribute to a life-like appearance and resistance to wearing. BPS system uses a controlled heat/pressure polymerization procedure during which time the exact amount of material flows into the flask to compensate for shrinkage, which ensures a perfect fit. This pressure also optimizes the physical properties of the denture. \[[@CIT2]\] A 60-year-old edentulous woman with a chief complaint of compromised function and esthetics was treated in the clinic. Intraoral examination showed resorbed ridges and masticatory dysfunction \[[Figure 1](#F0001){ref-type="fig"}\]. An extraoral examination revealed flattened mandibular plane. She was wearing dentures with attrited teeth and worn out denture base. A significant loss of vertical dimension affected the temporomandibular joint. Hence, a BPS denture was planned to give her a better fitted prosthesis. Figure 1Resorbed ridges The BPS recommends impression making similar in principle to the mucostatic method that minimally compresses tissues, using a combination of irreversible hydrocolloids of varying densities together in the same impression.\[[@CIT3]\] Low-density impression material (syringe Acc Gel) was syringed into the vestibular area and the occlusal centric tray was loaded with high-density hydrocolloid and inserted into the patient's mouth to get the initial vertical dimension \[[Figure 2](#F0002){ref-type="fig"}\]. This vertical dimension was used for mounting the casts obtained from initial impressions, taken with Accu-trays (different from conventional denture trays) with an extra flange to cover the vestibular areas and extended distal part to cover the retromandibular pad area more efficiently \[[Figure 3](#F0003){ref-type="fig"}\]. Custom trays were made on the primary casts. The Gnathometer M tracing device was attached to the casts, which facilitates the clinical procedures of secondary impression making, face-bow record and jaw registration \[[Figure 4](#F0004){ref-type="fig"}\]. Figure 2Occlussal centric tray loaded with impression for recording initial vertical dimension Figure 3Biofunctional prosthetic system impression trays Figure 4Bite registration through Gnathometer M The secondary impression was taken with zinc oxide eugenol impression paste \[[Figure 5](#F0005){ref-type="fig"}\]. Casts were poured and a wax-up denture was made for the trial \[[Figure 6](#F0006){ref-type="fig"}\]. After checking the fit and occlussal relations, the denture was sent to the laboratory. Dentures were cured with injection molding technique \[[Figure 7](#F0007){ref-type="fig"}\] using Ivocap high-impact plus denture base material.\[[@CIT4]\] Necessary adjustments were done and the dentures were delivered to the patient. Figure 5Secondary impression-making with zinc oxide eugenol paste Figure 6Wax-up trial for the patient Figure 7Acrylized denture The patient was recalled after 6 months and examined. There was no occlusal disharmony or sore spots. The patient was very much satisfied with her new prosthesis and she showed her gratification for the comfortable prosthesis and a younger look. We are grateful to Mr. Chauhan, Dental Technician, Chauhan Dental Lab, Sec-32, Chandigarh, India, for his laboratory work.

PubMed Central

2024-06-05T04:04:19.966604

2011-01-01

PMC3053519

New Delhi, India, August 30, 2010 - The Journal of Pharmacy and BioAllied Sciences (JPBS) today announced the launch of a special issue on Chemical, Biological, Radiological, and Nuclear (CBRN) Disaster Management,, edited by Dr. Rakesh Kumar Sharma, Scientist 'G', Additional Director and Head, CBRN Defence, Institute of Nuclear Medicine and Allied Sciences (INMAS), Defence Research and Development Organization, Brig. SK Mazumdar Marg, Delhi, India. Lt. Gen. J. R. Bhardwaj, Honorable Member, National Disaster Management Authority released the issue on August 30, 2010, during the inauguration of the Conference on 'Emergency Planning in Industries including Halma Water Management (HWM) and Transportation of Petroleum, Petroleum Products, Natural Gas by Pipelines and Petroleum, Oil, and Lubricant (POL) Tankers,' organized by the Federation of Indian Chambers of Commerce and Industry (FICCI), in association with the National Disaster Management Authority (NDMA) and Ministry of Environment and Forests (MoEF), Government of India, and the Petroleum and Natural Gas Regulatory Board (PNGRB). Release of the Special Issue of JPBS on CBRN Disaster Management. Seen in the picture from left to right are Dr. Amit Mitra, Secretary General, FICCI, Lt. Gen. J. R. Bhardwaj, Honorable Member, NDMA, Dr. Rakesh Kumar Sharma, Additional Director - CBRN Defence, INMAS / DRDO and Editor of the Special Issue of JPBS, Gen. N C Vij, Honorable Vice Chairman, NDMA and Shri L. Mansingh, Chairperson of the Petroleum and Natural Gas Regulatory Board. For further information please contact

PubMed Central

2024-06-05T04:04:19.967303

2011-01-01

PMC3053520

Differential Scanning Calorimetry (DSC), is a straight forward, non-perturbing technique, first developed in the early1960s. This method measures the thermodynamic properties of thermally induced transitions and has been applied to a variety of biological macromolecules such as lipids or proteins.\[[@CIT1][@CIT2]\] Examples of these applications have involved conformational states of proteins, DNA binding to protein,\[[@CIT3]\] biopolymer melting, lipid phase transitions, and lipid-protein interactions.\[[@CIT1][@CIT4]\] Differential Scanning Calorimetry is primarily used to determine the energetics of phase transitions and conformational changes and allows quantification of their temperature dependence.\[[@CIT5]\] Technical improvements over time have resulted in high sensitivity instruments, which also make DSC a very relevant tool for investigating the thermodynamic properties of various pharmaceutical products, such as, biopolymers, proteins, peptides, and lipid carriers.\[[@CIT1][@CIT4]\] Many reviews are available on protein conformation,\[[@CIT4]\] biopolymers stabilization,\[[@CIT6]\] thermodynamic properties of lipids,\[[@CIT1]\] and lipid-protein interactions,\[[@CIT7]\] however, this article will focus on the application of DSC in the pharmaceutical field, with an emphasis on drug-lipid interactions. Many groups have made relevant contributions and no overview can be fully comprehensive to acknowledge that. Most references in this article are reviews that will provide the reader with sources for a wealth of detailed references. History {#sec1-1} ======= The evolution of scanning microcalorimeters has progressed rapidly since first described in a publication in 1964 \[[Figure 1a](#F0001){ref-type="fig"}\].\[[@CIT8]\] Initially designed for measuring temperature-induced heat-release from conformational changes, the instruments were applied to biopolymers and the melting of DNA double helices. The introduction of differential adiabatic scanning microcalorimeters (DASM), in 1963, allowed continuous measurements of heat capacity as a function of a set heating rate.\[[@CIT8]\] Adiabatic processes are defined as the absence of heat transfer between a system and the environment, and early DSCs used shields, vacuum or water jackets to protect temperature feedback loops to the outside environment \[[Figure 1b](#F0001){ref-type="fig"}\].\[[@CIT8]\] Moreover, the continuous measurement over a set temperature range was a major advancement allowing for the comprehensive analysis of temperature dependence on thermally induced events.\[[@CIT8]\] Furthermore, the development of differential heating abilities enabled comparison of the energy difference between a reference and sample cell, which effectively canceled contributions from extraneous factors or solvents.\[[@CIT8]\] ::: {#F0001 .fig} Figure 1 ::: {.caption} ###### a\) Figure of the first DSC used for studying liquids b) Schematic of an adiabatization system using thermal shields, vacuum and water jackets. Reprinted from Thermochimica Acta, Vol. 139, P.L. Privalov V.V. Plotnikov, Three generations of scanning microcalorimeters for liquids, 257-277, 1989, with permission from Elsevier :::  ::: The next key breakthrough was the miniaturization of the cells, which improved the sensitivity, by drastically reducing temperature gradients that occurred in larger samples. Moreover, this design change also allowed the elimination of stirring mechanisms.\[[@CIT9]\] By replacing cylindrical cells with capillary tubes that had a very high surface-to-volume ratio, the effects of viscosity and gradient heating were minimized. Furthermore, the capillaries could withstand higher internal pressure than other cells of the same thickness. The increased pressure resistance increased the operational temperature and made DSC a very versatile tool, with a large temperature range.\[[@CIT8][@CIT9]\] Subsequently, non-adiabatic differential scanning microcalorimeters were designed, as they were simpler to manufacture and were more applicable in the industry. The use of cells that were removed for washing and required adiabatization after loading resulted in baseline instability and irreproducibility. Moroever, altering the heat capacity measurements from mass to volume minimized the large error associated with loading, and increased the accuracy and reproducibility of the machine (more details in the next section).\[[@CIT8][@CIT9]\] The cells were fitted with sensors to determine the volume, rather than relying on user measurements.\[[@CIT9]\] Two main systems are used to control cell temperatures. The first is a power compensation unit, which independently controls and monitors the temperature of the reference and sample cells. Constant energy is provided to both cells, hence, the temperature increases at a steady rate.\[[@CIT10]\] However, a thermally induced transition that requires heat results in a temperature lag in the sample cell compared to the reference. The extra heat required to maintain the same temperature between two cells is used to calculate the excess heat capacity.\[[@CIT10]\] Independent controls utilize two heating / cooling units (one for each cell) to maintain the temperature. The second system is referred to as a heat-flux or heat leak principle, where both cells are connected via a low resistance heating flow-path (usually a metal disk). The recorded difference in voltage of the temperature-measuring device is proportional to the temperature difference that is used in the heat capacity calculation.\[[@CIT10]\] There are multiple different designs of scanning calorimeters based on the applications and samples tested, however, they all share three main characteristics.\[[@CIT3]\] The first is the fact that calorimeters must be able to measure temperature changes, keep constant heating or cooling rates, and take accurate temperature measurements.\[[@CIT2][@CIT3]\] Second, the instruments must accurately measure the differential heat flow between the sample and reference cell, which results in better baseline stability and reduced noise.\[[@CIT3][@CIT9]\] Finally, the cell contents are measured usually in volume (older instruments may still use mass), which is essential for reproducible and accurate values.\[[@CIT3]\] The various modern calorimeters also retain twin cells and a differential heating mode, where cells are heated or cooled quasi-adiabatically at a constant rate.\[[@CIT5]\] Current calorimeters have become exceptionally more accurate with advancements in sample size, baseline stability, and sensitivity.\[[@CIT3][@CIT5]\] The temperature range of operation has also been increased using high pressure to scan to about 100°C and super cooling to measure below 0°C.\[[@CIT3]\] Moreover, many different DSC models are available based on their application. Examples include Hyper DSC, which allow very high scanning rates such as 400 -- 500°C/min and modulated DSC, which can separate heat flows from reversible and non-reversible events. Nano DSC can operate with very small quantities of sample per trial, approximately 130 µL or 100 µg, while maintaining the accuracy of larger volume calorimeters.\[[@CIT11][@CIT12]\] In recent times, fully automated cleaning and loading devices have been incorporated in many DSC models, which enable computer-controlled sample addition, cell cleaning, and sample degassing. Such instruments can test 50 samples a day with increased accuracy and minimal systematic errors.\[[@CIT11][@CIT13]\] Theory {#sec1-2} ====== Differential Scanning Calorimetry is used to measure the specific heat capacity of thermally induced events as a function of temperature.\[[@CIT5]\] The apparent specific heat (c~2~) of a solution is calculated by the following equation: $$c_{2}\ = \ c_{1}\ + \ 1/w_{2}\left( {c - c_{1}} \right)$$ where c is the specific heat of the solution, c~1~ is the specific heat of the solvent, and w~2~ is the weight fraction of the solute.\[[@CIT4]\] DSC measures the excess apparent specific heat (c~ex~), which is the value (c-c~1~) in [equation 1](#FD1){ref-type="disp-formula"}. Expanding the definition of c~ex~ (c-c~1~), the measured heat capacity of the buffer (c~1~) can be written as: $$C_{b}\ = \ m_{b} \times {C_{b}}^{deg}$$ where m~b~ and C~b~° are the mass and the specific heat capacity of the buffer, respectively. Equally, the heat capacity of the sample solution (c) can be expressed as: $$C_{s}\ = \ m_{s} \times {C_{s}}^{deg}$$ with 's' denoting the sample. By subtracting these two values the c~ex~ can be determined.\[[@CIT2]\] The value (m~b~-m~s~) can be replaced by the partial specific volume, removing mass from the equation, as new calorimeters use the more precise volume over mass measurements.\[[@CIT2]\] The differential heat flow from the calorimeter is temperaturedependent and is referred to as a thermoanalytical curve. As the scan rate is constant, the time integral of the measured differential heat flow provides the energy of the sample.\[[@CIT3]\] As the c~ex~ is usually quite small (about 0.7% for a 1% aqueous protein solution), using equal volumes of solution and proper shielding from external effects is of paramount importance.\[[@CIT4]\] The excess specific heat is plotted against temperature, revealing the respective transitions. Integration of c~ex~ over the temperature range results in specific calorimetric enthalpy ∆h~cal~.\[[@CIT10]\] However, traditionally, problems arise when performing integrations.\[[@CIT2][@CIT4]\] For example, the course of the baseline is not necessarily obvious during a phase transition and may change after the transition, thus, artificial baselines and sophisticated software tools are necessary. Experimental Procedures {#sec1-3} ======================= Sample preparation differs depending on the type of sample to be analyzed although in most cases the compound of interest is studied in buffered aqueous solutions. Sharp peaks such as the first order gel-to-liquid crystalline phase transition (L~α~) seen for high purity phosphatidylcholines (PC)\[[@CIT4]\] require very low scan rates of around 0.1 K min^-1^ or less, so as to avoid the broadening commonly seen with faster scan rates. Slower scan rates are also beneficial as they enhance the resolution, thus enabling the resolution of closely spaced DSC peaks that may arise from single phospholipid phase transitions. However, slow scanning rates result in decreased signals and more sensitive calorimeters are required.\[[@CIT4]\] Most modern DSC instruments have two cells one as a sample and one as a reference, but some calorimeters have three samples cells that can be scanned against the same reference.\[[@CIT10]\] As the volume is used to determine the c~ex,~ sample and reference solutions are normally degassed prior to being loaded into the cell. This is important, to avoid the formation of bubbles that will affect the accuracy of the volume and add spikes and experimental noise to the thermograms. However, a disadvantage is that the small capillary cells will make the cleaning more difficult, which may also result in bubble formation.\[[@CIT8]\] State-of-the-art instruments allow setting a variety of experimental parameters such as the post scan temperature, the number of scans, their range, scan rate, and feedback strength. As discussed earlier, slower scan rates provide higher resolution. Furthermore, a high feedback will give optimal sensitivity to the reaction, but may increase the noise levels in the experiments. Finally, it is important to reach equilibrium before the thermotropic data are analyzed. To ensure that this has been reached, sufficient scans are recorded until two scans are superimposed. Once the parameters have been chosen for an experiment, the temperature is scanned at the set rate in the heating or cooling mode. Initially the temperatures in both cells increase linearly to the same extent, resulting in a zero baseline.\[[@CIT1][@CIT10]\] However, once the sample undergoes a phase transition a temperature difference is observed. During endothermic events the recorder will move upward, indicating that energy input is required and in an exothermic event a downward deflection is seen as less energy being required from the DSC, to maintain the temperature. The size of the deflection is dependent on the heating or cooling rate, and following the thermal event, the signal returns to baseline or a new baseline can be detected if there has been a change in the specific heat.\[[@CIT1]\] Analysis {#sec1-4} ======== Differential Scanning Calorimetry analysis is performed on equilibrium data.\[[@CIT4][@CIT14][@CIT15]\] Depending on the system investigated, different means of analysis and different models have been devised. Most interpretations are based on the van't Hoff equation: $$\left( {d\ln\ K\ /\ dT} \right)\ = \ \Delta H_{vH}/\ {RT}^{2}$$ where K is the equilibrium constant of the process, T is the absolute temperature, and ∆H~vH~ is the van't Hoff enthalpy.\[[@CIT4][@CIT14]\] This equation is only applicable to two state processes, without significant intermediate populations, during the transition.\[[@CIT4]\] This model is normally applied as most systems have an initial state, some intermediate state during the transition, and a final state. There are even differences in the two state models based on whether there is a change in specific heat after the transition, as observed for the denaturation of the T4 lysozyme.\[[@CIT4]\] More complex models are also used for multi-state changes such as gradual unfolding and the presence of different intermediate states that make data analysis more complicated. In such cases a different equation, incorporating the entire transition, is utilized, where each step has its own set of parameters, such as, van't Hoff Enthalpy and T~1/2~. For a comprehensive review on the different models refer to.\[[@CIT2][@CIT4][@CIT14]\] The enthalpy of the endothermic or exothermic event is determined by the integration of the area under the DSC peak, which is often reported in kcal / mol \[[Figure 2](#F0002){ref-type="fig"}\].\[[@CIT1][@CIT10][@CIT14][@CIT16]\] Initially this was performed by means of a planimeter or even by cutting and weighing the paper traces, to determine the values to use in the van't Hoff equation.\[[@CIT1][@CIT4]\] Today, various iterative processes in the modern software are used, with different equations, based on the type of process (two state, irreversible, etc.). Moreover, the instruments are calibrated with known standards and a buffer blank is subtracted to provide accurate enthalpy values. ::: {#F0002 .fig} Figure 2 ::: {.caption} ###### Enthalpy, T~1/2~ and T~m~ shown on a DSC endotherm. Reprinted from Chemistry and Physics of Lipids, Vol. 30, R.N. McElhaney, The use of differential scannning calorimetry and differential thermal analysis in studies of model and biological membranes, 229-259, 1982, with permission from Elsevier :::  ::: The maximum height of the transition (also maximum heat capacity) occurs at the phase transition.\[[@CIT1][@CIT16]\] In the case of lipids the peak of a symmetrical curve represents the temperature at which the gel-to-liquid-crystalline state is half complete \[[Figure 2](#F0002){ref-type="fig"}\]. However, many biological extracts and pure phospholipid thermograms are asymmetrical and the T~m~ is not longer the midpoint of phase transition,\[[@CIT1]\] and in this case the width of the distribution is considered (see a little further in the text). The shape of the thermally induced event is described by the width of the transition at half height of the peak (T~1/2~), whereby, the peak is defined by the difference between the lower (T~S~) and upper boundaries (T~L~) of the phase transition \[[Figure 2](#F0002){ref-type="fig"}\]. Values can range from 0.1°C for pure phospholipids to over 15°C for biological membranes.\[[@CIT1][@CIT16]\] T~1/2~ is a valuable tool to gauge purity, protein-lipid interactions, as well as lipid-lipid interactions, and provides information about the cooperativity of the phase transition.\[[@CIT10]\] Cooperativity of a pure lipid transition is related to the shape and sharpness of the peak and is described by a cooperative unit (CU), the number of lipids involved in the transition.\[[@CIT10]\] Furthermore, CU can be calculated by the ratio of ∆H~vH~/∆H~cal~, where ∆H~cal~ is the enthalpy of the transition (cal / mol) and ∆H ~vH~ is the van't Hoff enthalpy.\[[@CIT4]\] The van't Hoff enthalpy can be determined using an approximate relationship relation: H~vh~ \~ format 4RT~m^2^/T~1/2 ([equation 4](#FD4){ref-type="disp-formula"}).\[[@CIT1]\] For a purely cooperative first order transition, cooperativity would reach nearly infinity, whereas, a non-cooperative process will reach zero\[[@CIT1]\] Highly purified synthetic phospholipids can yield almost fully cooperative transitions, but as even small impurities can have a significant impact, the cooperativity value should be interpreted with caution.\[[@CIT1][@CIT16]\] Besides the three main values immediately apparent from the DSC trace, other important thermotropic parameters can also be calculated. As free energy (G) is zero at the phase transition T~m~, the enthalpy can be calculated using the equation $$\Delta S\ = \ H_{Cal}\ /\ T_{m}$$ Where H~cal~ is the enthalpy that corresponds to the area under the transition peak. Moreover, the partition function for a macromolecule system can be found by a double integration of the apparent heat capacity.\[[@CIT4]\] Fractional occupancy of different states has also been calculated, based on DSC thermograms, assuming that only two distinct states exist.\[[@CIT10]\] This is performed using the equilibrium constant: $$K = \ \left\lbrack B \right\rbrack\ /\ \left\lbrack A \right\rbrack\ = \ f\left( {1 - f} \right)$$ where K represents the equilibrium constant, A and B are the respective states, and f is the fractional occupancy.\[[@CIT10]\] As K can be determined via: $$\Delta G^{deg}\ = \ - {RT}\ln K\ = \ \Delta H^{deg}\ - \ T\Delta S^{deg}$$ the absolute heat capacity difference between the unfolded and folded states can be used to show solvent accessible polar and apolar surfaces between the states.\[[@CIT2]\] Deconvolution analysis of the heat capacity function can yield the number of states that will be populated during the denaturation of the protein, which allows a more detailed analysis of this process.\[[@CIT17][@CIT18]\] For a full review on protein analysis using DSC refer to review.\[[@CIT2]\] Applications {#sec1-5} ============ Considering the ability to measure enthalpy changes and phase transitions, there are multiple applications for such a versatile tool. There are good reviews on its application to proteins,\[[@CIT3]\] protein for pharmaceutical interest,\[[@CIT19]\] protein mutations,\[[@CIT20]\] protein-ligand interactions,\[[@CIT11][@CIT21][@CIT22]\] protein folding,\[[@CIT23]--[@CIT26]\] nucleotides,\[[@CIT4]\] other macromoleculesdont,\[[@CIT6][@CIT27]\] lipids,\[[@CIT28][@CIT29]\] drug-lipid interactions,\[[@CIT30]\] and protein-lipid interactions.\[[@CIT31]\] This review will start with a brief overview of other pharmaceutical applications and will focus on lipid-drug interactions such as antimicrobial peptides. A good review on drug development using DSC is presented in\[[@CIT32]--[@CIT34]\] and for drug development uses for DSC.\[[@CIT33]\] Proteins {#sec1-6} ======== As pharmaceutical products can come in the form of proteins (e.g., enzymes), their thermodynamic properties are important, and one of the earliest DSC applications was to study thermally induced, cooperative conformational changes of small proteins.\[[@CIT6][@CIT35][@CIT36]\] However, small molecules do not yield good data unless they aggregate, showing intermolecular cooperation. The application of DSC to protein denaturations was described by Freire and Biltonen,\[[@CIT37][@CIT38]\] who reported that thermal transition was synonymous with the protein partition function, suggesting that the thermogram can be used to identify the states in denaturation.\[[@CIT37][@CIT38]\] Thus, protein thermodynamics, during unfolding, is measured as an enthalpy change, as a function of temperature, to determine the partition coefficient.\[[@CIT14]\] For a full review of the thermodynamic calculations for different types of denaturation see.\[[@CIT4][@CIT14][@CIT35][@CIT36][@CIT39]\] Differential Scanning Calorimetry-based analysis of the thermal denaturation of proteins provides an insight into the unfolding process and forces involved in conformation stability.\[[@CIT4][@CIT40]\] For comprehensive reviews on protein denaturation refer to\[[@CIT3][@CIT4][@CIT41]\] and for protein folding.\[[@CIT14][@CIT42]\] During protein denaturation there are different thermodynamic states, with many microscopic states. This process is highly cooperative with disruption of many forces and bonds, including hydrogen bonds, hydrophobic interactions, and many non-covalent interactions.\[[@CIT41]\] DSC allows for the direct study of thermal stability, over a very large concentration range, in the absence of light, thus photosensitive proteins such as bovine lens crystallins can be analyzed.\[[@CIT41]\] Conversely protein folding can also be studied, investigating thermotropic changes in different environments. The energetics and heat capacity, ∆C~p,~ of the protein, refolding into different conformations such as α-helix or β-barrel structures\[[@CIT25]\] is used for this purpose. Such analysis has been performed on the α-helical membrane protein, bacteriorhodopsin, which yields a high transition temperature and low unfolding enthalpy.\[[@CIT25]\] For a review on the energetic states of protein conformations, refer to.\[[@CIT25]\] Furthermore, the enthalpy of relaxation (∆H ^\*^) can be investigated by using DSC for the characterization of the structural relaxation of a protein.\[[@CIT43]\] Heat capacity for thermally induced protein denaturation has revealed thermodynamic information about the different states,\[[@CIT9][@CIT35]\] as it depends on three major factors. The first relates to the primary structure of the protein and contributes from stretching and bending to the rotating of internal bonds.\[[@CIT6][@CIT35][@CIT36]\] The second factor is based on non-covalent interactions from the secondary and tertiary structures. Finally contributions from the hydration affect the heat capacity. The primary structure provides the most significant contribution, followed by hydration, and less impact from the non-covalent secondary and tertiary interactions.\[[@CIT6]\] Such denaturation processes can be categorized into either two-state denaturations or multi-state denaturations. The former can be further broken into multiple different groups, such as, those with self-dissociation, ligand dissociation, and large permanent specific heat changes.\[[@CIT4]\] Multi-state denaturation has been observed for many proteins, including histones H1 and H5.\[[@CIT4][@CIT36]\] Different trends in the T~m~, T~1/2~, and enthalpy are observed, and hence, allow the classification of a given protein. Once the denaturation process has been established, the stabilizing factors and conformations can be more easily assessed. Accompanying protein denaturation is the study of protein stability, which is of great importance in understanding its role and in its screening, for improved stability of proteins, for pharmaceutical applications.\[[@CIT37]\] DSC can be used to study two types of protein stabilities, thermodynamic stability or kinetic stability.\[[@CIT37]\] Most calorimetric protein studies involve the thermodynamic stability, relating to the equilibrium between the native folded and the unfolded or partially unfolded states.\[[@CIT37]\] The focus on thermodynamic stability is due to the ease of studying small proteins and the availability of software and algorithms that can be easily applied.\[[@CIT37]\] Kinetic stability relates to the Gibbs energy between the folded and unfolded states, reached with progressive scans. The amount of time required to adopt the proper state or lose the adopted conformation is essential in pharmaceutical applications, for drug shelf life and potency. Furthermore, the ability to adopt the active state under non-ideal intracellular and extracellular environments may vary and different environments can be investigated using DSC.\[[@CIT37]\] Low kinetic stability drugs have been improved through mutations, and reassessed using calorimetry, as a quick comparison with the original can be performed. Moreover protein-ligand stability has also been studied to screen for undesirable effects such as aggregation or proteolysis.\[[@CIT11][@CIT12][@CIT37][@CIT44]\] A current review has discussed the role of DSC in the kinetic stabilization of proteins.\[[@CIT11][@CIT12][@CIT37]\] The solid state of proteins has also been studied using DSC, as the chemical and physical degradation is significantly reduced. The solid state provides the ability for improved drug delivery without the need of an excipient.\[[@CIT45]\] However, DSC is better suited for proteins dissolved in a solution. This has been directly applied to investigate the propensity of liquid protein therapeutics, to aggregate during storage.\[[@CIT46]\] DSC allows for a quick scanning procedure to detect the presence of aggregates without the need for extended stability trials. However, problems arise when studying protein denaturations, as the majority of transitions are deemed calorimetrically irreversible, as upon denaturing, a subsequent scan will show no transition or a significantly reduced one.\[[@CIT37][@CIT47]\] Most protein denaturation analyses are performed assuming equilibrium thermodynamics, hence, suitable analysis is only available for kinetic stability.\[[@CIT37]\] Further problems with the conformational analysis of proteins relate to dilute sample solutions of the protein. The high background heat capacity of the system may overshadow signals from dilute samples and require very sensitive and precise calorimeters.\[[@CIT2][@CIT9][@CIT35]\] Good reviews on applications of proteins and potential problems associated with calorimetry are presented in.\[[@CIT11][@CIT12][@CIT47]\] Applications of DSC for proteins are not limited to structural changes, as shown in a few examples. The polymerization steps of tobacco mosaic virus coat protein include thermally induced reversible conformational changes, which can be investigated by both heating and cooling scans \[[Figure 3](#F0003){ref-type="fig"}\].\[[@CIT4][@CIT12][@CIT35]\] DSC has also been used with proteins of the plant photosynthetic system, to study the effects of temperature on the heat inactivation process of photosystem 2.\[[@CIT6]\] Moreover, molecular recognition studies have been performed by DSC as ligand or drug association alter intermolecular interactions, which result in changes of T~m~, enthalpy and free energy associated with such interactions.\[[@CIT21][@CIT48]\] Such protein ligand studies have been reported for glucose transporter GLUT-1 and ATP as well as bovine serum albumin and anilinonaphtalene sulfonate (ANS)\[[@CIT48]\] ::: {#F0003 .fig} Figure 3 ::: {.caption} ###### Heating and Cooling Scans of an alpha helix forming peptide. Reprinted from Methods in Enzymology, Vol. 323, G.P. Privalov P.L. Privalov, Problems and prospects in microcalorimetry of biological macromolecules, 31-62, 2000, with permission from Elsevier :::  ::: Protein-lipid interactions such as the interaction of various apoproteins with different lipid mixtures have been investigated showing preferential binding to specific matrices.\[[@CIT49][@CIT50]\] Studies on the lipid interaction of cytochrome C oxidase showed that one oxidase molecule perturbed over 70 lipid molecules corresponding to the lipids surrounding the protein.\[[@CIT51]\] For a review on membrane proteins and DSC refer to.\[[@CIT51]\] Protein effects on surfactant lipid systems have also been studied with SP-B and SP-C.\[[@CIT52]\] Furthermore, the binding stability has shown that most DNA-binding proteins are typically unstable without DNA.\[[@CIT3]\] This is used to determine the conformation of the protein upon binding DNA. DNA Drugs {#sec1-7} ========= Base stacking enthalpies and helix-coil enthalpies have been used to determine conformations of DNA.\[[@CIT4]\] Generally it has been found that an increase in enthalpy of 8 - 10 kcal (mol / base pair)^-1^ is observed for a helix coil transition.\[[@CIT4]\] These enthalpies have even been used to predict quaternary and quinternary structures of DNA in liver nuclei.\[[@CIT53]\] The investigation of the heat capacity of DNA and RNA identified water clusters on the nucleic acid matrix.\[[@CIT6]\] A change in hydration can be used to explain the exposure of polar or apolar groups revealing the possible drug-binding sites.\[[@CIT3]\] For a review on DNA thermogram analysis refer to.\[[@CIT54]\] Many of the DNA melting curves are typically quite broad and contain overlapping regions,\[[@CIT55]\] because DSC only measures the overall enthalpy changes and cannot distinguish between enthalpies from different thermodynamic events. Statistical deconvolution has been applied to many of these thermograms, by essentially 'desmearing' low resolution overlapping transitions by fitting them to individual peaks that contribute to the enthalpic endotherm.\[[@CIT41]\] Deconvolution has been utilized to provide a direct means of obtaining a partition function and properties of intermediate states \[[Figure 4](#F0004){ref-type="fig"}\].\[[@CIT18][@CIT54]\] First described by Freire and Biltonen, deconvolution can be used to establish the partition function of the thermal unfolding event by using a mathematical algorithm.\[[@CIT18][@CIT38][@CIT54]\] Once the partition function is determined, properties such as cooperative melting and information about more complex structures such as oligomeric hairpins, can be analyzed.\[[@CIT54]\] Typical deconvolution of DNA melting profiles yields biphasic and triphasic transitions and allows for a thermodynamical description of the transitions for each complex, by indicating the favorable enthalpic contribution due to base stacking and the effects of environment, such as, pH and ionic strength.\[[@CIT55]\] ::: {#F0004 .fig} Figure 4 ::: {.caption} ###### Deconvolution of a thermogram with (\.....) representing experimental data and (\-\-\-\--) representing the deconvoluted data for excess heat capacity of myosin rod in 0.5M KCl, 0.2 M Phosphate pH 7.0. Reprinted from Thermochimica Acta, Vol. 193, G. Castronuovo, Proteins in aqueous solutions. Calorimetric studies and thermodynamic characterization, 363-390, 1991, with permission from Elsevier :::  ::: Thermal stability of nucleic acids and their melting behavior in various duplex or triplex conformations has been studied using DSC.\[[@CIT55]\] Denaturation of triplex structures has shown that initially the third strand is removed followed by the unfolding of the duplex.\[[@CIT17][@CIT55]\] Furthermore, a melting analysis of the different oliognucleotides revealed forces involved in the structural stability as well as the effect of ions, pH, and temperature.\[[@CIT6]\] Ligand-DNA interactions similar to protein-ligand interactions have been used to test the pharmaceutical development potential of anti tumor drugs, by assessing their binding to DNA.\[[@CIT4]\] For a review of the calorimetric binding of anti-tumor drugs to DNA, see.\[[@CIT4]\] Differential scanning calorimetry has also been applied to analyze *in vitro* interactions of antitumor drugs, with human epithelial cell nuclei that exhibit a characteristic melting profile, with four structural transitions.\[[@CIT56]\] A loss of the fourth transitional peak upon drug treatment is correlated with the inhibition of cell division induced by different DNA strand breakers and alkylating drugs. The primary mode of action of many antitumor drugs is DNA damage and cleavage and the observed changes to the four transitions provides an insight into the mechanism of breakage.\[[@CIT56]\] Hence, DSC is a quick screening tool to observe the effect of intercalating drugs on a nucleosome structure. Changes in the supercoiled loops can also be used to study DNA strand breakage and to assess the effect of intercalating drugs on base pair stability. This has been observed for belomycin and streptonigrin, which destabilize the supercoiled DNA to a relaxed form, characterized by a drop in enthalpy and the T~m~ of the fourth transition.\[[@CIT56]\] Moreover, different mechanisms of DNA interaction can be elucidated as alkylating drugs produce a kinetic intermediate peak, and intercalating drugs reduce the melting temperatures of transition II, but increase the T~m~ of transition IV.\[[@CIT56]\] Intercalating drugs from the anthracycline group of antibiotics such as ethidium bromide and actinomycin D showed a characteristic shift of the seven DNA melting peaks observed in a plasmid, to higher temperatures.\[[@CIT57]\] The magnitude of the shifts depended on the strength and concentration of the drug. Furthermore, the binding sites could be determined from the peaks that diminish with increasing concentration.\[[@CIT57]\] Examples include the binding of danomycin to the 5'CG-3"-rich region of DNA sequences.\[[@CIT57]\] This insight can help to distinguish between minor or major groove binders and to identify other specific binding sequences, which aid in rational drug design. DNA-drug interactions have been studied for many non-steroid anti-inflammatory drugs (NSAIDs).\[[@CIT58]\] Variations in the calorimetric data such as enthalpies and temperatures for the unfolding of DNA provide information about the type of drug interaction. For example a primarily electrostatic interaction results in a decrease in enthalpy with increased drug addition.\[[@CIT58]\] The overall stability of the DNA is affected by many compounds that impact the observed scans. A stabilizing effect (e.g., seen for urea) will shift the calorimetric peak to higher temperatures.\[[@CIT58][@CIT59]\] Furthermore, the presence of more Guanine-Cytosine base pairs increases the enthalpy due to additional hydrogen bonds that stabilize the double helix. Addition of NSAID drugs such as naproxen and ketoprofen lower the T~m~, which indicates a reduction in the energy required for denaturations, suggesting a destabilizing drug interference between base-pair interactions.\[[@CIT58]\] Moreover, the effects of novel methods such as virus-induced gene silencing can also be investigated by DSC. The treatment of many genetic disorders is envisioned via the delivery of plasmid DNA, which has been studied *in vitro* and *in vivo*.\[[@CIT60]\] Typical DNA transfection techniques suffer from cellular toxicity and the safety of such retroviral delivery systems is not well-established. pH-sensitive liposomes have been utilized as potential plasmid delivery mechanisms.\[[@CIT60]\] Plasmid pPTCK-6A was encapsulated in a DOPE / Cholesterol and an antigen, resulting in immunoliposomes. The gene was shown to be successfully delivered into the cell by monitoring the reporter gene.\[[@CIT60]\] DSC can be used to monitor the interaction of the plasmid with the liposome and to ensure that aggregation does not exist. Lipids {#sec1-8} ====== Phospholipids are one of the most studied lipids by DSC.\[[@CIT61]\] One of their major advantages is that pure synthetic phospholipids undergo transitions at well-defined temperatures based on their structure.\[[@CIT10][@CIT61]\] Hence, the transitions are easily reproducible and trends can be established within systematically altered lipids (e.g., progressively increasing chain length). Pure lipids are analyzed as aqueous dispersions, formed from a lipid film, by mechanical agitation, such as vortexing. They contain multilamellar vesicles (MLV), which are closed multi sheaths comprised of concentric bilayers that are separated by aqueous spaces.\[[@CIT1][@CIT62][@CIT63]\] MLVs are the predominate form used to investigate lipids, as they provide the clearest resolution of phase transitions with accurate enthalpy values. Different vesicle preparations alter the observed thermograms, as single unilamellar vesicles (SUVs) produce a lower resolution peak than MLVs. Sonicated disaturated PC thermograms reveal less enthalpic transitions with a greater T~1/2~ and no notable pre-transition. The increase in peak width is likely from a reduced enthalpic component rather than a decrease in CU. However, a decrease in cooperativity can be attributed to the smaller radius of SUVs over MLVs, resulting in a less ordered orientation, which increases the free motion of the hydrocarbon chains.\[[@CIT1]\] The affect of the radius coincides with other calorimetric studies indicating that the thermogram is dependent on the size of the vesicle.\[[@CIT10]\] When DPPC SUVs are studied in the cooling mode their main T~m~ decreases to 37°C (MLV 41°C) accompanied by a lower enthalpy and a substantially larger T~1/2~.\[[@CIT10][@CIT64]\] However, complications arise when the SUVs are studied in the heating mode, as they tend to fuse by forming large unilamellar vesicles (LUVs). The thermograms of LUVs are nearly identical to those of MLVs, although with slightly broader endotherms attributed to the size inhomogeneity.\[[@CIT10][@CIT64]\] Multilamellar vesicle thermograms produce reversible and highly cooperative transitions Phospholipids exhibit three main groups of phase transitions, however, they are not always detectable \[[Figure 5](#F0005){ref-type="fig"}\].\[[@CIT10]\] The first is the most observed and best characterized gel-to-liquid crystalline transition, L~α~, which occurs at the T~m~. This transition is quite rapid and is the conversion from a gel to liquid crystalline state.\[[@CIT10]\] The second transition is only seen for some phospholipids and usually occurs below the T~m~. It is much slower and exhibits much less enthalpy when compared with the L~α~. This so called pre-transition is from a gel to a rippled gel phase and is sensitive to impurities and has been used to gauge vesicle preparation.\[[@CIT10]\] A review on using DSC to evaluate liposome preparation is available.\[[@CIT10]\] The last transition is not very well characterized and usually occurs below the operational range of most conventional DSCs. This subgel transition is very slow and does not reveal a lot of thermodynamic information.\[[@CIT10]\] Each of these transitions is characterized by its own temperature (T~m~, T~p~, T~s~, respectively), their own enthalpy (∆H ~m~, ∆H ~p~, ∆Hs respectively), and their own half width T~1/2~.\[[@CIT10]\] The gel-to-liquid crystalline lipid phase transition is the most well-understood, however, the DSC data has also suggested that the pre-melting and pre-freezing phenomena can provide information about the liquid-liquid phase separation and boundary defects in the solid state.\[[@CIT1]\] The main transition is where the lipid membrane changes from a relatively ordered crystalline-like gel state to a disordered fluid-like state.\[[@CIT1][@CIT61]\] This transition is due to the cooperative melting of the hydrocarbon chains, which retains the lamellar structure. It includes a conformational change of the hydrocarbon chains from all trans in the rigid gel state to a disordered state that allows gauge conformations.\[[@CIT1]\] Accompanying these changes in hydrocarbon orientation, are a lateral expansions due to increased mobility, and a concomitant decrease in the bilayer thickness.\[[@CIT1]\] Moreover, the increasing chain length and saturation, results in higher enthalpy values for the L~α~ transition.\[[@CIT6]\] Hence the phase transition enthalpy of lipids depends on the structure of the lipid, especially with the position of unsaturated bonds and length of the fatty acid chain.\[[@CIT1][@CIT6]\] Moreover, shifts in T~m~, enthalpy, and increased T~1/2~ values are good indicators of sample purity and liposomes size distribution.\[[@CIT10]\] Many of the thermodynamic properties of synthetic and biologically derived lipid phase transitions are available through an online database LIPIDAT <http://www.lipidat.chemistry.ohio-state.edu>\[[@CIT65]\] Phosphatidylcholine {#sec1-9} =================== Phosphatidylcholines (PCs) are among the most common components of mammalian membranes\[[@CIT65]\] and have mainly structural roles. A comprehensive review of the phase transitions of PC is available.\[[@CIT10][@CIT65]\] The most important thermodynamic event is the gel-to-liquid crystalline transition, which is a two-step, first-order endothermic process.\[[@CIT1][@CIT65]\] Fully saturated phosphatidylcholines (PCs) with identical fatty acid tails are among the most common lipids studied by DSC and exhibit sharp and symmetric chain-melting transition. The more commonly studied lipids (DMPC)(di-14:0) and DPPC (di-16 : 0) exhibit peaks at 24°C and 41°C \[[Figure 5](#F0005){ref-type="fig"}\].\[[@CIT10][@CIT65]--[@CIT67]\] Depending on the temperature range, scan rate, and fatty acid length (\> 14 carbons), a pre-transition peak (T~p~) that is typically of lower enthalpy and broader endothermic transition than L~α~ can be seen.\[[@CIT1][@CIT10][@CIT65]\] The temperature interval between the pre-transition and the main transition peaks decreases with increasing fatty acid chain length and both coincide at about 22 carbons.\[[@CIT1][@CIT65][@CIT68]\] However, the values reported for pre-transition show more disparity than the L~α~ data, due to a larger dependence on the scan rate and often the values are higher than the equilibrated data. The affects of the scan rate on T~p~ are stronger for PCs with more than 16 carbons.\[[@CIT1][@CIT10][@CIT65]\] Furthermore, even minor additions or impurities diminish or abolish the pre-transitional peak. The associated heat (∆H~cal~) is between 1.0 and 1.8 kcal / mol, and this transition is highly cooperative, involving several hundred lipids, independent of the chain length.\[[@CIT1][@CIT10][@CIT65]\] ::: {#F0005 .fig} Figure 5 ::: {.caption} ###### DSC Heating endotherm for DPPC MLV. All three transitions can be seen. (DPPC was equilibrated at 5°C for 2 days prior). Reprinted from Biochemistry 24, M. J. Ruocco, D. J. Siminovitch, and R. G. Griffin, Comparative Study of the Gel Phases of Ether- and Ester-Linked Phosphatidylcholine, 2406-241, 1985. With permission from American Chemical Society :::  ::: [Figure 5](#F0005){ref-type="fig"} illustrates all three transitions in a DSC scan of DPPC. From left to right one can find sub-transition (T~s~ = 21°C), pre-transition (T~p~ = 36°C), and the main transition (T~m~ = 41.3°C).\[[@CIT10][@CIT68]\] These values are dependent on the scan rate, as lower scan rates have resulted in lower T~p~ temperatures. PCs found in biological membranes, have both saturated and unsaturated fatty acid tails and exhibit considerably lower T~m~ values compared to disaturated PCs.\[[@CIT1][@CIT65]\] For 1-palmitoyl-2-oleoyl phosphatidylcholine (POPC), the T~m~ values have been reported to be between -5°C and 3°C, with an enthalpy (∆H ~cal~) of approximately 8 - 8.1 kcal.mol. Although there is a significant difference between the reported T~m~ of POPC and the disaturated DPPC (\~41°C), the ∆H~cal~ values for both are similar.\[[@CIT1][@CIT10][@CIT65]\] These PC bilayers have been used as model eukaryotic systems.\[[@CIT69]\] Typically unsaturated PCs are not studied using DSC, as the main transition falls below the operating range of most instruments, however, a review on mixed acyl chain PCs is available.\[[@CIT70]\] The impact of double bonds on the lipid phase behavior depend on their location and type.\[[@CIT10][@CIT65]\] Trans double bonds tend to have fewer effects on lipid packing than cis double bonds.\[[@CIT10][@CIT65][@CIT71]\] A systematic DSC-based screen of double bond position in an unsaturated fatty acid shows a characteristic U-shape when the T~m~ versus the double bond position is plotted. The minimum T~m~ is found when the double bond is in the center of the fatty acid chain.\[[@CIT1][@CIT10]\] This trend also holds true for the ∆H~cal~ with higher enthalpy found when the double bond is located at the beginning or the end of the fatty acid. Cis double bonds tend to decrease in order, which results in increased entropy due to an increase in free volume and the rotational degree of freedom, which is not seen for trans double bonds. This results in larger decreases of T~m~ for lipids containing cis double bonds.\[[@CIT10][@CIT65]\] The decreasing T~m~ directly affects entropy as follows: $$T_{m}\ = \ \Delta H^{deg}\ /\ \Delta S^{deg}.$$\[[@CIT10]\] Differential scanning calorimetry has also been used to investigate the effect of different salts on the thermotropic behavior of the PCs. Monovalent cations such as Na^+^ or K^+^ did not show much affect, even at high concentrations (1 M), on the properties of the pre- or main transitions.\[[@CIT1][@CIT65][@CIT72]\] However, the divalent Mg^2+^ and Ca^2+^ substantially changed the lipid phase behavior of the lipids.\[[@CIT73]\] 1M Mg^2+^ increased the melting temperature of the pre-transition, the main transition, and the enthalpy, whereas, low concentrations of Ca^2+^ (1 mM) have been shown to decrease the enthalpy of the pre-transition. The effects of Ca^2+^ are considerably stronger compared to Mg^2+^, as concentration above 10 mM induce a substantial increase of T~p~ and a moderate increase of T~m~.\[[@CIT1][@CIT65]\] At large concentrations (250 mM Ca^2+^) the pre-transition and main transition peaks merge together. Salt affects are more commonly seen for negatively charged phospholipids, as zwitterionic PCs tend to be less sensitive to cations.\[[@CIT1][@CIT65][@CIT72]\] The combination of the lipid head group structure as well as pH, salt concentration, and ionization states affect the thermodynamic properties.\[[@CIT10]\] Other Lipid Classes {#sec1-10} =================== The chemical structure of the polar head group affects the L~α~ transition via hydrogen bonding capabilities and electrostatic interactions.\[[@CIT1][@CIT6][@CIT10]\] DSC results for the polar and zwitterionic phosphatidylethanolamine (PE) vary with pH, due to different protonation states, however, most studies are done at neutral pH values and show consistent trends. Disaturated PEs have a higher T~m~ than the corresponding PCs due to the hydrogen bonding capabilities of the PE headgroup that adds stability.\[[@CIT1][@CIT74]\] Furthermore, the smaller headgroup of PE allows for closer interactions of the lipid molecules resulting in a more stabilized gel state. The T~m~ values increase with increasing chain length, similar to PC, and similar ∆H~cal~ values were reported for both disaturated PCs and PEs.\[[@CIT1][@CIT68]\] However, contrary to PC the disaturated PEs do not show any pre-transitions and an asymmetric main transition is evident. Differences in enthalpy have been observed for di-unsaturated lipids, whereby, the PE values are approximately half of what is seen for the corresponding PCs.\[[@CIT1][@CIT74]\] The cooperativity of the main transition is also reduced as disaturated PEs have CU values that are only about half of those for the equivalent PCs. Contrary to PC it has been found that due to the extra hydrogen bonding capabilities from NH~3~^+^ and PO~4~^-^ between separate bilayers, a tight interaction is formed, reducing the hydration levels,\[[@CIT74]\] in contrast to many other lipids. DSC has been used to analyze the hydration energetics of different PC and PE bilayers and the impact on the non-lamellar properties.\[[@CIT74]\] The pH of the solution will affect the T~m~ of the transition depending on the protonation / deprotonation state of the amino group. Deprotonation at low pH reduces the hydrogen bonding capabilities, and thus, decreases the T~m~ of DPPE from 63°C to 41°C.\[[@CIT74]\] Moreover, pH values below 8 increase the propensity for PEs undergoing a sharp transition from a lamellar to a hexagonal (H~11~) phase, within or above the L~α~ temperature.\[[@CIT74]\] The enthalpy of the non-lamellar phase is not easily detected by DSC and thus pH and temperature range needs to be considered when studying PEs.\[[@CIT1]\] Phosphatidylethanolamine can form either a lamellar or hexaganol phase depending on the type of acyl chains.\[[@CIT74]--[@CIT76]\] Short diacyl chains typically less than 12 carbons form lamellar phases, whereas, unsaturated systems yield hexagonal conformations, with the lamellar to H~II~ phase transition dependent on the number of hydrocarbons and frequency and position of the unsaturation.\[[@CIT74][@CIT77][@CIT78]\] PE lipids have a cylindrical shape, indicating equal size from the tails and headgroup, typically from short, fully saturated, hydrocarbon tails. A cone-shaped structure caused by a larger lipid tail region and smaller headgroup preferentially adopts a hexagonal phase.\[[@CIT74][@CIT75][@CIT79]\] The lamellar to hexagonal transition (T~H~) is found to occur at a minimum temperature when the unsaturation is closest to the middle of the acyl chain, as seen with T~m~ and PC.\[[@CIT80]\] The lowest T~H~ corresponded to an unsaturation at position 9, similar to the reported findings for the L~α~ transition for PCs and PEs.\[[@CIT80]\] Moreover, the non-lamellar properties of PE are being harnessed as a possible drug delivery mechanism, such that lipid-based nanoparticles incorporate PE hexagonal phase transition for drug release.\[[@CIT77]\] DSC has been found to show transition to hexagonal phases with higher sensitivity than ^31^P NMR or X-ray scattering, making it an ideal choice for many of the different drug studies.\[[@CIT78]\] An extensive review on the calorimetric behavior of different PE species is presented in\[[@CIT77]\] with the kinetics of PE transitions described in.\[[@CIT81][@CIT82]\] Furthermore, many different calorimetric studies have been compiled into the LIPIDAT database.\[[@CIT77]\] Phosphatidylglycerol (PG) is a major component of mitochondrial and chloroplast inner membranes as well as a pulmonary surfactant, but not a main structural component of mammalian membranes.\[[@CIT83][@CIT84]\] However, PG along with PE is one of the major lipids in bacterial membranes.\[[@CIT83][@CIT84]\] Thermograms of negatively charged PGs have generally been found to correlate well with PCs, as corresponding disaturated species sharing similar T~m~, ∆H~cal~, and entropy values.\[[@CIT1]\] The pre-transitional peak coincided with the PC studies, with disaturated PGs having a pre-transition with similar thermotropic properties and an absence of pre-transition, with di-unsaturated species. However, ion concentration and divalent Ca^2+^ and Mg^2+^ induce the formation of metastable complexes with PGs that are not seen with PCs.\[[@CIT85]\] PG has been found to exhibit a different melting regime with aqueous dispersions of DMPG at pHs higher than the pK~α~ and at high lipid concentrations of 70 -- 300 mM, revealing a very broad transition over an interval of about 10°C.\[[@CIT85]\] There appear to be at least two different phases existing, suggesting that DMPG forms a new phase at higher concentrations.\[[@CIT85]\] Using other biophysical techniques such as optical microscopy and X-ray scattering this phase has been identified at L~p~ (lamellar with pores), existing 3°C above the T~m~, prior to becoming a fluid phase past 30°C.\[[@CIT85]\] In addition to the main phases the stable subgel and the liquid crystalline lamellar phases, L~C~ and L~α,~ there are also metastable gel phases known as L~β'~ and P~β'~ under physiological conditions.\[[@CIT86]\] Low temperature incubation (4°C) of aqueous DMPG dispersions cause the lipids in the gel phase to transform into a highly metastable ordered solid quasi-crystalline bilayer, particularly for shorter chain lengths.\[[@CIT87]\] Freeze-fracture morphology has shown that two equal populations of a flat multilamellar sheet and a cylindrical shape occur when the phase transitions are monitored by DSC.\[[@CIT84]\] Upon cooling below the T~m~ the multilamellar aggregates dissociate forming unilamellar vesicles, which fuse to lamellar stacks upon low temperature storage forming a cylindrical shape.\[[@CIT84]\] Upon reheating, the main transition is considerably broader due to heterogeneous lipid conformations, and it occurs at a much higher temperature (40.3°C).\[[@CIT84]\] Due to polarity and charge of the head group, the pH values and ionic strength become major factors governing the T~m~ of the PG main transition. Low ionic strengths are characterized by a large gel-fluid transition approximately from 18 -- 35°C, which produces an optically transparent solution due to rearrangements in lipid packing.\[[@CIT83][@CIT88]\] This results in the transition usually being broken up into different calorimetric peaks called the T~m~^on^ and T~m~^off^, where structural changes occur between.\[[@CIT88]\] This usually correlates with a sharp decrease in turbidity at T~m~^on^ and an increase at T~m~^off^. The melting process is only fully completed at Tm^off^.\[[@CIT83][@CIT88][@CIT89]\] However, the exact structural characteristics of these transitions are still being determined. There is a hypothesis about a three-dimensional bilayer network as a possible structure.\[[@CIT88][@CIT89]\] Low pHs have induced T~m~ increases of 20°C for DPPG, attributed to a minimization of repulsive forces between the negatively charged headgroups. Anionic DMPG vesicles have been investigated with different Na^+^ concentrations, showing ionic-strength-dependent properties.\[[@CIT90]\] High salt concentrations result in a sharp shape indicating high cooperativity due to the shielding effect of the Na^+^ cation on a negative phosphate group.\[[@CIT90][@CIT91]\] On the contrary, significantly broader transitions were observed in distilled water, due to the absence of shielding.\[[@CIT90][@CIT91]\] Cardiolipin is a major component of mitochondrial membranes and regulates many different membrane bound enzymes.\[[@CIT92][@CIT93]\] Furthermore, it is present in the bacterial membranes as one of the anionic components.\[[@CIT93]\] For reviews on the thermotropic characteristics of CL and salt effects refer to.\[[@CIT92][@CIT94]\] CL still retains similar properties to other lipids with an increase in T ~m~, as there is an increase in acyl chains.\[[@CIT94]\] The T~m~ and enthalpy increases with a chain length similar to PG for CL, with the transition temperature being higher for PG.\[[@CIT92]\] Tetramyristoyl CL has two major endothermic transitions with similar enthalpy, however, the lower temperature transition is less cooperative and shows a cooling hysteresis.\[[@CIT93]\] CL has a propensity to form HII phases in the presence of high concentrations of salts or a decrease in pH, however, this is dependent on the amount of unsaturation and chain length.\[[@CIT94][@CIT95]\] Salts can be used to convert CL from a lamellar state to an inverted hexagonal phase.\[[@CIT95]\] CL typically converts to an inverted hexagonal phase at low pH values, when the phosphate group is protonated.\[[@CIT93]\] Tetraoleoyl cardiolipid showed this shift at NaCl concentrations of 3.5 M or higher.\[[@CIT95]\] Similar to PG, CL is sensitive to divalent cations, especially Ca^2+^, for which the pre-transition and main transition temperatures are raised.\[[@CIT94]\] Calorimetric studies with cholesterol have typically been studied with a lipid mixture check.\[[@CIT72][@CIT96]\] Cholesterol dramatically influences the phase transition by broadening the endotherm, with high concentrations eliminating the L~α~ transition.\[[@CIT72][@CIT96]\] Lower concentrations (\<10%) typically induce minimal phase separation. Added cholesterol increases the area of the gel monolayer due to the increasing disorder of the gel bilayer, while ordering the liquid crystalline state.\[[@CIT97]\] This behavior of the cholesterol facilitates the lamellar formation for many different lipid species including PG, PC, PE, and PS. There is a wealth of calorimetric studies on various cholesterol-lipid mixtures; this section will outline some finding from different mixtures, providing references to more comprehensive reviews. Two phases in PC and cholesterol mixtures are clearly present at 10 -- 25% cholesterol, revealing a liquid ordered cholesterol phase and a liquid disordered phase for the PC.\[[@CIT97]--[@CIT104]\] This has also been observed with SM.\[[@CIT98][@CIT105]--[@CIT109]\] Mixtures of cholesterol and PG bilayers show complete abolishment of the PG transition at 50% cholesterol for the dimyrstoyl species (DMPG).\[[@CIT99]\] However, longer acyl chains such as dipalmitoyl persist longer, with remnants of the transition still observable at above 50% cholesterol.\[[@CIT99]\] Differential scanning calorimetry of cholesterol has been applied to concepts such as the lipid-raft in cellular membranes and the existence of phase-separated fluid domains in cholesterol-lipid mixtures.\[[@CIT110]--[@CIT114]\] Cholesterol is one of the key lipids in eukaryotic cells, with essential roles in metabolism, hormone production, and formation of several vitamins.\[[@CIT115]\] Aside from the possible lipid rafts the role of cholesterol on ordering adjacent lipids has been studied with a recent review presented in.\[[@CIT113][@CIT115]\] Additionally mixtures of PC, SM, and cholesterol have been used to form raft micro-domains, as different concentrations of components result in different phase formations, which can act as potential targeting sites of pharmaceutical products.\[[@CIT96][@CIT110][@CIT111][@CIT114][@CIT116]--[@CIT118]\] A good review on calorimetry of lipid mixtures is presented in.\[[@CIT61]\] Such studies include the thermotropic analyses of the mixture of PC and PG of varying chain lengths, and the hydrate states examined have been at different pH.\[[@CIT83][@CIT119][@CIT120]\] Data from these studies have been used to formulate phase diagrams providing information about the mixing behavior of the different systems.\[[@CIT119][@CIT121]\] At neutral pH the phase boundaries are close together with a narrower coexistence between the two compared to pH 2.\[[@CIT119][@CIT120]\] Many comprehensive calorimetric reviews on different lipid mixtures are available from diacylglycerol mixtures with phospholipids,\[[@CIT122]\] phosphatidylserine, and cholesterol,\[[@CIT123][@CIT124]\] PC and PG,\[[@CIT125]\] and PE : PG mixtures, with differing chain length and pH.\[[@CIT126]\] Based on prior studies, mixing lipids with similar thermodynamic properties results in traces that retain similarities to the pure components, however, with an increased asymmetry and T~1/2~.\[[@CIT10]\] Moreover, the amount of similarity retained in pure components depends on the composition of the two components and the interaction between the polar and non-polar portions.\[[@CIT1][@CIT10]\] This has primarily been done by constructing phase diagrams \[[Figure 6a](#F0006){ref-type="fig"}\]. Phase diagrams use the onset and completion temperatures of phase transitions, the T~1/2~ and enthalpy for different lipid mixtures, and reveal the affect of different compositions on T~m~.\[[@CIT1][@CIT10]\] The comparison to theoretical curves is used to evaluate the phospholipid mixtures.\[[@CIT1]\] In cases where thermodynamic characters of the lipid components are quite different, the thermogram becomes complex and highly dependent on the concentration, as phase separation and demixing can occur \[[Figure 6b](#F0006){ref-type="fig"}\].\[[@CIT10]\] ::: {#F0006 .fig} Figure 6 ::: {.caption} ###### a\) A phase diagram showing the deviation from Ideal Mixing b) DSC thermogram of a DMPE DSPC mixture. Reprinted by permission from Proc. Natl. Acad. Sci. 73, S. Mabrey and J.M. Sturtevant (1976) Investigation of Phase Transitions of Lipids and Lipid Mixtures by High Sensitivity Differential Scanning Calorimetry 3862-3866 with permission from Proceedings of the National Academy of Science United States\[[@CIT191]\] :::  ::: Biomimetic liposome systems have long been used as simplified model membranes for many membrane-drug or membrane toxicity investigations.\[[@CIT28][@CIT29][@CIT127]\] A recent review correlates the toxicity of various compounds, such as, xenobiotics, detergents, and peptides, with established toxicity assays showing a good correlation between the two.\[[@CIT127]\] Analysis of multi-lipid membranes has been used to study biomembranes of eukaryotic cells, revealing domains and organization of lipids, for potential roles in signaling or recruitment.\[[@CIT128][@CIT129]\] Studies of domain components and phase behaviors allows for potential targeting sites for potential pharmaceuticals.\[[@CIT128]\] Differential scanning calorimetry has also been used to determine the lateral heterogeneity of membranes as preferential lipid-lipid interactions result in a clustering of lipids.\[[@CIT98]\] Such data provides an insight into the fluid phase of the membrane, as lateral organization and lipid targets may provide information for potential drug targets.\[[@CIT98][@CIT114][@CIT130]\] Clustering of lipid components can be revealed as pure lipid domains demixed in lipid mixtures. Furthermore broadening of the endotherms suggests mixing between the two components.\[[@CIT130]\] For reviews on lipid domains and calorimetry of different lipid mixtures see,\[[@CIT98][@CIT105][@CIT131]\] and in particular.\[[@CIT129]\] Calorimetric lipid analysis has also been applied to lipid components in biological membranes, such as aggregates of macrodomains in mammalian blood platelets, in order to evaluate the stability of the platelets during freeze drying, for therapeutic storage.\[[@CIT117]\] The first successful DSC studies were done on the prokaryote *Acholeplasma laidlawii*,\[[@CIT4]\] *Halobacterium Halobium*, and unsaturated fatty acid auxotrophs of *Escherichia coli* \[[Figure 7](#F0007){ref-type="fig"}\]. Analysis of the thermotropic data have shown that 90% of the membrane extracts undergo a cooperative transition, even with the multiple lipid species, and membrane protein DSC analysis of *E. coli* membranes showed the absence of a visible gel phase.\[[@CIT98][@CIT105][@CIT132]\] For reviews on the heterogeneity of biological membranes and the different mimetic lipid mixtures refer to.\[[@CIT133][@CIT134]\] ::: {#F0007 .fig} Figure 7 ::: {.caption} ###### DSC scans of a) A. laidlawii\[[@CIT192]\] a) whole cell, b) isolated membrane, c) protein denatured, d) aqueous MLV of membrane lipids. b) DSC of whole cell E Coli dashed trace is for E Coli ribosomes. Reprinted from Chemistry and Physics of Lipids, Vol. 30, R.N. McElhaney, The use of differential scannning calorimetry and differential thermal analysis in studies of model and biological membranes, 229-259, 1982, with permission from Elsevier :::  ::: Drug Purity {#sec1-11} =========== Physical constants and purity profiles of drugs have been determined by using differential scanning calorimetry.\[[@CIT14][@CIT135]\] The latter can be assessed by the melting behavior observed in the recorded thermograms. Peak integration, according to the van't Hoff relationship, and T~m~ values are used for batch-to-batch consistency and to test for impurities that will change the melting profile.\[[@CIT135]\] Although it is difficult to quantitatively measure the percentage or type of impurities, DSC provides a quick and reliable means of establishing batch variability and is a qualitative screen for contamination.\[[@CIT14][@CIT135][@CIT136]\] Nevertheless multiple techniques will be necessary to allow proper quantitative analysis.\[[@CIT135][@CIT136]\] The main application of DSC to purity relies on the notion that impurities reduce the melting temperature of the drug.\[[@CIT137]\] The melting temperature is a strong indication of drug purity and DSC not only allows for a quick screening of the T~m~, but the resolution of the peak (T~1/2~) will relate to populations of drug that may be in a different conformation or interacting with an excipient resulting in a shoulder region.\[[@CIT14]\] The amount of impurities is derived from van't Hoffs's law for diluted solutions: $$X\ = \ \left( {- \Delta T\ \times \ \Delta H_{f}} \right)\ /\ {RT}.^{2}$$ with X equaling the mol fraction of impurity, ∆T representative of the melting point depression, T. equal to the melting point of the pure substance, R, the gas constant, and ∆H~f~, the enthalpy of the pure material.\[[@CIT137]\] Most results have been highly complementary to chromatographic data.\[[@CIT137]\] Nonetheless, DSC in purity analysis has become increasingly popular, due to the low quantity of sample required (1 -- 2 mg) and the relatively quick analysis time.\[[@CIT137]\] Another aspect of purity is drug polymorphism, which is related to the different crystalline states.\[[@CIT19][@CIT33][@CIT138]\] As pharmaceutical processing results in multiple polymorphs, the bioavailability of the key state of the drug as well as the potential health risks of different states require scrutiny in testing.\[[@CIT33]\] The polymorphic transitions can be measured using DSC and phase diagrams can be constructed, respectively.\[[@CIT33][@CIT121][@CIT138]\] Polymorphs typically exhibit similar properties in the gaseous and liquid states, however, show differences depending on the solid state. The most commonly analyzed states are the amorphous state, crystalline state, and glassy state.\[[@CIT19][@CIT138]\] Amorphous relates to a non-ordered system, whereas, glassy state refers to an amorphous solid that undergoes a glass transition, forming a rubber like appearance.\[[@CIT33]\] The glass transition (T~g~) is the transition that occurs in amorphous materials, as the heat capacity undergoes a quasi-discontinuous change to a higher value. Another analyzed transition is exothermic crystallization, which occurs as the amorphous solid turns crystalline or semi-crystalline, usually lies in between the glass transition and the T~m.~\[[@CIT16]\] For reviews on distinguishing between amorphous and crystalline states of a drug by DSC, refer to.\[[@CIT43][@CIT139]\] Amorphous pharmaceutical solids are typically less stable than their crystalline counterparts and the addition of excipients have a tendency to exist as an amorphous solid.\[[@CIT43]\] Typical pharmaceutical preparative techniques such as lyophilization, milling or wet granulation lead to amorphous conformations.\[[@CIT19][@CIT33][@CIT139]\] Hence, DSC has been used to study the thermodynamic differences between the amorphous form and crystalline form as well as to identify a coexistence between both.\[[@CIT43]\] Crystallization is often exothermic, whereas, amorphous compounds do not recrystallize and the enthalpy recorded can be analyzed quantitatively, to determine the drug state.\[[@CIT14][@CIT32][@CIT43]\] The enthalpy of the peak can be used to determine the purity of the peak. As an advantage of DSC over capillary melting point approaches, separate melting transition or polymorphs and recrystallization events can also be observed, which provide information about sample fusion or impurities.\[[@CIT32]\] In theory, a completely pure crystalline sample should yield an infinitely narrow transition, whereas, increased broadening is associated with impurities.\[[@CIT136]\] Using van't Hoff's law of melting point depression, a straight line is seen when temperature is plotted against the inverse of the molten fraction of a sample.\[[@CIT136]\] For a full review on van't Hoff's law applied to DSC purity profiles, see van Dooren and Muller, 1984.\[[@CIT136]\] Subsequent to linearization, deviations from a profile expected for a pure compound can also provide information on sample stability. Moreover, other aspects such as changes in sample size or curvature due to the formation of precipitates will affect the validity of the analysis. Furthermore van't Hoff's analysis of impurities requires that its contents do not change over time, hence, evaporation or decomposition would affect the results.\[[@CIT136][@CIT140]\] Therefore, different scanning rates are usually compared, to observe the effects of evaporation or decomposition, especially around the melting region.\[[@CIT136]\] Nevertheless, impurities have been determined to be particularly accurate in 98% of the pure samples of various organic chemicals such as organophosphates, urea, amides, esters or halogenated compounds.\[[@CIT140]\] Many pharmaceutical products can be present in different conformations, with distinct chiral structures, which alter their desired effect.\[[@CIT48][@CIT141]\] Racemic compounds usually exhibit different thermal events, therefore, it is possible to detect as low as 1.5% of an isomer in a mixture of almost pure ephedrine hydrochloride.\[[@CIT141]\] Although difficulties may arise from overlapping thermal events within the same temperature range, deconvolution with non-linear regression has been successful to distinguish isomers.\[[@CIT141]\] Although there are other methods of determining chiral purity, DSC requires a minimal amount of material, as 1 -- 5 mg is adequate for most applications.\[[@CIT141]\] High sensitivity, reliability, and the relative speed of the assays provide a quick purity screen for different drug batches. This is very valuable for pharmaceutical applications, as isomers possess differential absorption, and altered potency and metabolism or pharmacological behavior.\[[@CIT141]\] Moreover, the determination of water and hydrate content is important as most drugs are hygroscopic and the primary solvent for crystallization is water.\[[@CIT32]\] During the crystallization process of many pharmaceutical compounds, solvents are incorporated into the crystal lattice affecting properties such as solubility, stability, and pharmokinetics.\[[@CIT142]\] Water content is a critical parameter in drug development as the water activity may vary with different hydrates existing in the same drug product (hydrate polymorphs).\[[@CIT33]\] Usually determination of the water content is done by thermogravimetric analysis (TGA), Karl Fischer titrimetry (KFT) or evolved gas analysis.\[[@CIT143]\] However, DSC has been applied, to determine the water stoichometry in different drug hydrates, under the assumption that the enthalpy of dehydration (∆H~d~) is equivalent to the enthalpy of vaporization (∆H~v~) of water.\[[@CIT142]\] Results correlated well with values from KFT, with the additional benefit of information on the potential location of water binding based on hydration enthalpies. Similar to chirality studies, the technique is limited to overlapping hydration peaks, however, used in conjunction with other techniques it provides a quick and reliable screening method of hydrate content.\[[@CIT142]\] Drug Stability {#sec1-12} ============== Product stability is essential and is usually described by the equilibrium constant (K) or the free energy (∆G°).\[[@CIT48]\] These values can be determined indirectly from the measured enthalpy through thermodynamic correlations such as the van't Hoff equation,\[[@CIT48]\] allowing the use of DSC to screen the stability of potential drugs or drug delivery systems. Progressive scans can be used to analyze stability and to assess the denaturation temperature, as gradually changing compounds will yield a different profile. Comprehensive reviews on drug stability, particularly liquid particles are presented in.\[[@CIT42][@CIT144][@CIT145]\] Pharmaceuticals applications of proteins depend on a properly folded state.\[[@CIT48]\] As a denatured protein has a higher heat capacity than its native form (∆C~p~), an increase in this parameter can be used to determine the extent of denaturation, with progressive cycling, over an extended period of time.\[[@CIT48]\] Moreover, the melting temperature can also be used as an indicator of thermostability as a higher T~m~ represents a more stable protein that is less susceptible to denaturation.\[[@CIT48]\] Drying of proteins for pharmaceutical applications can impact their conformation, and hence, reduce the potency of the drug.\[[@CIT146][@CIT147]\] DSC has been used to gauge different drying techniques for potential pharmaceutical applications. Techniques such as spray drying, lyophilization, super critical fluid technology, and many others have been proven challenged to maintain protein stability under high temperature, freezing, and dehydration.\[[@CIT19][@CIT146]--[@CIT149]\] A more in-depth analysis of these problems is presented in a review,\[[@CIT146]\] showing how DSC can be used to evaluate potential methods for pharmaceutical protein preparation. DSC also presents an advantage, as a high throughput screening means establishing protein changes quickly and easily, based on different mutations or preparations.\[[@CIT42][@CIT144][@CIT145]\] Lyophilization of liposomes lowers the potential hydrolysis of phospholipids and physical degradation of the vesicles extending the life of the drug carrying molecules.\[[@CIT19][@CIT149][@CIT150]\] However, such processes are not without faults, as physical changes may occur, resulting in the damage of the liposomes, releasing the encapsulated agent.\[[@CIT149][@CIT150]\] Lyophilization of liposomes is explained in detail in.\[[@CIT150]\] Furthermore drug-liposome stability is presented in,\[[@CIT16]\] with the kinetics of liposome phase transition in.\[[@CIT151][@CIT152]\] One particular relevant problem for the pharmaceutical industry is drug-excipient interaction.\[[@CIT153]\] The latter is an inactive substance used to carry the active compound or to minimize drug degradation upon delivery. Drugs and excipients were incubated at a set temperature for a period of time, followed by an increase to a higher temperature, and subsequent isothermal incubation. Analysis of the thermograms will illustrate any changes to the compound, such as, degradation or interactions between the excipient and drug at higher temperatures.\[[@CIT42][@CIT153]\] Furthermore, the sample environment could be easily controlled in the instruments, allowing incubation at high humidity or temperature, to simulate long-term exposure.\[[@CIT153]\] In addition to the simulated storage of the drug products, choices of different salts and drying methods have been investigated using DSC.\[[@CIT14]\] The effects of the coating on different drugs and delivery systems such as nanoparticles (mentioned in the next section) have resulted in characteristic shifts and decreases in enthalpy or the T~m.~\[[@CIT14]\] Stability under a wide range of conditions has to be studied, as batch-to-batch variation can result in different polymorphs, as observed with different photochemical stabilities.\[[@CIT33]\] Problems may arise when using DSC to screen for excipient compatibilities, as it is required to use high temperatures at set heat rates. Hence, inconsistencies between reactions at ambient temperatures and pressurized cells can occur.\[[@CIT34]\] The properties of vitamin B6 in different excipients, such as mannitol, were used to gauge changes to the properties of the drug.\[[@CIT154]\] In conjunction with other highly sensitive thermal techniques, such as Micro-thermal analysis (µTA), the changes in thermotropic properties could be used to select the best suited excipient.\[[@CIT154]\] The study of cyclodextrins as an excipient has been of great interest in the pharmaceutical field as the torus-shaped, cyclic structure allows for encapsulation of drug molecules inside a less hydrophilic cavity, compared to the aqueous solvent\[[@CIT155]\] Three different drugs trimethoprim, sulfadiazine, and sulfamethoxazole, with natural cyclodextrins (α,β,γ) were studied in both the aqueous and solid states, showing lower stability of the drugs in the amorphous state and solubilizing properties depending on the carrier size of the cyclodextrin.\[[@CIT155]\] DSC-derived excipient compatibility is usually compared to spectroscopic results obtained by UV or IR or to chromatographic HPLC analysis.\[[@CIT156]\] Differential scanning calorimetry has been very effective in determining the physiochemical properties of different pharmaceutical products, thus facilitating design of new drugs or improving modifications of the existing compounds.\[[@CIT154]\] With the abilities to test both drug and excipient for purity, stability or pharmacological properties, DSC is becoming increasingly popular in the pharmaceutical industry.\[[@CIT154]\] Nanoparticles {#sec1-13} ============= Thermal analysis can also be used to analyze the incorporation of drugs into nanoparticles via examining enthalpy change.\[[@CIT15][@CIT157]--[@CIT159]\] Liposomes have been used to penetrate skin for drug delivery and localized drug delivery.\[[@CIT160]\] DSC is one of the primary tools used for the characterization of the matrix state, with polymorphism and drug incorporation in lipid dispersions.\[[@CIT161]\] Nanoparticles tend to have a decreased melting temperature compared to bulk material that is not in the nanometer size.\[[@CIT161]\] Lipid polymorphism is commonly found in lipid nanoparticle dispersions with various components affecting molecular packing, which is reflected in the different melting points and enthalpies.\[[@CIT161]\] Furthermore, the smaller radius prevents optimal lipid packing of the lipid acyl chains, thus lowering the energy required for the phase transitions. Broadened profiles are usually attributed to the addition of multiple different lipid components, as well as size differences.\[[@CIT161]\] For a review on liposome drug delivery refer to\[[@CIT162]\] and for nanoparticle drug interactions.\[[@CIT163]\] Analysis of drug loading efficiencies is quite complicated, as the drug typically interacts with the lipids inducing a shift in the phase transition temperature.\[[@CIT161]\] Moreover, the enthalpy of the transition may also be reduced as a population of lipids is interacting with the drug solubilized in the matrix.\[[@CIT161]\] This can easily be used to identify if the drug is miscible in the melted state of the liposome. Most studies presume changes to the lipid thermogram and a negative shift of the matrix lipid T~m~ to be a sign of drug incorporation. However, in some cases it has been reported that decreases in enthalpy can be attributed to lipid dissolution or aggregation of drug molecules within the nanoparticles.\[[@CIT161]\] Improved efficacy of different drugs has been studied using nanoparticle delivery systems.\[[@CIT164]\] A potent cancer fighting drug, Paclitaxel, has difficulties in administration, due to poor solubility in water and with excipients. Nanoparticles composed of biodegradable polymers with poly(lactic-co-glycolic acid) have been used to encapsulate the drug within the nanoparticles, using emulsifiers such as cholesterol and phospholipids.\[[@CIT164]\] DSC allowed for comparison of the thermodynamic properties, as the T~m~ of Paclitaxel and the nanoparticle carriers were analyzed, to screen for undesirable changes to the drug.\[[@CIT164]\] DSC was also used to record the transition of non steroid anti-inflammatory drugs (NSAIDs) from a crystalline to an amorphous state upon encapsulation in polyethylene glycol (PEG), a solid drug carrier, accompanied by a decrease in endothermic transition over time and progressive scanning.\[[@CIT165]\] Similar studies were performed with solid lipid nanoparticles (SLN) prepared from oil-water microemulsions to encapsulate the drug diazepam.\[[@CIT166]\] Solid lipid nanoparticles are emerging as a potential application in drug delivery, due to their low toxicity and their ability to maximize drug incorporation for secondary and tertiary drug targeting.\[[@CIT157][@CIT158]\] Thermograms of crystalline diazepam and the drug loaded SLN particle, showed that the melting peak for the drug was not observed in the loaded nanoparticles, indicating an amorphous solid in the SLN.\[[@CIT166]\] Moreover, the solid state particles can exist in polymorphs, pseudopolymorphs, and even amorphous solids.\[[@CIT167][@CIT168]\] DSC melting profiles are essential for identifying the state of the drug, which can significantly influence bioavailability, stability, and water content.\[[@CIT157][@CIT169]\] This is essential for determining the proper state, for the active pharmaceutical ingredient.\[[@CIT169]\] Relating to drug stability, the lipids in SLNs are considered excipients, with different lipids and surfactants studied and presented in a recent review.\[[@CIT170]\] Thermotropic analysis of the SLN particles indicated that the chemical stability of the lipid is not affected during formation with a low level of degradation (2--5%) for the majority of lipids and a maximum of 10% reached after 24 months.\[[@CIT170]\] However, other lipid excipients such as lecithin have shown strong decomposition, minimizing its potential in SLN.\[[@CIT170]\] Structural properties and thermodynamic characteristics of nanocrystallization have been studied in detail in.\[[@CIT139]\] In addition, DSC was used to analyze stability and drug dissolution from nanostructure lipid carriers (NLC) composed of a solid lipid matrix with a liquid lipid nanocompartment core.\[[@CIT159]\] Drug release from three-dimensional polymer hydrogel systems is a growing field in biomedical drug delivery.\[[@CIT171]\] Site-specific targeting and release increases the bioavailability. Thus, it is important for pharmaceutical testing, to understand the interaction between nanoparticle carriers and drugs as well as nanoparticles and biological membranes. Different nanoparticle polymers and various cross-linkers have been used to modulate temperature-dependent drug release *in vitro* and some have been shown to obstruct drug diffusion and incorporation.\[[@CIT171]\] The application of dendrimers for drug delivery to the lungs was probed by studying their interactions with DPPC liposomes, as the latter is the main component of a lung surfactant.\[[@CIT52]\] Changes to the lipid phase transition are used to determine properties such as incorporation of the dendrimer into the bilayer as well as the strength of interaction, based on the overall structure and hydrophobicity of the dendrimer.\[[@CIT172]\] The manufacturing of plastics and rubber relies on plasticizers, to enhance the flexibility of polymers.\[[@CIT173]\] However, recent health concerns and increased industrial standards require more testing of the toxicity of these compounds. Thus, plasticizers such as dimethylsebacate (DMS), diethylsebacate (DES), and dibutylsebacate (DBS) were tested with DPPC,\[[@CIT173]\] which serves as a good model for the lung surfactant. The thermotropic data provided information on the extent of the interaction and potential penetration,\[[@CIT173]\] as large concentrations of plasticizers resulted in a complex transition and the coexistence of new phases and aggregates. Thus, these DSC-based results indicated negative health effects due to exposure to plasticizers.\[[@CIT173]\] Antimicrobial Peptides {#sec1-14} ====================== The increasing presence of antibiotic-resistance bacterial strains has increased the interest in antimicrobial peptides.\[[@CIT174]\] In the field of novel peptide antibiotics, which are known to elicit their properties on the biological membranes, the study of peptide-lipid interactions is crucial in their design and development, as well as, in the understanding of molecular mechanisms.\[[@CIT174]\] Many peptides and analogs have been designed with the intention of specifically targeting bacterial phospholipid classes. As mentioned earlier, the peptide-lipid interaction can be observed based on a change in the phase transition. DSC has provided quantitative information on the effect of the peptide interaction on the membrane structure by comparing the thermotropic data of the lipid blank and the sample with the peptide in a concentration dependant manner.\[[@CIT174]\] Immediately evident from the thermogram's preferential interaction with different lipid classes provides an insight into the type of binding.\[[@CIT174]\] Furthermore, the role of membrane perturbation and surface defects caused by protein-lipid interactions has implicated antimicrobial peptide activity based on phase separation, charge-charge interaction, membrane curvature strain, pore formation, and even detergent style effects.\[[@CIT174]\] Different strains of bacteria have different phospholipid compositions in their membrane. However, the primary phospholipids are PG, PE, and CL. Membranes composed of PG have been used as the main model for binding to bacterial cell membranes.\[[@CIT125]\] Depending on the species and whether the bacteria are gram negative or positive, the percentage of PG can range from 6 -- 90%, with just as large ranges for PE and CL.\[[@CIT174]\] Mammalian erythrocyte membranes contain mainly PC and SM on the outer leaflet and PE and PS are found in the inner leaflet.\[[@CIT174]\] Using different biomimetic lipid mixtures, liposomes can be used as a model of either human or bacterial membranes. Prior DSC studies have shown that depending on the antimicrobial peptide there may be preferential interaction with certain lipid components.\[[@CIT174]\] This has been documented with the peptide cinnamycin specifically interacting with PE and sapecin interacting with cardiolipin.\[[@CIT174]\] However, the affinity of antimicrobial peptides has been predominantly to negatively charged phospholipids such as magainins, with greater bactericidal activity for membranes with higher PG concentrations.\[[@CIT174][@CIT175]\] Studies with different antimicrobial peptides have shown a preference for binding PG as opposed to other negatively charged lipids, such as, phosphatidic acid, phosphatidylserine, and cardiolipin.\[[@CIT176]\] Phase separation was consistently seen with the different peptides and PE / PG mixtures, with a preferential interaction with PG. Less pronounced thermotropic changes were predicted to be due to the rigidity of the CL membrane reducing peptide penetration.\[[@CIT176]\] DSC has also been effective in showing that the interaction is not purely electrostatic, as strong hydrocarbon chain disruption is evident.\[[@CIT176]\] Lipid specificity and effects of lipid chain length have been identified with protegrin-1, as only minor thermotropic changes with PA have been observed, whereas, major perturbations were observed with PG.\[[@CIT177]\] Interaction with peptide-rich and peptide-poor regions are observed in PG membranes in analogy to other antimicrobial peptides, such as HNP-2.\[[@CIT177]\] Furthermore, a decrease in enthalpy with added peptide concentration suggests a concentration-dependent binding.\[[@CIT177]\] Interaction is not solely headgroup-dependent as these effects seen with DMPG and DPPG are not mirrored with DSPG, as only a slight increase in T~m~ was recorded. This suggests an impact of lipid packing as the increased chain length results in more non-covalent interactions.\[[@CIT177]\] Human neutrophils derived (Human Neutrophil Peptide) HNP-2 from the defensin class of peptides, showed a very high specificity for bacterial membranes over mammalian membranes.\[[@CIT178]\] This was accurately portrayed using model membranes and DSC with PC, SM, PE, and PG. Minimal changes to the PC or PC / SM mixtures were observed, with varying ratios of peptide, however, a proportional increase for the T~m~ of DPPG was found for the main transition.\[[@CIT178]\] Such behavior was consistent with a preferential stabilization of the gel phase by HNP-2. This was unique, as the majority of amphipathic peptides had limited interaction with the gel phase.\[[@CIT178]\] Gel phase interaction was confirmed by using DSC, by incubating HNP-2 / DPPG mixtures above and below the T~m~ and measuring the heat capacity of the L~α~ transition. Similar values were obtained with both incubations suggesting strong electrostatic interactions.\[[@CIT178]\] Such behavior has also been observed for tachyplesin I from horse shoe crab, which has a rigid β-sheet structure due to disulfide bridges.\[[@CIT178]\] Differential scanning calorimetry was also used to show PG lipid segregation upon interaction with PGLa or HNP-2 from peptide-rich and peptide-poor domains, with a new transition occurring above the original T~m.~\[[@CIT179]\] The peptide had a preferential interaction with PG liposomes, showing phase separation with peptide-rich and peptide-poor domains for PE / PG and PG membranes.\[[@CIT180]\] Furthermore, minimal peptide interaction was observed for biomimetic eukaryotic vesicles.\[[@CIT174]\] The peptides effect on the main transition was the aspect primarily studied using DSC, however, changes to the pretransition were also used to implicate packing defects caused by peptides.\[[@CIT181]\] Cathelicidins, indolicidin, and tritrpticin were all studied with varying model vesicle systems, such as, DMPC, DMPC / cholesterol, and DMPG, evaluating the different modes of interaction. In addition to abolishing the pre-transition of PC, cholesterol / PC demixing was observed suggesting the possibility of peptide-induced cholesterol domains.\[[@CIT182]\] Opposite effects have been observed for protegrin-1, where new phase transition occurred below the T~m~. Such examples elucidate the ability of scanning calorimetry to establish the induction of lateral separation into domains as a possible mechanism of action.\[[@CIT179]\] This suggest a preferential interaction with PG components of bacterial membranes resulting in a thermogram with strong PE contribution.\[[@CIT179]\] Such behavior has been recorded for many antimicrobial peptides such as magainin II, buforin II, tachyplesin, protegrin I, and gramicidin S.\[[@CIT179]\] Such similarities between peptides that adopt different conformations, suggest that this may be a key trait to search for, when screening potential peptides using DSC. Furthermore, this indicates the crucial role of PG interaction and domain formation for bactericidal effects.\[[@CIT179]\] Peptide LL-37, a member of the cathelicidin family, also showed chemotactic action on cancer and transformed cells.\[[@CIT183]\] With its high cationic charge (+ 6) and its alpha helical propensity upon binding to lipid membranes, its interaction with various model membranes was studied via DSC.\[[@CIT183][@CIT184]\] Typical loss of pre-transition followed by lower cooperativity prior to the formation of multi-phase transition\[[@CIT184]\] identified that the peptide interacted with the bilayer core of the membrane, reducing cooperativity and disrupting lipid packing.\[[@CIT184]\] The peptide was shown to be highly active reducing the sharp cooperative DPPC transition to a broad endotherm with a reduced enthalpy at an LL-37 ratio of 4%, correlating to small and wide angle X-ray scattering (SAXS). This broad region had been identified as the lamellar state, changing to disk-like micelles.\[[@CIT183]\] The peptide induced an even stronger effect with DPPG, with the main transition replaced with two overlapping transitions that were consistent with interdigitated and non-interdigitated domains.\[[@CIT183]\] The calorimetric data was utilized to make a phase diagram, indicating that the ability to form the quasi-interdigitated state depended on the helix topology, angle of hydrophobicity, and cationic distribution.\[[@CIT183]\] Interdigitation has also been observed with other peptides such as melittin, however, this was found with PCs at a mol ratio of 8%.\[[@CIT183]\] Formations of other peaks is quite common, as observed for cecropin B and B3.\[[@CIT185]\] Such interactions have been found to be concentration-dependent, with low concentrations (\~1 µM) typically broadening the profile, whereas, 20 times higher concentrations result in two shoulders above and below the pure lipid phase transition.\[[@CIT185]\] It has been suggested that the two phase transitions result from aggregation due to high peptide concentrations and pore formation, where two populations exist, one for lipids in the pore formation and the other for lipids not in the pore formation. Such examples of multiple peaks are found in defensins, magainins, and gramicidin S.\[[@CIT185]\] Poly(L-lysine) or polyarginine peptides were used as model peptides to study the electrostatic interaction between peptides and different model lipid systems.\[[@CIT186]\] One such study used poly(L-lysine) peptides with various anionic lipids and lipid mixtures, which were studied via DSC.\[[@CIT186]\] Using model membranes the effects of chain length, peptide concentration, and charge density were investigated with DPPG or DPPG / DPPC or DPPG / DMPC membranes.\[[@CIT186]\] Peptide binding increased the T~m~ of the pure DPPG lipid with a stronger shift for longer peptides, probably from charge stabilization of the PG lipids in the gel phase. Mixed PC and PG systems yielded similar trends for other antimicrobial peptides inducing phase separation in the gel phase of the membrane preferentially interacting with the PG in the system leaving a relatively undisrupted PC L~α~ domain.\[[@CIT186]\] Differential scanning calorimetry studies of the antimicrobial peptide Gramicidin S and MLVs composed of DMPC, DMPG, and DMPE have shown the dependence of interaction on the charge and structure of the headgroup\[[@CIT174]\] \[[Figure 8](#F0008){ref-type="fig"}\]. In respect to T~m~, enthalpy, and cooperativity, the peptide induced the largest differences with DMPG, whereas, DMPC had a moderate, and DMPE a comparably lower extent of interaction.\[[@CIT174]\] Interestingly, through repeated heating cycles of DSC, the GS bound to the DMPG vesicles protected the phospholipids from hydrolysis, due to their consistent exposure to high temperature\[[@CIT174]\] \[[Figure 8](#F0008){ref-type="fig"}\], suggesting a localization close to the phosphate groups. GS was also shown to suppress the pre-transition of DMPC at very low concentrations, increasing the width of the transition, which clearly indicated a perturbation of the bilayer.\[[@CIT187]\] Furthermore, GS induced an irreversible conversion to the L~α~ / HII phase with PE species after exposure to high temperatures, confirmed via ^31^P-NMR spectroscopy.\[[@CIT187]\] However, these results were not observed with the PG or PC species, indicating that the non-lamellar phase formation was headgroup-dependent.\[[@CIT187]\] Other peptides, such as protegrin-1 and PGLa also induced nonlamellar phases, like the cubic phase seen with GS.\[[@CIT183]\] For a recent review on nonlamellar phases and antimicrobial activity refer to.\[[@CIT188]\] ::: {#F0008 .fig} Figure 8 ::: {.caption} ###### DSC endotherms for the given lipid to peptide ratios for Gramicidin S and DMPG MLVs. Reprinted from Biochimica et Biophysica Acta (BBA) -- Biomembranes, Vol. 1417 Iss. 2, E.J. Prenner, R.N.A.H. Lewis, L.H. Kondejewski, R.S.Hodges, R.N. McElhaney, Differential scanning calorimetric study of the effect of the antimicrobial peptide gramicidin S on the thermotropic phase behavior of phosphatidylcholine, phosphatidylethanolamine and phosphatidylglycerol lipid bilayer membranes, 211-223, 1999, with permission from Elsevier\[[@CIT193]\] :::  ::: The number of scans required for the peptide-liposome interaction to equilibrate has been useful in determining the strength of the interaction as well as possible peptide redistribution.\[[@CIT182]\] This provides an insight into the mode of binding as DSC thermograms with tryptophan-rich cathelicidin show a progressive decrease in enthalpy and cooperativity of a specific component of the phase transition with each scan.\[[@CIT182]\] This can relate to the reduction of peptide-rich and peptide-poor regions, with increased cycling, due to increased lipid-peptide interaction. Also slow equilibration could be related to the possible release of peptide from the bilayer, upon cycling, indicating low affinity.\[[@CIT182]\] Antimicrobial peptide lytic activity was studied with the help of DSC. Using the phase separation data with liposomes composed of PE and CL, it was observed that different antimicrobial peptides were efficient in separating PE from CL seen with two PE populations, one resembling pure PE and the other a CL-containing broadened PE peak.\[[@CIT98][@CIT105][@CIT131]\] Furthermore, a preferential interaction of peptides with certain membrane components was observed, based on the presence of peptide-free and peptide-bound domains.\[[@CIT98][@CIT105][@CIT131]\] Based on thermotropic data the effects of hydrocarbon backing, headgroup mismatch, membrane curvature, bilayer destabilization, and electrostatic repulsion had all been shown to be relevant for peptide vesicle interaction.\[[@CIT174]\] DSC has shown that antimicrobial peptides discriminate between different types of phospholipids, have a preferential interaction with bacterial PG membranes, and show little interaction with major components of mammalian membranes. However, cases have been reported where minimal shouldering regions are observed, suggesting that there may be an equal distribution of peptides within the membrane, hence, all the lipids have a similar effect on the phase transition.\[[@CIT176]\] Furthermore, a complete translocation of the peptide may have occurred where the kinetics of peptide pore formation has been rapid, causing the peptide to reach an equilibrium on the outside and inside of the cell resulting in minimal changes to phase transition.\[[@CIT176]\] However, such possibilities can be determined by comparing the first and subsequent DSC scans. Differential scanning calorimetry was also used to study peptide-induced membrane curvature. MSI-78, an analog from the naturally derived magainin class of antimicrobial peptides was added to different PE species to monitor the L~α~ / H~II~ phase transition.\[[@CIT189]\] Initially for 1,2-dipalmitoleoyl-phosphatidylethanolamine (DiPoPE) the T~H~ transition occurred at 43°C. Small additions of peptide increased the transition with 0.4% peptide to 46.4°C.\[[@CIT189]\] Such experiments were confirmed using ^31^P-NMR, showing that the peptide induced a positive curvature strain of DiPoPE membranes, while the peptide prevented POPE from transforming from the lamellar state even at high concentrations.\[[@CIT189]\] Low peptide concentrations of MSI-78 resulted in morphology changes of the bilayer, consistent with pore formations, whereas, high concentrations typically increased the population of the lipids in the hexagonal phase.\[[@CIT189]\] Insight into peptide penetration can also be determined based on the degree of disruption to the lipid acyl chains. Furthermore, comparison to other antimicrobial systems has been used to establish the mechanism. Such evaluations have been made for magainin, showing less disruption to acyl chains of DPPG compared to melittin, suggesting less penetration into the hydrocarbon region.\[[@CIT190]\] Summary {#sec1-15} ======= Differential Scanning Calorimetry remains one of the primary tools for thermodynamic analysis. The rapid progression to modern nano and automated DSC instruments has shown the development and improvements in multiple applications. Advancements with software allow for easy interpretation of thermodynamic data, which again make DSC a very attractive technique. The instrument is comparatively inexpensive and the high sensitivity models only require a relatively dilute suspension in the aqueous phase. Even as most studies focus on protein conformation, DNA binding, lipid studies, and lipid-peptide or lipid-protein interaction, DSC has also been used in industry testing, health concern with plasticizers, and biomimetic lung membranes.\[[@CIT173]\] DSC is one of the most powerful techniques for the routine measurement of gel-to-liquid crystalline phase transition in the lipid bilayers and biological membranes, and changes in interactions of antimicrobial peptides are used to assess peptide-membrane interactions.\[[@CIT174]\] Applications to the pharmaceutical industry such as drug purity, stability, DNA drugs, lipid targets and different drug delivery models have improved our ability to study different compounds.\[[@CIT32]\] The ability to determine the physical and energetic properties of a compound has made DSC increasingly popular in drug development.\[[@CIT32]\] The authors would like to thank Patrick Lai for his assistance with figures. **Source of Support:** MHC was supported by a NSERC CGS-M and a CIHR team grant "novel alternatives to antibiotics" (EJP is one of the team PIs), **Conflict of Interest:** None declared.

PubMed Central

2024-06-05T04:04:19.967795

2011-01-01

PMC3053521

Drug-biomembrane Interaction Studies in Biomedical and Pharmaceutical Research {#sec1-1} ============================================================================== The human body possesses a variety of biomembranes, whose functions range from protecting tissues and cells from foreign molecules, and to select the cellular penetration of compounds with a biochemical or physiological role. After its administration, a drug molecule rapidly meets many of these biomembranes, from the circulating macrophage cells to the vessel endothelium, to the more complex blood-brain or blood-retinal barriers. Consequently, drug distribution in the body is largely affected in terms of time and concentration. Besides a structural role, biomembranes play other essential functions: control the passage of selected compounds, thus maintaining the biochemical integrity of cytosol; communication, allowing the exchange of information between the extra- and intracellular environments, and the physical interaction with the extracellular phase; biochemical-active surface, due to the amount of associated enzymes, receptors, ion channels, signaling molecules, and supramolecular structures that have a role in cellular homeostasis, metabolism, growth, and even death. Therefore, pathological alterations in the structure or functions of cell membranes and other biomembranes are often caught up in the etiology of many diseases. Studying the molecular events occurring on cell membranes, as well as the multiplicity of interactions with bioactive compounds, in either physiological or pathological situations, is therefore of paramount importance, to enlarge our knowledge of many diseases and to identify further potential therapeutic targets. Interactions of drugs and biological compounds with biomembranes are complex phenomena from the chemical or physicochemical point of view. They can represent a final stage, in case the biomembrane embodies a barrier to drug passage or is the site of action for the drug (e.g., at the level of the membrane receptors). In many cases, however, drug-membrane interaction represents only a preliminary step to the biological (or toxic) activity, as it can affect the rate of penetration and partitioning of the biomolecule in the cytoplasm, to reach a specific target cell organelle or system. In other words, both partitioning into and binding with cell membranes deserve to be accurately studied and characterized, for the old as well as the novel bioactive compounds.\[[@CIT1]\] It becomes evident that a drug-membrane interaction can be considered either as a partitioning phenomenon or a binding pathway. To generalize a situation that is instead often complex and multivalent, when the membrane acts as a barrier to the drug penetration into cells, that is, when the pharmacokinetic aspects are considered, the partitioning phenomena are more important. Conversely, when the cell membrane represents the site of action for the drug (pharmacodynamic), the drug binding processes deserve to be explored. These aspects become greatly relevant when the right model biomembranes are to be designed for *in vitro* studies. The forces underlying both kinds of interactions are the same, that is, polar and hydrophobic chemical interactions.\[[@CIT1]\] On this basis, although there is a great complexity of biochemical phenomena occurring in living cells, it is relatively simple to stress on such simplification, and design an efficient experimental model, suitable for investigating or even predicting the possible drug-membrane interactions. The results of such interactions can be reciprocal, in the sense that the biomolecule can alter the structure and function of the membrane, for instance change its permeability, charge potential, fluidity, and so on; but, on the other hand, the structure and properties of the drug can also be affected by its interaction with the membrane components, in terms of stereochemistry, molecular conformation, time of onset, and duration of the biological activity, for instance.\[[@CIT2][@CIT3]\] A passive, relatively easy diffusion through the lipid domains of the biomembranes has been considered since a long time, as being the main process regulating the permeability of drugs across membranes and the whole cell internalization process. The pharmacokinetics and pharmacodynamic patterns of drugs have been shaped according to such a statement. More recently, however, the role of membrane transporters has been highlighted. Large superfamilies of transporter proteins have been found in every living cell.\[[@CIT4]\] In a recent review, Dobson *et al*. have discussed the role of such proteins in the cell uptake of drugs, and of the modeling strategies that can be used to forecast drug pharmacokinetic features.\[[@CIT5]\] Nevertheless, it is still valid to deem that interaction with the cell surface is a prerequisite for drug activity, even when such an interaction is not followed by its internalization into the cytoplasm. The so-called 'non-specific' interaction of drugs with the biomembrane components involves a contact with the phospholipid (PL) structures, which can affect the highly organized lipid compartment and cause changes in the membrane protein conformation and work.\[[@CIT6]\] For analogous reasons, drug-biomembrane interactions can be the basis for adverse or toxic side effects,\[[@CIT7]\] such as phospholipidosis.\[[@CIT4][@CIT8]\] The presence of a foreign compound can damage membrane integrity: amphiphilic compounds, in particular, like many classes of drugs are, can exert a detergent-like activity on membranes, causing their disruption and dysfunction.\[[@CIT9][@CIT10]\] As natural cell membranes present a great complexity of structure, cross-connections and functionality, artificial model membrane systems, under intensive development, help the scientists to understand the effects of membrane lipids in drug transport and uptake into cells, drug activity, and even toxicity. In respect to living cell membranes, these models can be used under conditions that will not allow the cells to remain live and biochemically integral. This review will illustrate some of the analytical techniques applied in the study of interactions among bioactive compounds, with biological membrane models, with particular emphasis on calorimetric methods. This kind of data can be a powerful tool for medicinal chemistry and pharmaceutical technology, to design and optimize the activity and tolerability profiles of new drugs. Moreover, 'old' compounds also, whose clinical usefulness has been limited by their physicochemical properties, and thereby by pharmacokinetic problems, can be drag to a second life by optimizing precise details in their chemical structure. The Structure and Functions of Cell Membranes {#sec1-2} ============================================= Membranes of eukaryotic cells are made of three major components: lipids, proteins, and sugars. All membranes have a common general structure \[[Figure 1a](#F0001){ref-type="fig"}\], with structural or functional proteins (e.g., enzymes, receptors, channels, etc.) embedded within the two-layered sheets of lipid molecules. The lipid and protein molecules are held together mainly by non-covalent interactions, and alterations of their dynamics and strength are often associated with diseases. Sugars are attached by covalent bonds to some of the lipids and proteins. They are found on one side of the membrane only, for example, on the outer surface of the plasma membrane. ::: {#F0001 .fig} Figure 1 ::: {.caption} ###### The relations are shown among cell membranes (a), phospholipid bilayers (b), and liposomal vesicles (c) :::  ::: Biological membranes contain three major kinds of lipids: PL, glycolipids, and cholesterol. PL is generally formed by glycerol linked to two fatty acids, a phosphate group, and a basic head group; the fatty acid chains usually contain 14 to 24 carbon atoms. One chain can be unsaturated, containing from one to four *cis* double bonds. The three major glycerol-based phospholipids contain choline, serine, or ethanolamine attached to the phosphate; another type of phospholipid contains sphingosine instead of glycerol, such as sphingomyelin. About 40% of the lipids in eukaryotic cells are phosphatidylcholines (= lecithins), which are zwitterionic in a pH range from 4 to 10. Thus, they carry one negative and one positive charge in the physiological pH range. A common attribute of membrane lipids is their amphipathic nature. Both PL and glycolipids have a hydrophilic head and two hydrophobic tails: in an aqueous medium, these molecules spontaneously associate to form bilayers, with their hydrophobic tails sandwiched between the hydrophilic heads \[[Figure 1b](#F0001){ref-type="fig"}\]. The PL bilayers can be further simplified as a fluid phase containing three specific domains: a nonpolar hydrocarbon core, an interfacial region containing the uncharged PL acyl ester groups, showing an intermediate polarity, and the highly polar membrane surface that contains the charged PL head groups that is exposed to the aqueous exterior. These bilayers tend to close on themselves to form sealed compartments called liposomes, to eliminate the edges where the tails would be in contact with the water \[[Figure 1c](#F0001){ref-type="fig"}\]. A small drug molecule or an ion that migrates from the surface of a PL bilayer to the inner domains encounters a remarkable decrease in polar solvation and dielectric constant. Indeed, the low polarity of the hydrocarbon chain domain hinders the penetration of the charged or polar species, and most proteins that span a bilayer membrane have a sequence of nonpolar amino acids that match the thickness of the hydrocarbon region. Cholesterol contains a four-ring steroid structure together with a short hydrocarbon side-chain and a hydroxyl group. Cholesterol is found in some mammalian membranes, but not in most bacterial membranes, nor in plant membranes. As an amphipathic molecule, it can be incorporated into PL bilayers, but cannot form a bilayer on its own\[[Figure 2](#F0002){ref-type="fig"}\]. ::: {#F0002 .fig} Figure 2 ::: {.caption} ###### Schematic structure of a multilamellar liposome, showing the possible location of the host compounds: lipid soluble or amphiphilic compounds (gray ellipsoids) allocate completely or in part among the PL acyl chains and hydrophilic compounds (black spots) are retained in the aqueous spaces between the bilayers :::  ::: The cell membrane is currently not considered as the homogeneous and static bilayer, as described in the classical Singer and Nicolson's 'fluid mosaic model', but as a heterogeneous medium with a complex and dynamic lipid organization at a nanoscale level.\[[@CIT11]\] Simons and Ikonen proposed the concept of 'functional rafts' to describe the lateral domains rich in saturated lipids and cholesterol, dispersed within a phase rich in unsaturated lipids.\[[@CIT12]\] These structures are endowed with membrane proteins and can further be involved in chemical interactions with other functional proteins and sugars. Moreover, their size and structure may dynamically change under specific signals or stimuli (or also in some disease states) underlying the multiplicity of cellular responses.\[[@CIT13]\] Both the lipids and proteins are able to move within their own monolayer in the membrane. Lipid membranes can undergo transitions from a rigid to a liquid crystal lamellar state, where both the lateral order of the lipid molecules and the order of the lipid hydrocarbon chains change. In biological membranes this occurs at temperatures slightly below the body temperature. Both states form two-dimensional membranes, but their physical properties may be quite different. The main factors that determine the fluidity of cell membranes, apart from the temperature, are the length and the degree of unsaturation in the fatty acid tails of PL, the characteristics of their head groups, and the concentration of cholesterol in the membrane. This aspect retains a relevant role in the development and validation of biomembrane models. In contrast with the rapid lateral diffusion, lipid molecules rarely move from the monolayer that they are in to the opposite one, and often the lipid composition of the two layers is quite different. The transfer of a PL molecule from one layer to the other - known as transverse diffusion or *flip-flop* - is rare, because the polar, hydrophilic head would have to penetrate the non-polar, hydrophobic hydrocarbon core of the bilayer. The Elaboration of Artificial Biomembrane Models {#sec1-3} ================================================ Cell and biological membranes essentially consist of a lipid environment where liposoluble compounds can dissolve and pass through. Therefore, among the physicochemical features that characterize a small bioactive molecule in terms of interaction with living cells, solubility and partition coefficients are the most important.\[[@CIT14]\] As the interaction with membranes can represent only the first step of a cascade of chemical and physical processes, such as relations with protein receptors, enzymes or nucleotides, it is actually the balance between the hydrophilic and lipophilic characters of a molecule, that is, its *amphiphilicity* that determines a successful interaction with biomembranes. Many studies have shown that the simple parameter of lipophilicity cannot result in an overall improvement of cellular uptake of a drug or its passage through the barriers. The composite structure of cell membranes, as previously described, involves that drugs must possess an amphiphilic character to be able to modulate their movements inside both the lipid domains and polar spaces of the membranes.\[[@CIT15]\] The complex phenomena linked to partitioning into and binding of the drug to the cell membranes or barriers are better related to the so-called 'anisotropic lipophilicity'.\[[@CIT1]\] It not only derives from the hydrophobicity of drugs, but also from its ability to make polar and ionic bonds with the membranes. The traditional way to express the lipophilicity / amphiphilicity of a compound has been the determination of its partition coefficient (logP) between two immiscible liquid phases, typically, a water or a buffered aqueous solution and an organic solvent.\[[@CIT16]--[@CIT18]\] The cross-evaluation of the logP value in a four-solvent series allowed to draw a more complete lipophilicity/hydrophilicity profile of the test compounds.\[[@CIT19]\] All these 2-D systems share the relative simplicity of use and remain of great value when, for instance, a chemically related series of analogs must be compared or quantitative structure-activity relationship (QSAR) studies on large sets of compounds must be developed. However, when the aim is the prediction of the interaction with biomembranes, solvent--solvent partition experiments suffer the absence of a tridimensional scaffold, similar to that involved in side-interactions between living cell membrane components and a drug. Hydrogen bonds or van der Waals interactions are, for instance, difficult to reproduce or simulate in an isotropic 2-D liquid system. Moreover, the partitioning of chargeable or charged compounds is almost impossible to measure and reproduce.\[[@CIT20]\] In the last few years, many alternative models have thus been developed, in which a 3-D structure is built to allow as many interactions as possible with the test compound. The well-defined composition and structure of these complex assemblies can allow to better reproduce the phenomena that occur *in vivo* during drug transport, distribution, biological activity, and even resistance. Brilliant examples of experimental comparison between *n*-octanol-water logP data and 3-D membrane models can be found in the fundamental work of Seydel and Wiese (2002).\[[@CIT6]\] Another excellent review by Peetla *et al*.\[[@CIT7]\] has recently examined the available biomembrane models and discussed their utility in studying the interaction with different classes of drugs, as well as in predicting their efficacy or toxicity. Artificial membrane models allow the carrying out of particular studies in experimental conditions, such as temperature or osmolarity, at which the living cells cannot be driven without damaging their functionality. In these models some cell membrane properties are obviously lost, such as the presence of functionalized membrane proteins, receptors, endocytosis, and other active processes; thereby, the biophysical interactions of drugs with artificial membranes may not exactly reproduce all aspects of the biological environment. However, a good reproducibility of results and correlation with *in vitro* as also *in vivo* pharmacological or toxicological behavior has been observed in many investigations. Generally speaking, three main kinds of lipid membrane models have been identified: monolayers, vesicle-forming bilayers (liposomes), and supported bilayers. An interrogation of the Pubmed database on the articles published in the last 10 years using such biomembrane models gave approximately 500 hits, many of which made use of calorimetric techniques \[[Table 1](#T0001){ref-type="table"}\]. ::: {#T0001 .table-wrap} Table 1 ::: {.caption} ###### Literature overview of recent investigations using DSC and other techniques to study the interaction of drugs with various biomembrane models ::: Biomembrane model Analytical tool(s) Host drug/Compound Refs. --------------------------------------------------------------- ------------------------------------ ------------------------------------------------ ------------------------------------------------------------------------- Phospholipid vesicles (SUV, LUV, MLV) and Micelles DSC, LB, BAM, Fluo Melatonin De Lima *et al*., J. Pineal Res. 2010;49,169 DSC E-3,5,4'-trimethoxystilbene/β-CyD Sarpietro *et al*., Int J Pharm. 2010;388,144. DSC Labaditin Barbosa *et al*., Amino Acids 2010 DSC, BAM, PM-IRRAS Plasticins Joanne el al., Biochemistry 2009;48,9372 DSC, Fluo, DLS BC5 (N-pentadecylpiperidin-4-amine) Luciani *et al*., Mol. Biosyst. 2009;5,356 DSC, XRD NSAIDs Lucio *et al*., Langmuir. 2008;24,4132 ITC Cathelicidin antimicrobial peptides Andrushchenko *et al*., BBA 2008 1778;1004 DSC Xenobiotics Zepik *et al*, Crit Rev Toxicol. 2008;38,1 DSC Thymopentin prodrugs Pignatello & Pecora, Pharmazie 2007;62,663 DSC (-)-Epicatechin conjugates Lazaro *et al*., J Agric Food Chem. 2007;55,2901 DSC, XRD, FF Sulfadiazin Oszlánczi *et al*., Biophys Chem. 2007;125,334 DSC N-oxides of tertiary amines Kleszczynska *et al*., 1 Naturforsch C. 2005;60,567 DSC Gemcitabine Castelli *et al*., J Colloid Interface Sci. 2005;285,110 DSC β-interferon derivatives Larios *et al*., Biophys Chem. 2004;111/123 DSC Tocopherols and phenolic compounds Gutiérrez *et al*, Life Sci. 2003;72,2337 DSC rBPI(21) Domingues *et al*., PLoS One. 2009;4,e8385. DSC, Fluo Acyclovir and squalenoyl-acyclovtr Sarpietro etal., Int J Pharm. 2009;382,73 DSC Plantaricin 149 synthetic peptides Lopes *et al*., BBA 2009;1788,2252 DSC D-cycloserine Musumeci *et al*., J Liposome Res. 2008;18,211 DSC, TEM Bilirubin-IXalpha enantiomers Ceccacci *et al*., Bioorg Chem. 2008;36,252 DSC Chlorpromazine Zhang *et al*., Biochem. Biophys ResCommun. 2008;373,202 DSC Gemcitabine and prodrugs Castelli *et al*., J Colloid Interface Sci. 2007;3,1643 DSC Lipoamino acids (LAA) Pignatello *et al*., CurrDrug Deliv. 2007;4(109 DSC resveratrol and derivatives Sarpietro *et al*., J Agric Food Chem. 2007;55,3720 DSC β-Sitosterol/β-CyD Castelli *et al*., J Agric Food Chem. 2006;54,10228 DSC Oxicams Lúcio *et al*., Med Chem. 2006,2,447 DSC Docetaxel loaded-PLA/PLGA Musumeci *et al*., Int J Pharm. 2006;325,172 DSC Herbicides Librando *et al*., Environ Sci Technol. 2006,40,2462 DSC Idebenone amphiphilic prodrugs Pignatello *et al*., J Colloid Interface Sci. 2006,299,626 DSC Tranylcypromine amphiphilic conjugates Pignatello *et al*., Int J Pharm. 2006,310,53 Fluo Tocopherols and tocotrienols Sonnen *et al*., J Am Chem Soc.2005,127,15575 DSC Drug release from polymer conjugates Castelli *et al*., Drug Deliv. 2005,12,357 HPLC Biphenyl derivative Ceccacci *et al*., J Am Chem Soc. 2005,127,13762 DSC Arginine-based cationic surfactants Castillo *et al*., Langmuir.2004;20,3379 UV-Vis, Fluo Quinolones antibiotics Neves *et al*., Biophys Chem. 2005,113,123 DSC Hepatitis G virus envelope protein peptides Larios *et al*., Langmuir 2004;20,11149 NMR, ROESY Ditryptophan, diphenylalanine Bombelli *et al*., J Am Chem Soc. 2004,126,13354 DSC Micronized nimesulide Castelli *et al*., Eur J Pharm Sci. 2003;19,237 DSC, mol. modeling Ofloxacin Fresta *et al*., Bioorg Med Chem. 2002;10,3871 DSC Papaverine in CyDs Ventura *et al*., J Drug Target.2001;9,379 Fluo (-t-)-Totarol Mateo *et al*., BBA 2000;1509,167 Spectroscopic and other techniques HAV-VP3 peptides Sospedra *et al*., Biopolymers 2000,54,477 CFM, DSC Chitosan microspheres loaded with moxifloxacin Ventura *et al*, Eur J Pharm Biopharm. 2008;68,235 DSC Inulin-based hydrogel Castelli *et al*., Eur J Pharm Sci.2008;2,76 Monolayers LB Chromium(III) complexes Sella *et al*., BBA, 2010 LB Local anesthetics, alcohols Frangopol *et al*, Colloids Surf B Biointerfaces 2001;22:3. LB Acyclovir and prodrugs Sarpietro *et al*., Int J Pharm. 2010 LB Frutalin lectin Nobre *et al*., BBA 2010;1798;1547 LB Plantaricin 149 peptide analog Lopes *et al*., BBA 2009;1788,2252 LB Gemcitabine and squalene prodrug Castelli *et al*., J Colloid Interface Sci.2007;3,1643 LB Gemcitabine prodrugs Castelli *et al*., J Colloid Interface Sci. 2007; 313,363 DLS cationic liposome-DNA complexes Uchiyama *et al*., Anal Sci.2004,20,1537 Supported bilayers AFM Melatonin De Lima *et al*., J. Pineal Res. 2010;49,169 Real-time AFM Triton X-100 Morandatand El Kirat, Langmuir 2006;22,5786 HCM, Fluo Antimicrobial peptides Davis *et al*., J. Pept Sci. 2009;15,511 AFM Oritavancin Domenech *et al*., BBA 20091788,1832 AFM Protegrin-1 Lame *et al*., J. Phys. Chem. B. 2006;110,21282 Phospholipidcoated columns (Immobilized artificial membranes) HPLC (-)-Epicatechin conjugates Lazaro *et al* J Agric Food Chem. 2007;55,2901 HPLC -- Zhang *et al*., J. Sep. Sci., 2010, in press HPLC Various model drugs Zhang *et al*., Talanta. 2005;67,1023 HPLC Flavonoids Ollila *et al*., Arch Biochem Biophys. 2002;399,103 Molecular chormatography Statins Sarr t al., J Chromatogr B Analyt Technol Biomed Life Sci. 2008; 868,20 HPLC Various model drugs Barbato *et al*., Eur J Pharm Sci. 2004;22,261 Immobilized phospholipid capillary electrophoresis HPLC NSAIDs Mei *et al*., Talanta 2008;75,104 Abbreviations: AFM: atomic force microscopy; BAM: Brewster angle microscopy; CD: circular dichroism; CFM: Confocal fluorescence microscopy; CyD: cyclodextrin; DLS: dynamic (quasi-elastic) light scattering; DSC: differential scanning calorimetry; FF: freeze-fracture; Fluo: fluorescence spectroscopy methods; HCM: hyperspectral confocal microscopy; ITC: Isothermal titration calorimetry; LB: Langmuir-Blodgett monolayer films; LUV: large unilamellar vesicles; MLV: multilamellar liposomes; NMR: nuclear magnetic resonance spectroscopy; PM-IRRAS: Polarization modulation infrared reflection absorption spectroscopy; ROESY: rotational nuclear overhauser effect spectroscopy; SEM: scanning electron microscopy; SUV: small unilamellar liposomes; TEM: transmission electron microscopy; TR-QELS: time-resolved quasi-elastic laser scattering; XRD: X-ray diffraction methods. ::: As amphiphilic molecules, PL form monomolecular insoluble films (called *Langmuir monolayers*) on the surface of a liquid subphase. Such films are excellent model systems to study membrane biophysics, because a biological membrane can be considered as two weakly coupled monolayers.\[[@CIT21][@CIT22]\] They can be useful models for analyzing the structure and the mixing interactions of drugs with PL monolayers, at the air-water interface.\[[@CIT15][@CIT23]--[@CIT27]\] Moreover, their extemporary method of production allows to use lipids with the necessary physicochemical properties. To study the interaction of drugs with PL, a lipid / drug mixture is spread over water or an aqueous buffered phase to form a monolayer in a Langmuir trough. The monolayer is then compressed, and the surface pressure-area isotherms are measured and compared to the one of pure lipids.\[[@CIT22]\] Two main thermodynamic parameters, namely surface pressure and temperature, can be easily controlled in this application. Apart from being valid drug nanocarriers, *liposomes* have become interesting models of biomembranes to study the interaction with drugs and to make partitioning experiments.\[[@CIT27]--[@CIT30]\] Liposomes are self-assembling systems usually composed of PL molecules. Bilayers made of native PL, extracted from different kinds of cell membranes, have also been described.\[[@CIT2]\] Hydration of these lipids with an aqueous medium spontaneously produce concentric multilamellar vesicles (MLV). The mechanical handling of the latter by sonication or extrusion through membrane filters, yields unilamellar liposomes of different size (SUV, LUV). Partitioning experiments can be performed using either MLV or SUV suspensions, in which a drug or another foreign compound can be incorporated during vesicle production, or, as revealed in the review of Castelli *et al*. in this issue, allowed to diffuse inside preformed, void vesicles. The amphiphilic nature of PL produces liposomes with an ordered bilayer arrangement, with the acyl chains aligned to form a lipid domain and the polar head groups (e.g., choline, serine) facing the external or interlamellar aqueous spaces \[[Figure 3](#F0003){ref-type="fig"}\]. According to its chemical nature, an extraneous compound can thus find room within the acyl chains (as hydrophobic and aromatic molecules do), in the aqueous spaces (for hydrophilic or polar compounds), or even between them, if an amphiphilic compound is hosted \[[Figure 4](#F0004){ref-type="fig"}\]. Artificial bilayers show many physical features very close or comparable to natural membranes, such as, thickness, electrical resistance, interfacial tension, and so on.\[[@CIT31]\] ::: {#F0003 .fig} Figure 3 ::: {.caption} ###### Outline of the main analytical techniques applied for studying the interaction between drug or drug carriers and biomembrane models :::  ::: ::: {#F0004 .fig} Figure 4 ::: {.caption} ###### Typical DSC curve of phospholipid bilayers undergoing gel-to-liquid crystal phase transition under controlled heating :::  ::: In spite of specific problems linked to the solubility of the host compound or to the compatibility or degradation issues, liposomes generally represent a valid means to study the partition of biologically related substances. The main advantage of liposome models is their anisotropicity, which allows the occurrence of tridimensional phenomena between the host molecule and the PL bilayers. Thus, steric hindrance and spatial-related troubles, which can affect the *in vivo* behavior of a drug, can be definitely observed and exploited. Lúcio *et al*.,\[[@CIT32]\] for example, has shown that as the evaluation of the antioxidant properties, correlated with the activity of non-steroidal anti-inflammatory drugs, receives an additional value using a 3-D liposome model, the effects of the drugs on the lipoperoxidation of membrane components can be analyzed as a function of drug-membrane interactions. In other words, while *partitioning* is the only phenomenon that can be reproduced in solvent--solvent models, with 3-D anisotropic models also, the physicochemical and chemical processes can be generated, to simulate at least in part the *binding* phenomena that occur in living systems.\[[@CIT1]\] A further, very intriguing issue is that of the electrical charge of membranes and / or host molecules. Although on one hand the differences in the partition profiles of neutral and protonated species, often noticed with the octanol-water system, are basically reduced with liposomes\[[@CIT6]\] so that experiments can be carried out within a large range of physiologically-related pH values; on the other hand, cell membranes show a net surface charge, which can influence electrostatic interactions with drugs and biomolecules. Such types of interactions might not be predictable with the bidimensional solvent / solvent partition experiments, but they can be measured or predicted using 3-D systems made of suitably charged PL or lipids.\[[@CIT33]\] *Supported lipid bilayers* can be formed on atomically smooth solid supports, such as silicon, by the Langmuir-Blodgett / Langmuir-Schäfer (LB / LS) technique,\[[@CIT34]\] by the vesicle fusion (VF) technique\[[@CIT35]\] or by a mixed Langmuir-Blodgett / vesicle fusion (LB / VF) technique.\[[@CIT36]\] These systems are also valid models for reproducing the thermodynamics of cell membranes, even when different liquid lipid phases coexist.\[[@CIT37]\] On contact and interaction with the test drugs, changes in morphology, structure, and chemistry of these supported bilayers can be monitored by coupled analytical techniques, such as Fourier transform infrared spectroscopy (FTIR), scanning electron microscopy (SEM), transmission electron microscopy (TEM), X-ray diffractometry, atomic force microscopy (AFM), and so on.\[[@CIT38]\] The use of these various 3-D membrane models has contributed to the unraveling of many limits associated with isotropic solvent-solvent systems, such as the impact of an electrical charge or the stereochemistry and spatial orientation of bulky parts of the molecules, on their interactions with the membranes. However, all these models still present severe limitations when a direct transfer of the experimental information to living biological systems or when a prediction of the *in vivo* behavior of a drug is endeavored. Analytical Techniques Available to Study Drug-Membrane Model Interactions {#sec1-4} ========================================================================= This paragraph will give some quick references to the principal methodologies that have been explored or proposed in the last few years, for this special type of study. More detailed information on how each technique works and can be used, may be found in some of the cited references. Many analytical methods have been investigated with the aim of determining the affinity of compounds for biomembranes \[[Figure 3](#F0003){ref-type="fig"}\]. Chromatography (HPLC), spectroscopy (FT-IR, UV, ESR, EPR, NMR, fluorescence depolarization, circular dichroism, SIMS), and other biophysical techniques (X-ray and neutron diffraction methods, differential scanning calorimetry (DSC), isothermal titration calorimetry, surface pressure changes, potentiometry, etc.), often in combination with each other. Of late in-depth investigations have been allowed, to find out the effects of drug molecules on the structural and functional aspects of cell membranes. Using radiolabeled salts (displacement of ^45^Ca ^++^) is also a relatively simple tool to carry out drug-membrane affinity experiments. In recent years, different computational modeling approaches have also been directed to these studies.\[[@CIT39]\] Some of these applications have been recently reviewed by Seddon *et al*.\[[@CIT40]\] and more information can be found in the textbook of Seydel and Wiese.\[[@CIT6]\] We basically agree with other authors in their consideration that to characterize the interactions of drugs with biomembrane models well, from a qualitative and quantitative point of view, no single technique can give all the required data, however, the association of complementary experiments could provide utmost benefits. Among the chromatographic methods set up for this particular application, Immobilized Artificial Membrane (IAM) chromatography allows to efficiently simulate liposome / water partitioning and cell membrane permeation.\[[@CIT41]--[@CIT44]\] IAM stationary phases are solid-phase systems, where a phospholipid monolayer is covalently bonded to a propylaminosilica support material. With IAM columns, simple aqueous mobile phases (e.g., PBS) can be used, without addition of organic modifiers. With respect to liposomal vesicles, the hydrophobic domain is half large in the IAM surface and the PL is more ordered and less mobile. Solute partitioning between the eluent and stationary phases seems to be the mean retention mechanism in IAM retention.\[[@CIT41]\] However, polar interactions are often involved, depending on the structural properties of the analytes. Thus, protonated basic compounds are stronger retained, because of their interaction with the phosphate anionic groups of the stationary phase.\[[@CIT45]\] The potential of IAM chromatography is to predict passive transport through various biological barriers, as well as to estimate drug pharmacokinetic and pharmacodynamic properties, which has been recently reviewed.\[[@CIT46]--[@CIT48]\] As for liposomes, the results of IAM chromatography give mixed information on passive diffusion (permeation) of drugs and drug-membrane interactions (binding), although the contribution of electrostatic forces and hydrogen bonds has been reported to be weaker in IAM chromatography than in liposome partitioning\[[@CIT49][@CIT50]\] Correlation with *n*-octanol-water lipophilicity\[[@CIT51]\] and the quantitative structure-retention relationship with IAM chromatography, using classical physicochemical molecular descriptors, has been elucidated.\[[@CIT52]\] However, it would seem that the results obtained with these techniques can be efficaciously compared to the interaction profile with biomembranes only for strictly structure-related classes of molecules.\[[@CIT53]\] In more recent years, other types of biomimetic stationary phases have been developed, like α~1~-acid glycoprotein or albumin derivatized surfaces, widening the applications of biochromatography as a reliable tool to study drug interaction with cells and biosystems.\[[@CIT54]\] A further advance in this field is represented by biopartitioning chromatography (BPC), a technique in which a chromatographic method is combined with biomembrane-mimetic structures like PL vesicles or monolayers, polymer micelles, microemulsions, niosomes, and so on. In the last few years BPC has become a high-throughput screening platform to study drug-membrane interaction and permeability and their correlation with the biological effects.\[[@CIT55]\] Also, the capillary electrophoresis technique (EC) has been tailored to create a valid model for drug-membrane interaction studies.\[[@CIT56]\] EC experiments have been used to characterize the size, surface properties, encapsulation volumes, and the electrophoretic mobility of colloidal lipid vesicles and of lipoprotein particles. Interactions between biologically-related compounds and lipid vesicles that serve as pseudostationary phases, or as coated stationary phases, in electrokinetic chromatography, can be used to investigate the biophysical nature of drug-membrane model interactions. Among the spectroscopic techniques, NMR analysis, and especially the new, more sensitive techniques, such as Nuclear Overhauser Effect (NOE) or transfer NOE, can add more detailed information about the interactions occurring at a molecular level between a drug and lipid and PL molecules.\[[@CIT57]\] These experiments, in some cases, allow to identify the molecular specificities responsible for the interaction with and / or permeation of drugs through the membrane bilayers.\[[@CIT58]\] Solid-state NMR (SS-NMR) is also a valid technique to explore, in depth, the effects of the host compounds on the package, and inter-connections among biomembrane components. Thus, ^2^H, ^13^C, ^31^P, and ^15^N SS-NMR, or a combination of the various techniques, allow to determinate the degree and level of interaction of model compounds with the PL bilayers, as well as the effects on the morphology and physicochemical property of the resulting membranes. In many cases, it becomes possible to discriminate between the effect exerted on the polar surface or on the hydrophobic alkyl chain domain of the bilayers, monitoring the interaction with foreign molecules, for example, as a function of PL composition and charge, temperature, and pH.\[[@CIT59][@CIT60]\] A fluorescence study, using the fluorescent probe 1,6-diphenyl-1, 3, 5-hexatriene (DPH), has been applied to test the effects of NSAIDs on different membrane systems, including liposomes, mouse macrophages, a human leukemia monocyte cell line, granulocytes, and mononuclear cells. The tested NSAIDs were able to efficiently quench the probe located in the membrane hydrocarbon region and to enhance the membrane fluidity, proving their interaction with the membrane lipids. Authors suggested that the induced changes in lipid dynamics could affect the activity of inflammatory enzymes or could be related to the local side effects of NSAIDs on the stomach mucosa.\[[@CIT61]\] Fluorescence assays have been also used to follow the binding and transport of fatty acids (FA) in model and biological membranes. Partitioning of FA between membranes and the aqueous buffer, insertion of the FA acyl chain into the hydrophobic core of the phospholipid bilayer, and the presence of the FA carboxyl group at the outer leaflet of the membrane have been studied using different probes.\[[@CIT62]\] A luminescence assay, which is based on the energy transfer of a permeant to liposomal terbium (III) has also been used to examine drug permeation in the membrane bilayers.\[[@CIT63]\] The experimental results on model acidic molecules led to the important conclusion that depending on single variables like membrane composition and rigidity or electrostatic interactions, and on the geometry of the model system, lipid bilayer permeation may positively, negatively or not correlate with the bilayer affinity of the tested molecule. Small angle X-ray diffraction and small angle neutron scattering techniques have been utilized to identify the position of drugs in model and native multi-bilayer vesicles.\[[@CIT2]\] An improved potentiometric evaluation of the lipid membrane - water partition coefficient of ionizable drugs has been recently described, in which the data analysis was corrected on the basis of Coulomb electrostatic phenomena.\[[@CIT64]\] Voltammetric methods have been also applied to analyze the ionic transfer kinetics of ionizable drugs across the lipid-modified liquid-liquid interfaces.\[[@CIT65]\] This technology promises interesting developments in the field of high-throughput assessment of the ADMET properties of drugs. Calorimetric Techniques {#sec1-5} ======================= Calorimetric approaches are among the most used techniques to study drug-biomembrane interactions. DSC is a diffuse technique for such studies.\[[@CIT6][@CIT66]--[@CIT68]\] In DSC experiments, the thermotropic changes, eventually occurring in a liposome sample, in the presence of a drug molecule, are registered and quantified. The concept statement of DSC is that PL vesicles undergo a reversible phase transition, under the effect of increasing temperature, from a 'gel' state, in which the acyl chains are orderly packed within the bilayers, to a 'liquid crystal' state, associated with an increase of spatial disorder in the bilayers \[[Figure 4](#F0004){ref-type="fig"}\]. Such a transition is joined with heat uptake, therefore, it is an endothermic process. A detailed description of the technique can be found in many reviews (e.g.,\[[@CIT68]--[@CIT71]\]) and is illustrated in other articles published in the present issue. Extensive DSC studies from some of us and from Castelli and coworkers have been focused on the effects on DPPC or DMPC (dimyristoylphosphatidylcholine) uni- and multi-lamellar liposomes or different series of analogs.\[[@CIT26][@CIT72]--[@CIT78]\] The DSC analysis of the interactions of xenobiotics, such as drugs, with liposomal models may also correlate with their biological and toxicological behavior.\[[@CIT1][@CIT79]--[@CIT84]\] Also the toxicological pattern of drugs can be marked out;\[[@CIT85]\] for instance, Jelokhani-Niaraki *et al*. used isothermal titration calorimetry to assess the lytic effects on model PL membranes, for the conformational changes occurring in gramicidin S analogs.\[[@CIT86]\] When the fatty acid chains of PL are in their (solid) crystalline state, they are in an extended conformation and in *all trans* configuration. They form a regular crystal lattice, stabilized by low energy forces and electrostatic interactions between the polar head groups. During the phase transition to liquid crystal, the melting of the acyl chains causes a passage to the gauche configuration, leading to an increase of chain mobility. Addition of an extraneous compound, based on its chemical nature and thus its location within the bilayers, is able to induce a defined change in the thermotropic parameters of the liposomes, with respect to pure PL vesicles. For instance, a reduction in the temperature of the main phase transition (Tm) can be essentially ascribed to an interference with the polar (choline) head groups of the PLs; similarly, variations in calorimetric enthalpy changes (ΔH) associated to this phase transition rather denote the degree of interaction with PL acyl chains. In [Figure 5](#F0005){ref-type="fig"} a typical example is given, where increasing the molar fractions of a host compound can either induce a progressive lowering of the Tm of pure dimyristoylphosphatidylcholine (DMPC) MLV liposomes or not, along with a reduction in the associated ΔH changes, according to the degree of interaction with the bilayers. ::: {#F0005 .fig} Figure 5 ::: {.caption} ###### DSC curves of MLV liposomes made of pure DMPC or containing increasing molar fractions (Xo = 0-0.15) of a model host compound. (a), a compound exhibiting a reduced interaction with the PL bilayers (e.g., a polar molecule) induced limited thermotropic effects; (b) strong deformation of the DSC curves of PL bilayers under the influence of a lipophilic compound (F. Castelli, *personal data, with permission*) :::  ::: Other parameters can help to understand the occurring interaction, for instance, the shape of the endothermic transition peak, expressed as the width at half the height of the peak itself (T~½~), which gives a clear indication of the cooperativity of the system in the presence of a foreign compound, that is, the uniformity of distribution of the latter within the bilayers, from the surface through the core of liposomes.\[[@CIT87].[@CIT88]\] Also the transition interval, that is, the temperature range between the onset and the end of the phase transition curve, is indicative of the homogeneity of the host compound-liposome system and its cooperativity. A detailed explanation of the fundamentals of this technique can be found in this same issue, in the review of Chiu and Prenner. Conclusions {#sec1-6} =========== As this review belongs to an issue focused on special applications of calorimetric techniques, and DSC in particular, we have pointed out some examples derived from our study and studies of other authors, confirming the utility of such experimental approaches. Analysis of the interactions that can occur between a drug or a biologically active compound, and biomembranes are becoming a settled part of the design, discovery, and characterization of new drugs, and progressively at an earlier stage of development. Many different analytical techniques have been applied or developed on purpose to perform these kinds of studies. Using artificial membranes as simplified models for cell membranes has given a strong input to the understanding of the complex set of interactions that a biomolecule can develop toward biological membranes, and often also vice versa. However, the addition of an external compound to a PL bilayer or monolayer can induce physicochemical changes on its own, which will confuse the interpretation of the experimental data. Consequently, it is a lot more evident that reliable and constructive data may be acquired only on the basis of two elements: (i) the availability of new and more sophisticated models for cell and biological membranes, although simple enough to be used and reproduced; and (ii) the concurrent application of different analytical techniques, whose specific contribution will help to give an overall consciousness of these interactions. **Source of Support:** Nil **Conflict of Interest:** None declared.

PubMed Central

2024-06-05T04:04:19.978006

2011-01-01

PMC3053522

In recent years, a large number of studies have been directed to the evaluation of the action of antimicrobial peptide (AMP) on target pathogens, due to the huge increase in pathogenic microorganisms multiresistant to conventional antibiotics.\[[@CIT1]--[@CIT6]\] In this scenario, substantial efforts are being devoted to the discovery of new antibiotic resources and strategies, and antimicrobial peptides (AMPs) are receiving considerable attention as a new paradigm in antibiotic therapy. There is a general agreement that most peptides have the bacterial membrane as the main target of their antimicrobial action, which may proceed through different mechanisms, such as formation of stable pores (either barrel-stave or toroidal pore's type), membrane thinning (molecular electroporation or sinking rafts models) or micellization of the membrane (in a detergent-like action (carpet model).\[[@CIT7]--[@CIT13]\] Although most AMPs have a wide range of activity, subtle differences are found with regard to different spectra of activity, which must derive from a combination of factors such as size, amino acid sequence, charge, secondary structure upon membrane interaction, hydrophobicity, hydrophobic moment, and amphipathicity. These parameters are interdependent and a change in one of them can alter the structure-activity relationship, thereby, influencing the ability of the peptide to interact with the pathogen. Bovine Lactoferrin is a glycoprotein, composed of a polypeptide chain containing 703 amino acids folded into two globular lobes, called the C-- (carboxy) and N-- (amino) terminal regions, connected with an α-helix. Each lobe consists of two domains: C1 and C2, and N1 and N2. Most of the antimicrobial activity is attributed to the N1 domain, and the first group of peptides studied derived from this domain of Lactoferrin, comprising of amino acids 17 -- 41, and designated as Lactoferricin B (LFcin B).\[[@CIT14]--[@CIT18]\] Of late, another amino acid sequence, also located in the N1 domain was identified and studied in the Bolscher's group,\[[@CIT19][@CIT20]\] by searching for key determinants for antimicrobial activity, such as, the presence of stretches with alternating positively charged and uncharged residues with the potential to form a positively charged amphipathic α-helix.\[[@CIT19]\] The initial sequence contained amino acids 268--284, and thereafter, several peptides were synthesized, and obtained by truncation / extension of the original sequence.\[[@CIT21]\] Among these, we chose three peptides for the present study, namely LFampin 265 -- 284, LFampin 265 -- 280, and LFampin 270 -- 284. Taking LFampin 265 -- 284 as the lead peptide, the two others were obtained by truncation on the C side (LFampin 265 -- 280) and on the N side (LFampin 270 -- 284), as these two shortenings mainly affected two different parameters --- cutting on the C side reduced the charge, and on the N side the tendency to adopt a helical structure, as shown by van der Kraan *et al*,\[[@CIT21]\] In this manner, we attempted to discriminate the properties that were more important for antimicrobial potency and how the differences were reflected in the peptide's interaction with the membrane. Some properties of these peptides are summarized in [Table 1](#T0001){ref-type="table"}. ::: {#T0001 .table-wrap} Table 1 ::: {.caption} ###### Properties of synthetic LFampin peptides ::: Peptide Sequence \#AA[\*](#T000F1){ref-type="table-fn"} *M~r~*[\*\*](#T000F2){ref-type="table-fn"} Charge[a](#T000F3){ref-type="table-fn"} *\<μ\>*~b~[b](#T000F4){ref-type="table-fn"} *\<H\>*~c~[c](#T000F5){ref-type="table-fn"} -------------------- ---------------------- ---------------------------------------- -------------------------------------------- ----------------------------------------- --------------------------------------------- --------------------------------------------- LFampin 265 -- 284 DLIWKLLSKAQEKFGKNKSR 20 2389 4 + 0.30 \- 0.337 LFampin 265 -- 280 DLIWKLLSKAQEKFGK 16 1904 2 + 0.37 \- 0.186 LFampin 270 -- 284 LLSKAQEKFGKNKSR 15 1733 4 + 0.28 \- 0.437 \* \#AA= number of amino acids in the amino acids sequence. \*\* M~r~ = molar mass (kDa). a Net positive charge at neutral pH b *\<μ\>;* mean hydrophobic moment in a α-helical conformation.\[[@CIT41]\]. c \<*H*\>; mean hydrophobicity\[[@CIT49]\] ::: Materials and Methods {#sec1-1} ===================== Microorganisms and culture conditions {#sec2-1} ------------------------------------- Two strains --- *Streptococcus sanguinis* SK4 and *Escherichia coli* K12 --- were cultured aerobically at 37°C in brain heart infusion (BHI) medium from Difco (Becton Dickinson Microbiology). Yeast *Candida albicans* 315 was cultured aerobically at 30°C in Sabouraud dextrose broth source. The microorganisms were cultured overnight and subcultured for two-to-three hours to yield a mid-logarithmic growth culture at the time of harvesting. Synthesis and purification of peptides {#sec2-2} -------------------------------------- Bovine lactoferrin peptides \[[Table 1](#T0001){ref-type="table"}\] from the LFampin domain were synthesized with a Milli-Gen 9050 peptide synthesizer (MilliGen/Biosearch, Bedford, MA) according to the manufacturer's procedures. Peptides were purified to a purity of at least 95% with semi-preparative RP-HPLC (Jasco, Tokyo, Japan) on a Vydac C18-column (218MS510, Vydac, Hesperia, CA). The authenticity of the peptides was confirmed by ion trap mass spectrometry with an LCQ Deca XP (Thermo Finigan, San Jose, CA) as described previously.\[[@CIT21]\] Antimicrobial activity {#sec2-3} ---------------------- Bactericidal and candidacidal activity of the peptides was determined by peptide-mediated membrane permeabilization, monitored by the fluorescence enhancement of propidium iodide (PI, Invitrogen, Breda, The Netherlands) in dead cells, as described previously.\[[@CIT22]\] Briefly, a mid-log phase culture of bacterial suspensions (approximately 2.5×10^8^ CFU/mL) or C. *albicans* suspension (approximately 1.5×10^7^ CFU/mL) in 96-well U-bottom low affinity plates (Greiner Bio One) were supplemented with PI (final concentration 6 mM) and incubated with equal volumes of peptide solutions at final concentrations of 0.2-50 mM, at 37°C. Fluorescence was monitored at λ~exc~ 544 nm and λ~em~ 620 nm using a fluorescence reader (Fluostar Galaxy, BMG Labtechnologies, Offenburg, Germany) with five minute time intervals till 15 minutes followed by 15 minute intervals till one hour. LC~50~ values (mM) were the concentrations of the peptides resulting in 50% killing. All experiments were repeated at least twice in duplicate. Preparation of liposomes {#sec2-4} ------------------------ Appropriate amounts of 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC), 1,2-dimyristoyl-*sn*-glycero-3-\[phospho-*rac*-(1-glycerol)\] (DMPG) (Avanti Polar Lipids, Alabama, USA), and a DMPC:DMPG mixture at a molar ratio of 3 : 1 were dissolved in chloroform / methanol (3 : 1 v/v). The solution was dried under a slow nitrogen flow and the resulting lipid films were kept under vacuum for three hours to remove all traces of organic solvents. The lipid film was hydrated with 10 mM HEPES buffer (pH 7.4) containing 100 mM NaCl, at 10°C above the temperature of the *gel*, to *liquid crystalline* phase transition (T~m~). The resulting multilamellar vesicles (MLVs) were frozen in liquid nitrogen and thawed in a water bath at approximately 10°C above T~m~ (five cycles). Large unilamellar vesicles (LUVs) were obtained from the MLVs by extrusion in a 10 mL stainless steel extruder (Lipex Biomembranes Inc., Vancouver, Canada) and thermostated at 10°C above T~m~ . The samples were passed several times through polycarbonate filters (Whatman, Nucleopore, NJ, USA) of decreasing pore size (600, 200, and 100 nm; 5, 5, and 10 times, respectively), under inert (N~2~) atmosphere. The phospholipid concentration was determined for each preparation by the phosphomolybdate method.\[[@CIT23]\] ### Circular dichroism {#sec3-1} Circular Dichroism (CD) experiments were carried out in a Jasco 720 spectropolarimeter (Japan Spectroscopy Co., Tokyo) equipped with a rectangular cell, path length of 1 mm.\[[@CIT24]\] Scans were performed between 175 -- 250 nm, bandwidth 1.0 nm, and resolution of 100 mdeg. Measurements using pure buffer (2 mM HEPES, 100 mM NaCl, pH 7.4) were performed throughout, to test instrument reproducibility. Spectra of pure liposome preparations in the same solvent media at different concentrations were used in a blank experiment to be subtracted from the liposome / peptide spectra. The peptide concentration in buffer and in liposome / peptide mixtures was 36 *μ*M. Liposome concentrations were: 6000 *μ*M for DMPC (with 0.6% peptide), 1200 *μ*M for DMPG (with 3% peptide), and 3000 *μ*M DMPC : DMPG (3:1) (with 0.6% peptide), where peptide percentages were in mol / mol. The peptide / lipid ratio shown corresponded to the ones for which the best spectra definition was obtained (1:167 for the DMPC and for DMPC : DMPG; and 1:36 for DMPG). The desired amounts of peptide and liposome were mixed immediately prior to each measurement and incubated at 35°C for 30 minutes before the measurements, and performed at the same temperature. Each spectrum was always the average of nine accumulations. After blank correction, the observed ellipticity was converted to a mean residue molar ellipticity (θ) (deg^.^cm^2.^dmol^−1^), based on the total amount of peptide present in the mixture. Differential scanning calorimetry {#sec2-5} --------------------------------- Differential scanning calorimetry (DSC) was performed in a Micro-DSCIII microcalorimeter (SETARAM, Caluire, France) essentially as described previously.\[[@CIT24]\] Two successive up and down scans were performed for each sample, the up-scan at a scanning rate of 0.5°C/minute and the down-scan at 3°C/minute, over the temperature range 10 -- 35 °C. The sample mixtures were prepared immediately before the DSC run, by adding the desired amount of peptide (LFampin 265 -- 284, LFampin 265 -- 280 or LFampin 270 -- 284) stock solution to the LUVs suspension of DMPC, DMPG or DMPC : DMPG (3 : 1). Samples with 0.5, 0.75, 1.0, 2.0, and 3.0% (mol/mol) were used. All procedures regarding sample preparation and handling (lag time at low temperature, time between mixtures, and start of the experiment) were kept constant in all experiments, to ensure that all samples had the same thermal history. The instrument was electrically calibrated for temperature and the scan rate with the SETARAM Calibration Unit.\[[@CIT25]\] The Micro-DSCIII software was used for baseline subtraction (run with buffer solution on both cells (sample and reference)). The transition temperature Tm and the transition enthalpy change (∆~trans~*H*) were calculated by integration of the heat capacity versus temperature curve (Cp *versus* Temperature). A linear baseline was used to calculate the integral areas under the curves.\[[@CIT24][@CIT26][@CIT27]\] Results and Discussion {#sec1-2} ====================== Bactericidal and candidacidal activity of LFampin peptides {#sec2-6} ---------------------------------------------------------- The lead antimicrobial peptide LFampin 265 -- 284 comprises of a highly cationic C-terminal part and an α-helix facilitating N-terminal part.\[[@CIT21]\] To analyze the impact of either part on the antimicrobial activity of the bovine lactoferrin antimicrobial peptide LFampin 265 -- 284, its behavior was compared with two peptides truncated at either the N- or C-terminus of the LFampin 270 -- 284 and LFampin 265 -- 280, respectively. For representative target microorganisms, we used Gram-negative *Escherichia coli*, a rather harmless indigene of the lower intestine (although some strains could cause serious food poisoning in humans),\[[@CIT28]\] the Gram-positive *Streptococcus sanguinis*, which was a normal inhabitant of the healthy human mouth (although if it gained entrance into the bloodstream it was the most common cause of subacute bacterial endocarditis),\[[@CIT29]\] and the yeast *Candida albicans* which was an opportunistic pathogen and the causal agent of oral and genital infections in immunocompromised persons e.g., in AIDS, cancer chemotherapy, and transplantation patients.\[[@CIT30]\] When compared with the leading and most active peptide LFampin 265 - 284, although LFampin 270 -- 284 was found to be completely inactive, LFampin 265 -- 280 retained some activity toward the tested microorganisms. The levels of antimicrobial activities found with 50 *μ*M of LFampin 265 -- 280 were reached with less than 3 or 6 *μ*M LFampin against *C. albicans, E. coli*, and *S. sanguinis*, respectively \[[Figure 1](#F0001){ref-type="fig"}\]. Similar graphs were already found within 10 minutes of incubation and remained unchanged thereafter (not shown). ::: {#F0001 .fig} Figure 1 ::: {.caption} ###### Antimicrobial activity of LFampin peptides. Microorganisms were incubated with two-fold serially diluted peptides. Graphs represent fluorescence uptake after 1h of incubation :::  ::: Secondary structure of LFampin 265 -- 284, LFampin 265 -- 280, and LFampin 270 -- 284 peptides in buffer solution and in the presence of membranes, as studied by CD Indication of secondary structures of the three peptides was obtained by CD spectra, Figures [2a](#F0002){ref-type="fig"}--[c](#F0002){ref-type="fig"}. It can be seen that in Hepes buffer (2 mM and 100 mM NaCl) all peptides show a minimum, around 200 nm, characteristic of a random structure. ::: {#F0002 .fig} Figure 2 ::: {.caption} ###### CD spectra of LFampin 265-284 (a), LFampin 265-280 (b) and LFampin 270--284 (c) Buffer (36 μM peptide) (black stars); 6000 μM DPMC and 0.6% peptide (light gray squares); 1200 μM DMPG and 3% peptide (black triangles); in 3000 μM DMPC:DMPG (3:1) and 0.6% peptide (light gray circles) :::  ::: When in the presence of membranes, a different secondary structure is acquired, depending on the peptide and the nature of the model membrane. The CD spectrum of the lead peptide LFampin 265 -- 284, in the presence of DMPC, overlaps the one observed in the buffer, indicating no significant change of structure in the presence of liposomes for all concentrations tested (1200 -- 6000 mM, results showed only to 6000 mM). The other two peptides, LFampin 265 -- 280 and LFampin 270 -- 284, show a slight shift from 200 nm to the right, indicating that a small change of structure could be present, but without formation of the a well-defined different secondary structure. It must be noted that the peptides LFampin 265 -- 280 and LFampin 270 -- 284 in buffer do not present a CD spectra compatible with a pure random structure (as LFampin 265 -- 284 does), indicating that a mixture of structures is present. Therefore, the small shift cannot be over-interpreted, and thus there are no significant changes of secondary structure for any of the three peptides in the presence of DMPC membranes. In the presence of liposomes of DMPC : DMPG (3 : 1) (here used as model for pathogens) LFampin 265 -- 284 forms an α-helix structure, reflected in the two minima at wavelengths near 208 and 222 nm. LFampin 270 -- 284 presents a spectra close to the one obtained in the buffer, whereas, for LFampin 265 -- 280 some indication of an α-helix is apparent, although to quite a small extent (note that the obtained signal is always a weighted mixture of all the structures present) \[[Figure 2](#F0002){ref-type="fig"}\]. In order to evaluate the importance of charge effects and to compare them with the results we previously obtained for LFampin 265 -- 284 in pure DMPG membranes,\[[@CIT22]\] we did also study the truncated versions in the presence of this model membrane system. We could see \[[Figure 2](#F0002){ref-type="fig"}\] that LFampin 265 -- 284 also formed an α-helix in this membrane system \[[Figure 2a](#F0002){ref-type="fig"}\], perfectly super imposable with the spectra in DMPC : DMPG (3:1) discussed earlier, whereas, LFampin 265--280 clearly showed the presence of some α-helix structure (albeit to a smaller extent, for the same P:L ratio), and finally LFampin 270 -- 284 showed a shift in minima, but no clear α-helix structure. The fact that a helix was found for LFampin 265 -- 280 when the model membrane system was totally formed by the negatively charged DMPG, showed the importance of the membrane charge combined with the charge and amphipathic character of the peptide, as the presence of DMPC in the mixed membranes caused a charge distribution on the surface of the liposome, and thus a larger amount of peptide was necessary to induce the same amount of secondary structure for this peptide. In order to quantify the amount of each structure present, we calculated the percentage of the α-helix, β sheet, and randomized structures for each peptide in the presence of the three membrane systems. The percentages of each structure were calculated by fitting a linear weighted sum of structures, as proposed by Chen,\[[@CIT31]\] to the mean molar ellipticity per amino acid residue, using the Solver facility in Excel (Microsoft^TM^). The reference values for the average ellipticity of each structure were those provided by Greenfield and Fastman\[[@CIT32]\] based on synthetic polypeptides. The Excel sheet for this calculation was developed by us based on the references and values referred to, as most available software for these calculations was developed for proteins, and we found that this approach provided much better estimates. The obtained values can be seen in [Table 2](#T0002){ref-type="table"}. The results show, as expected from the spectra, that all peptides are still predominantly random in the presence of DMPC membranes, although LFampin 270 -- 284 presents a significant percentage of β sheet structure. ::: {#T0002 .table-wrap} Table 2 ::: {.caption} ###### Contribution (in %) of each secondary structure, α-helix, β-sheet, and random structure, to the total CD signal, calculated for each peptide by fitting procedures as described in the text ::: Peptides DMPC DMPG DMPC:DMPG (3:1) -------------------- ------ ------ ----------------- ---- ---- ---- ---- ---- ---- LFampin 265 -- 284 25 20 55 50 10 40 50 25 25 LFampin 265 -- 280 32 4 64 55 0 45 40 0 60 LFampin 270 -- 284 5 43 52 19 22 59 9 37 54 \* The estimated uncertainties from the fittings in the reported values are ± 2 ::: LFampin 265 -- 284 and LFampin 265 -- 280 show the highest percentage of α-helix (50 and 55%, respectively) in the presence of DMPG liposomes and that percentage is the same for LFampin 265 -- 284, in the presence of DMPC : DMPG (3 : 1), whereas for LFampin 265 -- 280 it is lower. This finding emphasizes the importance of using mixed membranes to simulate pathogen membranes (rather than purely negative membranes), as its use led us to differentiate between these two peptides' effects. LFampin 270 -- 284, on the other hand, still presents a higher percentage of random structure in all membrane systems studied. Indeed the peptide to lipid ratio also affects the formation of a secondary structure, as the structure formation is dependent on partitioning, and the amount of peptide in the membrane changes with the P : L ratio. The spectra presented were obtained for the most suitable P : L ratio, to produce a good signal in the CD spectra. The observed differences for the three peptides reflect their change in structure, and their strength relates to the degree of partition. The removal of the amino acid lysine (K) and arginine (R) (positively charged) and asparagine (N) and serine (S) (both polar) of the C-terminus side (from LFampin 265 -- 284 to LFampin 265 -- 280), decreased the peptide charge from + 4 to + 2, as well as its polarity, thus affecting the tendency to form an α-helix \[[Figure 3](#F0003){ref-type="fig"}\]. The absence of amino acids N, K, S, and R increased the hydrophobic moment (0.30 for LFampin 265 -- 284 and 0.37 for LFampin 265 -- 280) and increased the peptide hydrophobicity (- 0.337 for the LFampin 265 -- 284 and - 0.186 for the LFampin 265 -- 280) \[[Table 1](#T0001){ref-type="table"}\]. These values suggest a hydrophobic / hydrophilic balance more suitable for a good interaction between DMPG liposomes and the LFampin 265 -- 284 peptide. Moreover, the truncation of the amino acids sequence on the N-terminal side (from LFampin265 -- 284 to LFampin 270 -- 284) \[[Figure 3](#F0003){ref-type="fig"}\], with the removal of the amino acids aspartic acid (D) (residue 265, charged negatively), leucine (L) (residue 266, nonpolar), isoleucine (I) (residue 267, nonpolar), tryptophan (W) (268 residue, polar), and lysine (K) (269 residue, charged positively) did not change the charge of the peptide (+ 4), but decreased the hydrophobic moment (from 0.30 to 0.28 for LFampin 265 -- 284 and LFampin 270 -- 284, respectively) and decreased the hydrophobicity (from - 0.337 to - 0.437 for LFampin 265 -- 284 and LFampin 270 -- 284, respectively), which was reflected in an almost nonexistent tendency in LFampin 270 -- 284 to form the α-helix and partition into the membranes. Furthermore, the lack of tryptophan in LFampin 270 -- 284 was also important in the reduction of the partition, as this amino acid was often considered to have an important role in 'anchoring' the peptide in the membrane.\[[@CIT13][@CIT33]\] ::: {#F0003 .fig} Figure 3 ::: {.caption} ###### Helical wheel presentation of LFampin peptides, showings their amphipatic character. Neutral amino acids: light grey circles; positively charged amino acids: dark grey and negatively charged amino acids: white :::  ::: The CD results for the studied peptides are consistent with the literature results. NMR studies using negatively charged SDS micelles (sodium dodecyl sulfate) and zwitterionic DPC (dodecylphosphocholine) liposomes with the LFampin 268 -- 284 peptide (the original sequence of LFampin), showed that the peptide formed an α-helix involving only residues 1 -- 11 and the final six C-terminal residues remained relatively unstructured.\[[@CIT34]\] We observed a well-defined α-helix for the LFampin 265 -- 284 in DMPG and DMPC : DMPG liposomes, confirming the influence of the amino acids aspartic acid (D), leucine (L), and isoleucine (I) in the formation of the secondary structure. The secondary structures of LFampin 265 -- 284, LFampin 265--280, and LFampin 270--284 were determined in the presence of trifluoroethanol / water (TFE / water) by van der Kraan *et al*,\[[@CIT21]\] In this study the authors found that the three peptides were able to form an α-helix, but to a different extent: LFampin 265 -- 284 showed the highest tendency to form an α-helix, followed by LFampin 265 -- 280, and finally LFampin 270 -- 284. The residual tendency of LFampin 270 -- 284 to form an α-helix in helix-inducing solvents was not apparent in the presence of membranes (at the P:L ratios used), and this reinforced the need for these studies to be performed with membranes, when a biological correlation was aimed at. Moreover, studies with mimetic membrane allowed the differentiation of the secondary structures formed and their dependence on the composition of the membrane, which was a fundamental aspect in the possible biological implications. Interaction of the peptides with liposomes as studied by differential scanning calorimetry {#sec2-7} ------------------------------------------------------------------------------------------ The thermodynamic characterization of peptide / liposome interactions by DSC is based on the changes in the thermal profile and in the thermodynamic parameters characterizing the thermally induced transitions in liposome systems (T~m~ and ∆~trans~*H*), due to the presence of the peptides. The DSC curves of pure DMPC liposomes (LUVs) and peptide / lipid mixtures at different P : L ratios are shown in [Figure 4](#F0004){ref-type="fig"}. The derived phase transition temperature values T~m~ and change in transition enthalpy ∆~trans~*H* are reported in [Table 3](#T0003){ref-type="table"}. In all cases, we see that the peptide affects neither the thermal profile nor the derived thermodynamic parameters for the gel to liquid-crystalline transition of DMPC liposomes \[[Figure 4](#F0004){ref-type="fig"}\]. We can thus conclude that for the studied P : L ratios, none of the peptides partition to the zwitterionic membranes. The pre-transition is observed when MLVs are used (temperatures between 16 and 17.5°C)\[[@CIT35][@CIT36]\] are usually not so clear with LUVs, as it appears superimposed with the main transtition.\[[@CIT37]\] Its change is not discriminatory, as it disappears in the presence of most added drugs,\[[@CIT35][@CIT38]--[@CIT41]\] hence, we have not deconvoluted the two transitions and will not address it here. ::: {#T0003 .table-wrap} Table 3 ::: {.caption} ###### Transition temperature) Tm and transition enthalpy change ∆trans*H* values for the three peptides in the three model membrane systems studies, DMPC, DMPG, and DMPC : DMPG (3 : 1) liposomes. The results are presented as a function of peptide % (or P : L ratio) ::: *T~m~ (°C) / ∆~trans~*H* (kJ mol^-1^)* ---------------------------------------- ------- --------- --------- --------- --------- --------- --------- --------- --------- --------- 0 0 24.5/19 24.3/20 24.4/21 23.4/23 23.9/23 23.3/24 25.0/21 24.9/22 24.8/22 0 0 24.5/19 24.3/20 24.4/21 23.4/23 23.9/23 23.3/24 25.0/21 24.9/22 24.8/22 0.50 1:196 24.5/21 24.3/21 24.4/20 23.1/25 23.4/21 23.2/24 24.9/21 24.8/20 24.9/22 0.75 1:129 24.5/21 24.3/19 24.4/20 23.0/21 23.3/20 23.1/24 24.9/19 24.8/19 24.8/21 1.0 1:96 24.5/19 24.3/19 24.4/20 23.0/20 22.9/17 22.9/24 24.8/19 24.9/21 24.8/21 2.0 1:46 24.5/18 24.3/20 24.4/19 21.0/17 25.4/20 22.4/22 24.4/20 24.8/22 24.8/21 3.0 1:29 24.4/18 24.3/20 24.4/19 \- 25.5/17 21.9/22 23.9/18 24.6/22 24.8/20 Estimated uncertainties: within sample is ± 0.1°C for T~m~ and ± 0.5 kJ mol^-1^ for ∆~trans~*H*; and between samples is ± 0.3°C for T~m~ and ± 3 kJ mol^-1^ for ∆trans*H* ::: ::: {#F0004 .fig} Figure 4 ::: {.caption} ###### DSC curves for DMPC liposomes and the studied peptides at different peptide molar ratios. Black: pure DMPC; light gray: 0.5 mol% peptide; dark gray: 0.75 mol% peptide; gray: 1.0 mol% peptide; dash black: 2.0 mol% peptide; dash light gray: 3.0 mol% peptide. Lipid concentration was in all cases 3 mM :::  ::: The DSC curves obtained in the presence of DMPG liposomes at different P : L ratios are shown in the [Figure 5](#F0005){ref-type="fig"} and the respective thermodynamic parameters are reported in [Table 3](#T0003){ref-type="table"}. In this case the peptides alter both the profiles, the T~m~ and the ∆~trans~*H*(at the highest P : L ratios). Also here, the largest changes are observed for LFampin 265 -- 284\[[@CIT22]\] and LFampin 265 -- 280, and the smallest for LFampin 270 -- 284. At the highest percentage of peptide, 2 mol% (P : L = 1 : 46) and 3 mol% (P : L = 1:29), the ∆~trans~*H* decreased for the first two peptides, whereas, the parameters were the same (within uncertainty limits) for the later one. Furthermore, a broadening of the curves (decrease in cooperativity) and the appearance of a shoulder at higher temperatures was also observed in all the cases. Finally, it should be noted that for LFampin 265--284 and LFampin 265--280 the transition was totally distorted at the highest ratios, whereas, LFampin 270--284 still presented a transition curve, again showing that this peptide had the weakest interaction. ::: {#F0005 .fig} Figure 5 ::: {.caption} ###### DSC curves for DMPG liposomes and the studied peptides at different peptide molar ratios. Black: pure DMPG; light gray: 0.5 mol% peptide; dark gray: 0.75 mol% peptide; gray: 1.0 mol% peptide; dash black: 2.0 mol% peptide; dash light gray: 3.0 mol% peptide. Lipid concentration was in all cases 3 mM :::  ::: The shoulder at higher temperatures can be interpreted as reflecting the electrostatic attraction between the peptide and the lipid head, leading to a stabilization of the gel phase. Precipitation / aggregation was observed for the three peptides at the highest P : L ratios (2 and 3%). Finally the results for DMPC : DMPG (3 : 1) can be seen in [Figure 6](#F0006){ref-type="fig"}. For the studied P : L ratios, neither LFampin 265 -- 280 nor LFampin 270 -- 284 show significant changes in the DSC profile or parameters. The thermotropic behavior shown with LFampin 265 -- 284 is intermediate between the one observed for the zwitterionic DMPC and the anionic DMPG liposomes, and both Tm and ∆~trans~*H* decrease gradually as the P : L ratio increases [Table 3](#T0003){ref-type="table"}. At the highest P : L ratio a shoulder at higher temperatures is apparent, which is a strong indication of lipid segregation within the membrane, due to the preferential interaction of the peptide with the negatively charged lipid DMPG. Prenner *et al*,\[[@CIT42]\] also observed a more strong interaction with anionic DMPG than zwitterionic DMPC or DMPE phospholipid bilayers, with the cationic peptide gramicidin S (GS), due to electrostatic effects. At higher peptide concentrations, gramicidin S (GS) reduced the temperature, enthalpy, and cooperativity of the main phase transition in the DMPG bilayers. Furthermore, a decrease in cooperativity in the phase transition is observed when DMPG liposomes are involved, further substantiating a better interaction and higher disturbance of membrane patches, richer in DMPG when the mixed DMPC / DMPG system is used. The decrease in the transition enthalpy at higher P : L ratios is compatible with the partial insertion of the peptide in the lipid bilayer, which causes a change in the packaging of carbon chains, by disruption of the van der Waals inter- and intrα-molecular interactions.\[[@CIT24]\] ::: {#F0006 .fig} Figure 6 ::: {.caption} ###### DSC curves for DMPC:DMPG (3:1) liposomes and the studied peptides at different peptide molar ratios. Black: pure DMPC:DMPG; light gray: with 0.5 mol% peptide; dark gray: with 0.75 mol% peptide; gray: with 1.0 mol% peptide; dash black: with 2.0 mol% peptide; dash light gray: with 3.0 mol% peptide. Lipid concentration was in all cases 3 mM :::  ::: The analysis of the thermotropic profile of liposomes of different compositions in the presence of peptides, allows us to conclude that the interaction between antimicrobial peptides and lipid bilayers involves factors related to the characteristics of the peptide and also to the lipid composition of the membrane. The DSC data reflect the low affinity of these peptides to DMPC liposomes, and this is confirmed by our CD results where the peptides remain unstructured in the presence of this membrane system. Finally, the results obtained by DSC confirm that the presence of anionic lipid boosts the action of the peptide by the initial electrostatic interactions, facilitating peptide insertion and membrane destabilization. The characteristics of the peptides responsible for the distinct behavior include: the peptide charge, the tendency to form α-helix in the presence of the membrane, and the amphipathicity of the helix formed. The three peptides are capable of electrostatic interactions with the negatively charged membranes because all have a positive charge. LFampin 265 -- 284 and LFampin 270 -- 284 have the same nominal charge (+4) and LFampin 265 -- 280 has the lowest (+2), but nevertheless it disturbs DMPG to a much larger extent than LFampin 270 -- 284. This confirms that the charge alone is not the key factor in differentiating the effect of peptides on membranes and our results show that the secondary structure formed in their presence is a more important factor for a larger partition and consequently a more effective interaction. CD results \[[Figure 2](#F0002){ref-type="fig"}\] show that the LFampin 265 -- 284 forms the highest percentage of α-helix, followed by LFampin265 -- 280. In [Figure 3](#F0003){ref-type="fig"} it can be seen that LFampin 265 -- 284 has a helical structure, with the positive charges more clustered on one side of the helix, thus forming a more amphipathic α-helix, more suited to interact with the polar heads of the lipid bilayer. Thus, although both effects (charge and secondary structure) are important, the results indicate that the absolute charge is less significant in an effective interaction with the membranes, than the ability to form a well-defined secondary structure. The interaction of the original sequence of the LFampin peptide (residues 268 -- 284) with the multilamellar liposomes (MLVs) of DPPC and DPPG was studied by DSC, by Vogel and colleagues.\[[@CIT34]\] The results of this study showed that the peptide did not affect the thermotropic profile of zwitterionic liposomes (DPPC) across the range of the studied P : L ratios. In anionic liposomes (DPPG) the authors did not observe any significant change in the phase transition up to the highest P : L ratio studied. Our results for LFampin 265 -- 284, LFampin 265 -- 280, and LFampin 270 -- 284 peptides with DMPC liposomes led to similar conclusions, whereas, for charged liposomes they showed that an interaction occurred in the DMPG and DMPC / DMPG (3 : 1) membranes. The lack of change in the thermotropic profile of DPPG was explained by Vogel *et al*,\[[@CIT34]\] as being due to the fact that the LFampin 268 -- 284 peptide (charge + 5) could form electrostatic interactions only at the polar head level, without significantly altering the packing of the chains. A reasonable explanation for the differences between the two studies could be found in the membrane composition (DMPC vs. DPPC). DPPC had a much higher transition temperature (around 44°C) than the DMPC and the phase transition gel to *liquid-crystalline* was much more cooperative (particularly in the MLVs), due to the higher hydrocarbon chain length of the DPPC, which led to additional van der Waals interactions. This had to be responsible for a lower partition of the peptide to this model membrane system. Ladokhin and White\[[@CIT43]\] studied the interaction profiles of melittin with zwitterionic and anionic model membranes. The authors suggested that, unlike with PC membranes, melittin should not adopt a trans-membrane configuration when interacting with anionic liposomes (PG), and that the permeabilization of these later membranes by melittin was possibly due to a mechanism of 'leaky fusion'. These authors also showed that the mechanism of permeabilization of the membrane was not an inherent characteristic of the peptide, but strongly depended on the nature of the lipid bilayer. Epand *et al*, had proposed that cationic peptides (α/β peptides) and polymers that mimic antimicrobial peptides could segregate the anionic lipids from mixed membranes, forming rich negative lipid / peptide domains, causing defects in the membrane with consequent loss of internal content.\[[@CIT44]--[@CIT46]\] The observation of domain formation between the cationic peptide and negative lipid membrane had been proposed earlier by McLaughin and colleagues.\[[@CIT47]\] They studied the interaction between MARCKS and pp60^Src^ (mimetic peptides from the charged region of the protein as phospholipase C and kinase C) with mixed phospholipid membranes of PC / PG and PC / PS, and found that the strength of the interaction was influenced by the cationic amino acid residues content and by the anionic lipid fraction in the membrane. Furthermore, they reported that for the peptides studied, the electrostatic interaction was independent of the nature of the anionic lipid (PS or PG) and of the cationic amino acid residue (lysine or arginine). The authors interpreted the sigmoid shape of their binding curves as a function of the negative lipid fraction in the membrane, and as resulting from the peptide-induced formation of domains rich in negative lipid. The formation of domains with negative lipids was also reported by Lohner *et al*,\[[@CIT48]\] for the peptide PGLa. Our results for the biophysical characterization of the interaction of peptides with model membrane systems were also in very good agreement with the ones we obtained for antimicrobial activity (against *E. coli, S. sanguinis, and C. albicans*), even considering the simplicity of the model membranes used, where we only introduced the lipid DMPG for modeling the pathogens, as it was well known to be one of their major components. The peptide with the highest antimicrobial activity was LFampin 265 -- 284, followed by LFampin 265 -- 280, whereas, LFampin 270 -- 284 proved to be inactive against the tested microorganisms \[[Figure 2](#F0002){ref-type="fig"}\]. Our CD results indicated that LFampin 265 -- 284 had the highest percentage of α-helix in the presence of negatively charged membranes, followed by LFampin 265 -- 280, whereas, LFampin 270 -- 284 remained unstructured in the presence of all the studied membranes. This confirmed the importance of a secondary structure on the antimicrobial activity, due to different interactions with the bacterial membrane. The amino acid composition also had a strong reflection in the interactions with membranes. The differences found for LFampin 265 -- 284 and LFampin 265 -- 280 peptides confirmed that the presence of some amino acids altered the structural arrangement of the peptide and influenced their behavior in the presence of membranes, as well their antimicrobial activity. Furthermore, as discussed earlier, both these peptides were capable of forming an α-helix in the presence of negatively charged membranes, albeit to a different extent. This was reflected in the microbiological results, in the much higher dose needed for LFampin 265 -- 280 to produce the same effect as LFampin 265 -- 284. The initial peptide / liposomes interaction was caused by the electrostatic attraction between the negative lipid and positively charged peptide, but the degree of interaction was differentiated by secondary structure propensities and by the amphipathicity of the peptide. As such, the secondary structure seemed to be more important than the peptide charge on peptide / pathogen (as well as peptide / membrane) interactions. Nevertheless, there was an optimum balance between charge and secondary structure (LFampin 265 -- 284 vs. LFampin 265 -- 280). It should be stressed that the excellent correlation obtained between the results derived from studies with mimetic membranes and the ones obtained in vitro against different pathogens and erythrocytes\[[@CIT21]\] confirmed that biophysical experiments could be used in the initial screening of new peptides, helping in the design of new and more active drugs. Thanks are due to FCT for the financial support to CIQ(UP), Unidade de Investigação 81, and for a Doctoral grant to R.A. (SFRH/BD/24055/2005). We also thank Prof. João Pessoa, IST, Lisboa, for access to the CD instrument. **Source of Support:** Nil **Conflict of Interest:** None declared.

PubMed Central

2024-06-05T04:04:19.982976

2011-01-01

PMC3053523

Important biological mechanisms like chromatin function, as well as biomedical applications involving liposomal carriers for DNA delivery for clinical use in gene therapy, are based upon the incorporation of nucleic acids within lipid membranes.\[[@CIT1]--[@CIT4]\] Lipoplexes, that is, complexes of lipids with DNA that have been successfully used as transfection agents for gene delivery applications and viral capsids, where DNA-- or RNA--lipid interactions take place, are two of the numerous examples manifesting the great importance of unraveling the detailed mechanisms of lipid--nucleic acid complexation.\[[@CIT5]--[@CIT9]\] Understanding the phases and the molecular organization of lipid membranes and the structures of their complexes with nucleic acids is essential for optimizing gene delivery efficiency and understanding the biological mechanisms at the molecular level. The enormous possibilities presented by liposomes, exhibiting versatility with respect to their size, fluidity, and surface charge, have over the last 20 years paved the way for systematic research in the direction of developing stable liposomal complexes with proteins, peptides, metabolites, and drugs.\[[@CIT10]\] A large amount of work has already been published presenting results from extensive studies of interactions of lipid membranes with biomolecules, primarily explored by DSC, where the molecular order of lipids and the thermally induced phase transitions were probed.\[[@CIT11]--[@CIT13]\] Even though DSC proved to be a powerful and easy-to-operate technique, it became clear that for the correct interpretation and the deeper understanding of the calorimetric data, that is, heat capacity vs. temperature, the correlation of DSC with structural information from a variety of combined techniques was necessary.\[[@CIT14]--[@CIT19]\] The negative charge density of the DNA molecule makes the direct electrostatic interaction with cationic liposomes possible. The addition of DNA to liposomes brings about substantial changes related to the molecular order and the structure of the lipid membrane. These changes heavily depend upon the fluidity of the membrane, the surface charge, and often the size of the preformed cationic liposomes. In this review, we summarize the nucleic acids--membrane interactions studied via DSC. These interactions depend upon parameters such as electrostatic interactions, lipid structure and membrane composition, entropic contributions, mesoscale conformations of membranes, and nucleic acids properties. DSC has definite limitations for studying thermal processes even more so for extracting detailed information about molecular order and structure. Despite these limitations, DSC is a sound technique for exploring membrane--biomolecules interactions through changes demonstrated in the calorimetric trace of lipid phase transitions. By employing thermodynamic analysis, high-sensitivity DSC has become a systematic method for investigating DNA or RNA interactions with lipid bilayers, shedding light into phase separation, lipid segregation, liposomal aggregation and fusion, as well as membrane-induced structural changes of the nucleic acids. Lipid Phases and Phase Transitions {#sec1-1} ================================== Lipid molecules, when dispersed in aqueous media, self-organize spontaneously to a great variety of supramolecular structures depending upon the physicochemical properties of the lipids and the dispersant. This amazing polymorphism of lipids mainly includes lamellar, micellar, hexagonal, cubic normal, and inverted phases.\[[@CIT20][@CIT21]\] These phases and their transitions have been a broad field of research for many years because of their structural and functional biological importance.\[[@CIT22][@CIT23]\] DSC is the experimental technique of choice to systematically characterize the thermodynamics of phase transitions and conformational changes of biological macromolecules and supramolecular structures. DSC measures isobaric changes in specific heat capacity as a function of temperature.\[[@CIT24]--[@CIT26]\] Thermal events like phase transitions absorb or release heat upon heating or cooling of the system, and thus information can be retrieved pertaining on the transition temperature, the enthalpy change, and the cooperativity of the transition as it can be inferred from the heat capacity peak with straightforward analysis. Single transitions, multiple transitions, phase separation, and aggregation phenomena can be distinguished. By comparing heat capacity curves and entailed molecular order changes -- baring a cooperative character -- we can extract qualitative and quantitative information about the physical chemistry of mechanisms and interactions on a molecular level. DNA and RNA of a great variety in size and origin have been used for interaction studies (bovine/plasmid/herring/salmon 1--40 kbp). Synthetic lipid membranes have been tested either as multilamellar vesicles (MLVs) or as large unilamellar vesicles (LUVs). Lipid compositions assayed varied from pure zwitterionic phosphatidylcholine (PC) membranes to three lipid systems containing cationic compounds, fusogenic-helper lipids, or anionic lipids. Lipid phases involved were mainly lamellar (gel or liquid-crystalline) as well as inverted hexagonal or cubic.\[[@CIT10]\] DSC could reveal the effect of nucleic acids binding to the lipid bilayers through the changes induced upon the molecular order, and thus upon the phases and phase transitions. DNA--membrane Interactions {#sec1-2} ========================== A systematic study of the complexation of cationic liposomes with DNA providing evidence that the lipid hydrophobic part plays a crucial role in the interaction with DNA was published by Subramanian *et al*.\[[@CIT27]\] The study straightforwardly demonstrated the lipid phase separation within the liposomal membranes as well as the lipid-induced abolishment of DNA thermal denaturing transition. More specifically, Subramanian *et al*. investigated the formation of lipoplexes from calf thymus DNA and binary lipid multilamellar liposomes composed of 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC) and one of the three synthetic cationic amphiphiles -- bis\[2-(11-phenoxyundecanoate)ethyl\] dimethylammonium bromide (BPDAB), bis\[2- (11-butyloxyundecanoate)ethyl\] dimethylammonium bromide (BBDAB), and *N*-hexadecyl-*N*-{10-\[*O*-(4-acetoxy)-phenylundecanoate\] ethyl} dimethylammonium bromide (HADAB). All cationic lipids (CLs) bear a dimethylammonium bromide headgroup (DAB) as the DNA-binding moiety yet different hydrophobic parts.\[[@CIT27][@CIT28]\] Importantly, the miscibility of the CLs with DMPC was assessed by scanning liposomes composed of CL-DMPC mixtures with molar ratios from 0 to 1. The DSC profiles manifested that maximum miscibility was achieved for low CL to DMPC molar ratios (≤0.15). At intermediate and high molar ratios, multiple overlapping peaks were observed, proof of the formation of phase-separated domains. Particularly interesting phase diagrams for the three binary mixtures were constructed, based on the DSC data, demonstrating the existence of extended gel and fluid-phase coexistence regions. The DSC results demonstrated very different phase behavior depending on the lipid hydrophobic part. A DASM-4 (Bioripor, Pushchino, Russia) calorimeter was employed for the study. The addition of even 0.2 mM calf thymus DNA to the cationic liposomes produced lipoplexes and induced dramatic effects in the DSC thermograms. Of course, the lack of interaction between the zwitterionic DMPC and DNA produced no changes in the DSC profiles of pure DMPC MLVs.\[[@CIT29][@CIT30]\] The hydrophobic part of the cationic components affected the degree of the DNA effect on the DSC thermograms. At low ratios, DNA-induced lipid demixing resulted to phase separation; a single cooperative thermal anomaly was split into two broader peaks in the presence of DNA. As DNA concentration increased, the main phase transition temperature was suppressed and the enthalpic content of the peak was decreased. An additional heat capacity peak emerged at a higher temperature than the main-phase transition attributed to CL-rich domains. Despite the main-phase transition enthalpy decrease, the total enthalpy of both transitions was larger than the initial. At higher nucleotide per CL ratio, the main-phase transition peak became broader, and the main-phase transition enthalpy increased, while the new peak had reduced enthalpy until complete abolishment at a molar ratio of 8. The DNA effect on vesicles composed of nonaromatic hydrophobic chains BBDAB was qualitatively the same as in the case of the diphenyl hydrophobic chains of BPDAB. In the case of the incompatible tail components, phenylamine in HADAB differentiated the effects enhancing the disorder induction by DNA even at low HADAB to DMPC ratio. The zwitterionic lipid 1,2-dioleoylphosphatidylethanolamine (DOPE) has been widely used as a helper lipid in cationic lipid membranes for lipoplex formation as it enhances gene transfection efficiency due to its fusogenic influence. It is thus essential to study the properties of DOPE-containing membranes interacting with DNA.\[[@CIT31]--[@CIT38]\] Several DSC studies have been guided in this direction. Barreleiro *et al*. used dioctadecyl-dimethylammonium bromide - 1,2-dioleoyl-phosphatidylethanolamine (DODAB-DOPE) liposomes scanned by DSC to show that the inclusion of DOPE increased the fluidity of the membranes leading to phase separation at high DOPE contents.\[[@CIT39]\] The addition of DNA resulted in increased cooperativity for the main lipid phase transition, in case of pure DODAB, but it enhanced phase separation in the DODAB-DOPE dispersions. Saunders *et al*. have studied the effect of DOPE in ethyldimyristoyl-*O*-phosphatidylcholine (EDMPC) LUVs at various ratios and subsequently the effect of plasmid DNA (9 kbp) binding, by employing high-sensitivity DSC.\[[@CIT40]\] EDMPC-DOPE liposomes showed increased transfection efficiency as found by Gorman *et al*.\[[@CIT41]\] EDMPC is derived from DMPC by placing an ethyl to one of the phosphate oxygens resulting in a cationic lipid from a zwitterionic. Pure EDMPC vesicles exhibit a relatively cooperative main-phase transition with a low transition temperature (*T*~m~) shoulder attributed to an interdigitated phase. As in the previously described case of DODAB-DOPE, the addition of DOPE at 12% severely alters the lipid organization in the membrane by inducing higher membrane fluidity. It suppresses the main-phase transition peak while abolishing lipid interdigitation. Interestingly, a low-enthalpy DSC peak at a higher temperature than the main-phase transition provides proof of pure EDMPC domains in the mixture. The addition of plasmid DNA, at a DNA-to-EDMPC charge ratio of 0.5, reduces the enthalpy by \~7%, while it decreases drastically (110%) the half width at half maximum of the heat capacity peak indicating an increase in cooperativity. As DNA binds electrostaticaly to EDMPC headgroup, long-range order may be enhanced leading to the observed increase in the cooperativity of the main-phase transition. On the other hand, the more fluid EDMPC-DOPE membranes exhibit more dramatic effects when bound to DNA. The main-phase transition peak is suppressed in both temperature and enthalpy, in favor of the EDMPC-rich phase-separated regions. Interestingly, for EDMPC-DOPE liposomes at 50% mol DOPE, while noncomplexed liposomes present no DSC features, the DNA-complexed ones exhibit two distinct low-enthalpy peaks characteristic of the structural changes induced by the binding of DNA. When DNA is incorporated within the lipid membranes, the changes that occur are quite dramatic, as described before. The interactions are also very slow and in order to characterize them properly several consecutive heating and cooling DSC scans are required. This was particularly pronounced in a case of DNA-mediated segregation of lipid binary mixtures that differ in their aliphatic chains, which was observed in a study by Giatrellis *et al*., of binary 1,2-dimyristoyl-trimethylammonium propane - 1,2-dioleoyl-trimethylammonium propane (DMTAP-DOTAP) lipoplexes.\[[@CIT42]\] Upon consecutive heating scans, the single thermal anomaly from the neat DMTAP-DOTAP unilamellar liposomes main-phase transition that was originally detected underwent changes in the transition temperature and also in the number of DSC peaks and their total enthalpy. During the very first cycle from gel to liquid-crystalline phase and back to gel phase, multiple DSC peaks emerged; the highest temperature peak coincides with the temperature of the phase transition of multilamellar DMTAP-DOTAP vesicles. This can have multiple interpretations; DNA interconnects and restricts trimethylammonium propane (TAP) headgroups, while at the same time DNA can induce multilamellarity, and finally DNA might be selective to the more fluid DOTAP lipids and thus creates DMTAP-rich domains.\[[@CIT43][@CIT44]\] All three scenarios can be concurrent, as they are also supported by light-scattering particle size measurements as well as by isothermal titration calorimetry experiments. Following a number of successive order--disorder bilayer transitions, the abolishment of the DSC peaks can be achieved; during each heating--cooling cycle, DNA has been incorporated in the membrane in higher and higher concentrations leading to the establishment of a "uniform" fluid phase without detectable phase transitions.\[[@CIT14]\] Typical thermograms from MLVs interacting with plasmid DNA are presented in [Figure 1](#F0001){ref-type="fig"}, where the previously mentioned thermal behavior is expressed in a lower degree than for LUVs because of multilamellarity. Even so, the complexity of the system and the interactions bring to surface the weaknesses of the DSC technique and point out the need for complementary techniques. ::: {#F0001 .fig} Figure 1 ::: {.caption} ###### Characteristic DSC traces for DMTAP-DOTAP 4:1 multilamellar vesicles (top curve), DMTAP-DOTAP 4:1 -- plasmid DNA complex first heating scan (middle curve) and second scan (bottom curve). Experimental procedures are described in Giatrellis *et al*.\[[@CIT43]\] :::  ::: Very important for understanding biological functions in the cell nuclei as well as for biotechnology applications are the interactions of RNA with lipid bilayers. A very systematic study was recently published aiming at understanding how tRNA (baker's yeast, 60-90 bp) associates with DMPC membranes doped with the anionic lipid 1,2-dimyristoyl-sn-glycero-3-phospho-l-serine (DMPS) or the cationic amphiphile DODAB. The work was carried out by Michanek *et al*. by combining DSC with a quartz crystal microbalance with dissipation (QCM-D) techniques.\[[@CIT45]\] Poly-A short single-stranded DNA (63 bp) and double-stranded salmon sperm DNA (2000±500 bp) were also assessed for adsorption on lipid bilayers for comparison. tRNA was found to adsorb on DMPC liposomes after long incubation times with maximum adsorption taking place when incubation is performed at the liquid-crystalline (*L~α~*) phase of the membrane as also observed by Koiv *et al*.\[[@CIT46]\] Divalent cations did not affect tRNA adsorption. Adsorbed tRNA insignificantly increased the main-phase transition temperature implying that neither hydrophobic nor strong electrostatic interaction was taking place, although an increase in cooperativity was observed, as ∆*T*~½~, that is, the transition peak temperature width at half ∆*C*~p~ maximum, decreased from 1.6°C to 1.3°C. This weak binding was attributed to apolar interactions between tRNA and the apolar parts of the lipid exposed at the hydrophilic zone of the bilayer. Miscibility of DMPS was also assessed with DSC, giving a single and strongly cooperative peak for all DMPC-DMPS mixtures implying narrow coexistence regions for the two phases L~β~ (gel tilted) and L~α~. The presence of tRNA in DMPS mixtures, at a monovalent Na^+^ buffer, increased *T*~m~ significantly, broadened the transition peak, and in some cases caused the splitting of the peak. The intensity of these effects increased with the relative DMPS content in the mixture and the tRNA concentration. The broadening and the splitting of the transition peak were explained by lipid segregation. In this case, the repulsion is driving this segregation and not the attractive forces as studied in other works presented. DMPS lipids are repelled by the nucleic acid negative charges and they are segregated from the DMPC lipids, leading to the formation of DMPC-rich domains exhibiting lower transition temperature than the ideally mixed DMPC-DMPS. Those effects were maximum and permanent after the third heating cycle, implying that the system tRNA--liposomes was not at a steady state immediately after mixing and incubating at a stable temperature, but it required the lipid bilayers to undergo the transition from order to disorder twice for a stable state to be established. This supported the hypothesis that the exposed apolar parts of the bilayer are interacting with apolar moieties of tRNA. In contrast to these results, DODAB (5%) doped DMPC membranes reached a stable state in only minutes after mixing with nucleic acids due to the strong interaction between the cationic compound and the negative charges of the nucleic acids. The effect of tRNA on the main-phase transition of DMPC-DODAB was analogous to the DMPC-DMPS membranes, but driven by electrostatic attraction. The main characteristic of the DSC thermograms is the peak splitting which indicated the tRNA mediated segregation and the formation of DODAB-rich domains; DODAB domains associated to tRNA were also confirmed with solid NMR experiments.\[[@CIT47]\] DODAB-rich and DMPC-rich domains were characterized by different temperatures shifted towards the pure DODAB (36.0°C and 44.0°C) or DMPC (24.0°C) melting temperatures, respectively.\[[@CIT48]\] A point of difference between the two segregation cases -- weak repulsive or strong attractive forces -- was the concentration of the tRNA which was 10 times lower for the cationic membranes but it was enough to cause drastic changes to the lipid bilayer. When assessing different nucleic acid interactions with DMPS containing membranes, similar effects have been observed only for the single-stranded DNA of comparable size to the tRNA (ssDNA~63~) most probably because it is more agile and it contains unpaired bases as tRNA. Larger DNA molecules, either double-stranded or single-stranded, left the main-phase transition of DMPC or DMPC-DMPS binary liposomes unaffected. Conclusively, it has been shown by DSC that tRNA induces lipid segregation either by attractive or repulsive forces. Details from the nucleic acids--membranes interaction as induction of disorder in the lipid bilayer or lipid headgroup interconnection and segregation can be inferred by DSC experiments. Nevertheless, a second experimental technique as nuclear magnetic resonance (NMR) or small-angle x-ray scattering (SAXS) must be employed in order to verify the proposed interaction model.\[[@CIT47]\] One of the earliest systematic studies of DNA and RNA effects on thermal phase behavior of multilamellar and unilamellar liposomes, employing DSC, was conducted by Koiv *et al*.\[[@CIT46]\] Liposomes composed of the zwitterionic lipid DMPC doped with the cationic single-tail lipid sphingosine at various ratios where studied for calf thymus DNA and RNA effects. Vesicles containing neat DMPC showed again no effect on both pretransition (gel to ripple phase -- *L*~β~ → *P*~β~) and main-phase transition (ripple to liquid-crystalline phase - *P*~β~ → *L*~α~) by the presence of DNA and RNA. Introducing sphingosine in the PC bilayers resulted in nucleic acid binding and significant changes in the thermal profiles. Phase separation was observed by the splitting of the main-phase transition peak. Lipid segregation was manifested by shifting the phase transitions toward higher temperatures implying the formation of sphingosine-rich domains. At the same time, the reduction of the total enthalpy at increasing DNA concentrations and the abolishment of the initial -- lower temperature -- phase transition peak indicates DNA-imposed disorder in the lipid bilayer. In connection to the previous studies of PC-containing membranes, calorimetry was combined with small-angle neutron scattering (SANS) and SAXD to study the interaction of 1,2-dipalmitoyl-sn-glycero-3-phospho-l-choline (DPPC) liposomes, mediated by Zn^2+^, with salmon sperm fragmented DNA (0.5--1 kbp) by Uhrikova *et al*.\[[@CIT30]\] Structural information from the DPPC-DNA-Zn ^2+^ complex was correlated to the thermotropic phase transitions in relation to Zn^2+^ concentration. Zinc cations bind strongly to PC and induce dehydration as well as conformational changes to the hydrophilic part of the DPPC lipid bilayers.\[[@CIT49][@CIT50]\] At the same time, zinc can condense DNA, like other divalent cations, and thus it can bridge PC headgroups with DNA polyanions.\[[@CIT51]\] The DPPC molecular order was almost unaffected by the presence of DNA in a monovalent buffer environment as indicated by the *T*~m~ of the pretransition and main-phase transition, in contrast to Suleymanoglu 2004 results where DNA had severe impact on DPPC bilayers.\[[@CIT52]\] At low zinc concentrations, the DSC trace of the DPPC-Zn^2+^-DNA complex was altered substantially. The pretransition peak was reduced to the detection limit and the main-phase transition had a slight temperature increase, and it appeared as a shoulder over a new high-enthalpy peak at higher temperature. The high-temperature peak was assigned to the phase transition of the DPPC-Zn ^2+^ -DNA complex, and the lower transition was the main-phase transition of DPPC-Zn ^2+^. These assignments were also inferred by SANS and SAXD experiments. Two lamellar phases were confirmed; the first was attributed to the DPPC-Zn^2+^ lamellar structure, and the second to the DPPC-DNA structure in which DNA strands are intercalated in the aqueous planar space between sequential DPPC bilayers. The enthalpy content for each transition could be deconvoluted to quantify the two phases. That analysis showed that at low zinc concentration, as of 1 mM, there was maximum DNA complexation. Further increase in zinc cations to 20 mM Zn^+2^ significantly decreased the amount of complexed DNA, as DNA was neutralized by the excess zinc cations. At high zinc content, a low-enthalpy peak manifested the presence of DPPC-DNA complex. DSC data analysis also defined the zinc--DPPC interaction saturation point and stoichiometry to three cations per DPPC molecule. In another recent study by Xu and Anchordoquy, DNA carriers composed of DOTAP and high cholesterol content exhibited interestingly high gene delivery efficiency.\[[@CIT53]\] DSC was employed to investigate the anhydrous crystalline phase of cholesterol formed over 60% and the effects of bound DNA.\[[@CIT54]\] The anhydrous neat cholesterol domains or "crystallites" exhibit a phase transition of around 40°C, which were not affected by DOTAP-bound DNA. Moreover the cholesterol domains transition enthalpy change was unaffected at high cholesterol content. Even though, an enthalpy decrease trend with high uncertainty was observed for increasing DNA amount in cholesterol 70% vesicles; this could be possibly due to DNA-induced disorder at the cholesterol domains interface. DMTAP-DMPC binary multilamellar liposomes and their complexes with calf thymus DNA have been thoroughly studied utilizing DSC in combination with SAXS and WAXS by Zantl *et al*.\[[@CIT31]\] The phase behavior of the binary lipid mixture at various molar ratios was explored with DSC as well as their complexes with DNA, and detailed phase diagrams have been created. The phases observed were lamellar as confirmed by the structural experiments. The basic DNA effect was the increase in transition temperatures as DNA stabilized the membranes at the isoelectric DNA to cationic lipid ratio. Additionally, at low TAP content, lipid demixing was reflected as a phase transition peak splitting toward higher temperatures compared to lipid mixtures in the absence of DNA. The applicability value of the phase diagrams is realized in the following study by Koynova *et al*., where binary lipids phase diagrams were correlated to transfection efficacies. Extensive DSC studies of numerous binary lipid formulations with potential transfection applications --ethylphosphatidylcholines, dimethylammonium bromides, DMTAP, egg PC, 1,2-dioleoyl-sn-phosphatidylglycerol (DOPG) -- was conducted by Koynova *et al*.\[[@CIT55][@CIT56]\] Phase diagrams of binary lipid membranes -- self-organized in lamellar phases -- were correlated to transfection efficiencies and revealed that maximum delivery efficiency was for binary lipid formulations that exhibit gel-liquid crystalline phase coexistence at physiological temperature. In line with the latter were the results of novel synthetic cationic amphiphiles applied for the *in vitro* lipofection of mouse melanoma cell lines.\[[@CIT34]\] It must be mentioned at this point that DSC has been employed in several cases to characterize the thermal denaturation of DNA in relation to membrane binding, where increasing melting temperature indicated enhanced stability for the formed lipoplex.\[[@CIT46][@CIT57]--[@CIT59]\] Lobo *et al*. have showed that DNA was thermally stabilized by DAB cationic lipids more than TAP lipids showing a shift from 109 °C to 122 °C.\[[@CIT58]\] This effect was bidirectional in terms of DNA-induced lipid order; DNA increased the main-phase transition temperature of DDAB by 4.8 °C, while the equivalent increase for 1,2-distearyl-trimethylammonium propane (DSTAP) was only 2.0 °C. Additionally, they have found that DOPE abolished the stabilization in contrast to cholesterol which had no effect. Finally, in another study, DSC has also been employed to investigate hydration levels of cationic lipids--DNA complexes in the absence and presence of helper lipids as a complementary technique to laurdan fluorescence generalized polarization by Hirsch-Lerner and Barenholtz.\[[@CIT37][@CIT60][@CIT61]\] Cationic lipids used in that study were DOTAP, DMTAP, and DC-Chol (3β-{*N*-\[(*N*´,*N*´-dimethylamino)ethane\]-carbamoyl}cholesterol) while for helper lipids DOPE, DOPC, or cholesterol where chosen. Correlation of DSC Data with Other Techniques {#sec1-3} ============================================= DSC can provide information regarding thermally induced phase transitions and with that as a probe information about DNA and membranes interactions can be extracted. DSC cannot give direct lipid--DNA structural information and must be combined with other experimental techniques in order to interpret the acquired calorimetric data with greater confidence. As it was mentioned earlier, in the study by Koynova *et al*., lipid phases were assessed relatively to their efficiency for DNA delivery.\[[@CIT55]\] In yet another extensive study though, DNA delivery efficiency could be correlated to lipid--DNA complex phases.\[[@CIT62]\] Structural information of the lipid phases was extracted by SAXS. Mixtures of the cationic lipids ethyldilauroyl-O-phosphatidylcholine-ethyldioleoyl-*O*-phosphatidylcholine (EDLPC-EDOPC), DOTAP-DOPE, and their lipoplexes were assessed. Anionic lipids -- phosphatidylglycerol, phosphatidylserine, and cardiolipin -- were mixed and characterized in connection with DNA dissociation and delivery. The molecular structure of the lipids and the consequential apolar-polar interface curvature regulated the final structural conformation. Lipoplexes of EDLPC-EDOPC exhibited lamellar conformation; EDLPC-EDOPC-cardiolipin exhibited inverted micellar cubic -- the highest interfacial curvature -- while the less-effective mixtures EDLPC-EDOPC-cardiolipin exhibited inverted hexagonal, cubic bilayered, and/or coexisting lamellar phases. The most widely studied mixture, DOTAP-DOPE, at low temperature appears to be, while at higher temperatures exhibits an inverted hexagonal phase. Early SAXS structural studies by Lasic *et al*. on DODAB-cholesterol-DNA complexes revealed a bilayered packing with dehydrated DNA ordered in a 2D layer between the cationic lipid bilayers.\[[@CIT63]\] Very few studies employ only DSC for the investigation of membrane--DNA interactions.\[[@CIT52]\] Most of the studies combine SAXS, WAXS, SANS, or other structural techniques like cryo-EM and solid-state NMR.\[[@CIT30][@CIT31][@CIT47][@CIT63][@CIT64]\] In some other cases, binding assays have been conducted employing QCM-D or ITC in order to add pieces to the lipids--DNA interaction puzzle.\[[@CIT43][@CIT46][@CIT65]\] Summary {#sec1-4} ======= The power of DSC as an experimental tool for characterizing nucleic acids--lipid membrane interactions is projected throughout all the studies reviewed here. DSC peaks associated with phase transitions or with the thermal denaturation of nucleic acids are used as the principal experimental evidence in all the studies. Straightforward analysis provides information on the enthalpy, the transition temperatures, and the cooperativity of the lipoplex phases. Most of the studies conducted involved binary lipid mixtures. It is essential, as a first step, to study the miscibility of the two lipids by DSC and characterize the phase transitions for safe interpretation of the experimental data from the DNA or RNA impact. In most cases and at low DNA concentration, surface-bound DNA will stabilize the lipid membranes through electrostatic interactions with the polar headgroup, though always depending on the membrane fluidity. When DNA penetrates the lipid membrane, the results are severe; the interactions can be slow and the fluidity of the lipid phase is almost always enhanced whilst lipid interdigitation is most surely abolished as is the lipid pretransition. In the binary lipid mixtures, one of the two components is the DNA-binding partner, while the other plays either a stabilizing or a destabilizing role for the long-range molecular organization. Thus, DNA is anticipated to demix lipids. Indeed, lipid demixing and segregation take place in most charged-zwitterionic binary systems resulting in phase-separated domains. Another issue accentuated is the concentration effect of nucleic acids; at low concentration, DNA can induce long-range order, while at high concentration it induces disorder. Information about the kinetics and the nature of the nucleic acids--membranes interactions can also be extracted with DSC. For very weak or slow-kinetics interactions, the establishment of a stable state can be monitored and demonstrated by successive heating and cooling scans and the repeatability of the so-obtained DSC thermograms. In several studies, the steady state is established after long incubation times and a large number of heating and cooling cycles. Since nucleic acid--membrane interactions also involve apolar interactions, several passes from order to disorder of the lipid bilayer will enhance apolar--apolar interactions and assist the establishment of a new phase for the complex. Recent developments in gene delivery involve novel cationic delivery reagents as amidines, polyethylenimines (PEI), and commercially available reagents (Polyplus-transfection SA, 67401 Illkirch Cedex, France) as well and exhibit increased transfection efficiencies.\[[@CIT66]--[@CIT68]\] An extensive biophysical approach on these novel gene carriers can also contribute to a better understanding of the mechanisms behind improved transfection.\[[@CIT69][@CIT70]\] DSC is a valuable technique for studying DNA--lipid complexes, which exhibits many advantages such as the relatively easy-to-use experimental setups and the straightforward data analysis. As a technique, DSC is efficient and relatively of low cost, reliable as an ensemble technique, and most importantly it can assess systems in physiological conditions. DSC should be used in combination with other techniques like structure analysis techniques or binding assays. Once the thermal behavior of a system is correlated with structure facts, then DSC can be used independently. Both qualitative and quantitative analysis can be carried out with DSC, contributing to the elucidation of biological mechanisms as well as to the development of improved gene carriers. **Source of Support:** Nil **Conflict of Interest:** None declared

PubMed Central

2024-06-05T04:04:19.987193

2011-01-01

PMC3053524

Biological membranes are constituted by lipid bilayers as their basic structure. Lipid bilayers, which are sheet-like assemblies of amphiphilic lipid molecules held together by hydrophobic interactions between their acyl chains, form the boundaries between intracellular cytoplasm and the cell's outside environment, as well as between the interior of many of the cellular organelles and the cytoplasm. This lipid bilayer structure was first recognized as the basis for cell membrane architecture in 1925;\[[@CIT1]\] however, only in 1972, Singer and Nicholson\[[@CIT2]\] first proposed a fluid mosaic model, to explain the membrane structure. According to this model, lipids and proteins diffuse freely within the plane of the cell membrane. Since then, large membrane domains (e.g., basal, lateral, and apical membrane regions of glandular, endothelial, and epithelial cells) and lateral microdomain structures (e.g., lipid rafts, caveolae, and coated pits) have been discovered, which reveal the complex nature of the cell membrane structure.\[[@CIT3]\] Membrane lipids belong to three groups: glycerol-based lipids (phospholipids), ceramide-based sphingolipids, and cholesterol. Phospholipids are further divided into different groups, depending on their hydrophilic head groups: phosphatidylcholine (PC), phosphatidylethanolamine (PE), and phosphatidylserine (which are largely present in the cellular membrane); in addition phosphatidylinositol and cardiolipin are present in smaller quantities. Membrane constituents are not always homogeneously arranged in the bilayer membrane of biological cells, but rather organized in complex lateral microdomains. This polymorphic nature of lipid arrangement, in addition to a significant variety of lipids, with distinctly different physical properties (i.e., cross-sectional area, fluidity, electric charge, molecular weight), make the lipid membranes extremely intricate structures. Furthermore, the covalent association of proteins and carbohydrates adds to the complexity of this membrane's structure. As a consequence of this intricacy of the cell membrane structure, along with the highly dynamic nature of the lipid--lipid and lipid--protein interactions in the cell membrane, the biophysical interactions with drugs and drug delivery systems are very difficult to investigate. Therefore, simplified artificial membrane systems, mimicking the natural bilayer lipid membrane, have been developed. In this article, we show how differential scanning calorimetry studies of biophysical interactions with lipid model membrane systems permit to better elucidate the absorption process of a drug by a biological membrane after the release of the drug from a delivery system or after the dissolution process. Biomembrane Models {#sec1-1} ================== Biomembrane models are systems in which the lipid organization mimics the arrangement of lipids in natural cell membranes. Supported lipid bilayers,\[[@CIT4]\] lipid monolayers,\[[@CIT5][@CIT6]\] and liposomes\[[@CIT7]\] are widely used biomembrane models. In this contest just a brief description of liposomes, which are extensively used in our research, will be provided. Liposomes are spherical lipid vesicles, with internal aqueous compartments. This simple structure still possesses two fundamental properties of the biological membranes: (i) due to the hydrophobic effects, membranes tend to form closed shapes, which are called vesicles; (ii) lipid molecules can move around (diffuse) rather rapidly and freely within the membranes, as they are usually in the fluid state.\[[@CIT8]\] The use of a liposome as a tool to determine the ability of biologically active compounds (BA), to interact, dissolve, and penetrate a lipidic bilayer has been studied over the years, and a vast amount of research has been carried out in an attempt to evaluate the ability of liposomes (MLV or LUV), to operate both as a barrier in mimicking biological membranes or as a carrier in the transport of BA. The structural units of liposomes are amphiphilic molecules, especially phospholipids. The most abundant lipids in liposomes are phosphatidylcholines (PCs), in which a glycerol bridge links a hydrophobic part with a hydrophilic polar head group. In contact with water, PCs form bilayers, due to their molecular tubular shape, differently from detergents, which form micelles because of their conical shape. Other polar lipids such as sphingolipids and amphiphiles can be introduced in the liposomal bilayers. Cholesterol or other sterols, isolated from natural sources, and several lipid conjugated polymers, may be found in the liposome bilayers. The phospholipids exhibit thermotropic mesomorphism. When heated, they undergo a number of phase changes before reaching the fusion. Frequently occurring among them are the lamellar subgel L~c~ and the gel L~β~ phases, which are stable at low temperature, and the lamellar liquid crystalline L~α~ phase, stable at higher temperatures. The biological significance of the latter phase is well known, as it is accepted that the cellular membranes are liquid crystalline bilayers with proteins embedded in them.\[[@CIT9]\] The thermodynamic properties of hydrated lipids depend on the molecular structure and on the composition of the lipid dispersion. The phosphatidylcholine phase behavior is affected by hydrocarbon chain length, unsaturation, asymmetry, and branching, as well as the type of chain--glycerol linkage and the position of chain attachment to the glycerol backbone, the head group modification, the stereochemical purity, and the morphology of the lipid aggregates (unilamellar and multi-lamellar vesicles). In addition the phase behavior is influenced by the composition of the aqueous dispersing medium.\[[@CIT10]\] A lipid thermodynamic database (LIPIDAT) collects, in one central location, all information on lipid mesomorphic and polymorphic transitions and miscibility. The database is considered comprehensive for glycerophospholipids, sphingolipids, glycoglycerolipids, and biological membrane lipid extracts.\[[@CIT11]\] This thermotropic behavior is investigated in depth by the calorimetric technique applied to the studies on lipids and biomembranes. A pronounced and easily detectable effect of solutes on lipid phase behavior is the shift, upward or downward, of the transition temperature. However, the other thermodynamic characteristics of the lipid phase transitions, enthalpy, transition width, and maximum specific heat, can be also influenced by solutes. As the biomembranes are constituted by a multitude of lipids, these studies were mainly carried out on synthetic and a single kind of lipids, so that the interaction between BA and lipids could be easily considered. It has been several years since our research group is involved in the study of the release of drugs from delivery systems to biomembrane models using the DSC technique. In this article the main results obtained from such studies will be presented. The steps of the protocol used are as follows: To evaluate the interaction between the drug under study and the biomembrane modelsTo determinate the real amount of drug present in the phospholipid and the aqueous phases of the biomembrane model dispersionTo evaluate the factors affecting the kinetic of absorption of the free drug by biomembrane modelsTo ascertain that the delivery system does not interact with the biomembrane modelsTo evaluate the release of the drug from the delivery system to the biomembrane model In addition a brief description of the DSC technique will be provided Differential Scanning Calorimetry {#sec1-2} ================================= Differential scanning calorimetry (DSC) is the most frequent technique used for determining the thermal effects of a variety of materials, including biological systems that are characterized by an enthalpy change and temperature variation. Besides being involved in the determination of the effect of hydration, pH, solvent, and kind of composition, on the phase transition, and of the changes of enthalpies of model lipid membranes and phospholipid bilayers, it is used in the thermal characterization of complex processes such as the denaturation of proteins and to study glass transition of polymers. Differential scanning calorimetry scans temperature and measures the difference between the heat flows to a sample and a reference pan that is under the same temperature program, at atmospheric pressure, and measures the heat capacity of a material. Differential scanning calorimetry measures the heat flow going into or being released by a material. From that, the heat capacity at constant pressure (*Cp*) can be calculated. Heat capacity units are cal °C^−1^ or J °C^−1^. It measures the amount of heat input (q) required to raise the temperature of a specimen by one degree Celsius while at constant pressure. Heat capacity is usually normalized by dividing the specimen heat capacity by the number of grams, to get the heat required to raise one gram of specimen by one degree Celsius. This corresponds then to the specific heat capacity *Cp*. If desired, the heat capacity can be normalized by the number of moles. Heat capacity is defined by *Cp* = (δ*q* / δ*T*)*p*,where T is the temperature and *q* is the heat input. If the temperature changes from T~0~ to T~1~, the enthalpy of the reaction $\left( {\mathrm{\Delta}H} \right)\ {is}\ \mathrm{\Delta}H\ = \ \int_{T_{0}}^{T_{1}}CpdT$ Usually, ∆T is small, and *Cp* is independent of temperature between *T*~0~ and *T*~1~ . The integral thus reduces to ∆*H* = *Cp* (*T*~1~ - *T*~0~) = *Cp∆T*.\[[@CIT12]\] Interaction Drug / Biomembrane Model {#sec1-3} ==================================== The first step to be taken into account in the study of the release of a BA from a delivery system to a biomembrane model is the evaluation of the interaction of the BA with the biomembrane model. In order to evaluate this interaction, the lipid vesicles are usually prepared in the absence and in the presence of increasing amounts of BA, and the measurement of the effect of the BA on the thermotropic behavior of the phospholipid bilayers (T~m~ and ∆H), applying the van't Hoff model, is carried out.\[[@CIT13]\] It is, in fact, known that, for dilute solutions, the presence of a solute in the solvent can modify the thermodynamic parameters (such as the melting temperature) of the solvent. The solute acts as an impurity toward the solvent and the modification is dependent on the amount of the solute. In a similar way, the presence of BA in the ordinate lipidic structure can affect the thermodynamic parameters of the transition from the ordinate gel phase to the disordered liquid crystalline phase.\[[@CIT14]--[@CIT17]\] The effect is correlated to the amount and to the collocation (for similar compounds) of BA in the lipidic structure. The biomembrane models (MLV or LUV) are generally prepared following the methods reported a little later in the text, depending on the solubility of the BA: \(a) BA soluble in organic solvent: To prepare the MLV, the phospholipid and the BA are separately dissolved in an organic solvent (generally chloroform / methanol, 1 : 1). Aliquots of the phospholipid solution are distributed in glass tubes in order to have the same amount of phospholipid in all the tubes. In the same tubes aliquots of the BA solution are added in order to have an increasing amount of BA. The solvents are removed under a nitrogen stream (at a temperature higher than the transition temperature of the phospholipid),and then, by freeze drying. To the obtained films a known amount of aqueous solvent at a well-defined pH (generally 50 mM Tris, pH 7.4) is added. The samples are vortexed for one minute and heated for one minute at a temperature higher than the transition temperature, thrice, and then left for one hour in a thermostated bath (at a temperature higher than T~m~). The latter step permits two important processes: the homogeneous repartition of the BA between the lipidic and the aqueous phases and the aggregation of eventual small unilamellar vesicles (SUV). For the LUV preparation, the MLV, with or without the BA, are repetitively (19 times) passed under moderate pressure at a temperature at least 5°C above the T~m~ through polycarbonate membranes (pores diameter 100 nm) in an extruder system (LiposoFastTM Basic, Avestin Inc.).\[[@CIT18][@CIT19]\] The membrane pores are almost cylindrical, and the vesicles (unilamellar or multi-lamellar) that are larger than the mean pore diameter are reduced in size and lamellarity during the passage through the pores, resulting in a final vesicle size that corresponds to the mean size of the pores.\[[@CIT20][@CIT21]\] \(b) BA soluble in aqueous solvent: To obtain the MLV, a phospholipid solution in an organic solvent is prepared. A BA solution in an aqueous solvent is prepared. Aliquots of the phospholipid solution are distributed in glass tubes so as to have the same amount of phospholipid in all the tubes. The solvents are removed under a nitrogen stream (at a temperature higher than the transition temperature of the phospholipid), and later, by freeze drying. To the obtained films, defined amounts of the BA solution are added, in order to have an increasing amount of BA. The samples are vortexed for one minute and heated for one minute, at a temperature higher than the transition temperature, thrice, and then left for one hour in a thermostated bath (at a temperature higher than the T~m~). The LUV is prepared as described herewith. To determine the real amount of phospholipid present in each sample, the phospholipid phosphorous content is assessed in the preparation by a phosphate assay, after destruction with perchloric acid.\[[@CIT22]\] The samples are submitted to DSC analysis and the transition temperature is reported as a variation with respect to the T~m~ of the biomembrane models prepared without BA, as a function of the amount of the BA. Usually, a linear correlation between the transition temperature variation and the amount of BA exists, as shown in [Figure 1](#F0001){ref-type="fig"}, where the transition temperature variation is reported as ∆T / T~m~^0^ (∆T = T~m~-T~m~^0^; where T~m~ is the transition temperature of the biomembrane models prepared in the presence of BA and T~m~^0^ is the transition temperature of the biomembrane models prepared without BA). The calorimetric curves obtained from these experiments will be used as reference in the experiments, where the release of a BA from a delivery system to the biomembrane models is studied, as successively described. ::: {#F0001 .fig} Figure 1 ::: {.caption} ###### Depression of biomembrane model transition temperature as a function of the molar fraction of BA compound present in the aqueous lipid dispersion. Adapted from Castelli *et al*.\[[@CIT23]\] :::  ::: Partition of the Drug between the Lipidic and the Aqueous Phases {#sec1-4} ================================================================ When the interaction BA / biomembrane model is evaluated, the repartition of the BA between the phospholipid bilayers and the aqueous medium where the biomembrane models are dispersed have to be taken in consideration. The obtained graph described earlier \[[Figure 1](#F0001){ref-type="fig"}\] reflects the effect exerted not by the entire amount of BA present in the biomembrane model sample (aqueous lipidic dispersion), but only the effect exerted by the amount of BA present in the phospholipid bilayer. Then, the amount of BA in the phospholipid phase, which really causes the effect has to be determined. To conduct this, the biomembrane model prepared in the presence of different molar fractions of BA is ultracentrifuged. The supernatant (containing the BA in the aqueous phase) is separated from the pellet (containing the BA in the lipidic phase), whose volume is corrected for the entrapped aqueous volume. The two aliquots are freeze dried and the obtained powders are solubilized in the opportune solvent. The amounts of BA in the two phases are determined by UV / VIS spectrometry. Then, the graph in [Figure 1](#F0001){ref-type="fig"} can be corrected by multiplying the values of the molar fractions (corresponding to the transition temperature variation) for the H~2~O / lipid partition coefficient, and subsequently, the effect on the transition temperature can be attributed to the real amount of BA present in the lipidic phase. In this manner, the obtained line (line b of [Figure 2](#F0002){ref-type="fig"}) can be used as the real calibration curve. The values of this curve permit to transform the effect exerted on the transition temperature of the biomembrane model with the amount of BA present in the phospholipid bilayer.\[[@CIT23]--[@CIT25]\] ::: {#F0002 .fig} Figure 2 ::: {.caption} ###### Calibration curve relating the depression of biomembrane model transition temperature to increasing concentration of BA present in: (a) the aqueous lipid dispersion; or (b) effectively dissolved in the lipid matrix. Adapted from Castelli *et al*.\[[@CIT23]\] :::  ::: In addition, the log P value can give useful information on the lipophilic character of the BA and it can be obtained from the bibliographic data or otherwise it can be assessed by using computational approaches (see for example Pallas 3.0 (CompuDrug International, Inc. San Francisco USA); OSIRIS Property Explorer).\[[@CIT26]\] Transfer of a BA from a Drug Delivery System to a Biomembrane Model {#sec1-5} =================================================================== The use of biomembrane models is not limited to getting information on the interaction of BA with the biomembranes. They can, in fact, be used to evaluate the release of drugs from drug delivery systems. Usually, the release kinetic of a drug from a delivery system (hydrogel polymeric systems, nanoparticles, and so on) is evaluated by the use of dissolution systems, which allow to get a graph, where the amount of drug released is plotted as a function of time. This route to evaluate the release does not take into account the absence of a system that is able to take up the drug released during the time, which is instead possible in the presence of a biomembrane model. To follow the release kinetic with an *in vivo*-like system, the differential scanning calorimetry can be used based on the following protocol. First of all, from the experiments where the interaction between the BA and the biomembrane model is studied, the molar fraction of the drug that has a strong interaction with the biomembrane model, but at the same time, produces a well-defined calorimetric curve, is chosen as a reference curve for the highest interaction between BA and the biomembrane model. This curve will be considered in all these experiments as the curve toward which that of the biomembrane model should tend if the BA was transferred to the lipid bilayers. Then, before following the transfer of a BA from a delivery system to a biomembrane model, the release (dissolution) and the interaction of the BA left in contact with the biomembrane model has to be considered. This experiment can clarify the effect, due to the dissolution and the migration and the successive uptake by the biomembrane model in the absence of a delivery system where the BA is molecularly dispersed. The experiments can be summarized as follows: An exact amount of the drug is weighted at the bottom of a DSC aluminum pan and a defined amount of biomembrane model dispersion is addedAn exact amount of drug loaded delivery system (containing the drug molar fraction chosen) is weighted at the bottom of a DSC aluminum pan and a defined amount of biomembrane model dispersion is addedThe unloaded delivery system (with the same amount of that at point b) is weighted at the bottom of a DSC aluminum pan and a defined amount of biomembrane model dispersion is added The pans are sealed and submitted to DSC analysis as follows: (a) A heating scan at the rate of 2°C/minute in a range of temperature starting about 20°C below the transition temperature of the phospholipid employed and ending about 20 C above the transition temperature of the phospholipid employed; (b) Next, the samples are left at this temperature for 60 minutes; (c) A cooling scan to bring the sample back to the starting temperature. These steps are repeated, at least, eight times. By this procedure the interaction of the delivery system and of the drug with the biomembrane model is followed both during the heating step and during the isotherm step, when the phospholipids are in a disordered liquid-crystalline phase. In fact, when the phospholipids are in a disordered phase, the drug is more able to penetrate the multibilayer (MLV) or the bilayer (LUV) and eventually to bind the phospholipids and localize among them. This process is slowed during the cooling step. The degree of interaction of the drug with the phospholipid and then the release from the carrier is quantified through the variation of the biomembrane model transition temperature. The protocols reported earlier in the text are aimed at obtaining the following information: When the free drug is put in contact with biomembrane dispersion, we can not only follow the uptake of the drug by the biomembrane model \[[Figure 3a](#F0003){ref-type="fig"}\], but also the dissolution and diffusion processes through the aqueous phase of the drug, which, in this case, are affected by the uptake process. We can also measure the effect of the pH on the dissolution of the drug and consequently on its interaction with the biomembrane model.When the drug loaded delivery system is put in contact with biomembrane model dispersion we can evaluate the uptake of the drug released from the delivery system, the process of hydration of the dissolution of the delivery system, the drug diffusion through the delivery system, and successively through the aqueous phase \[[Figure 3b](#F0003){ref-type="fig"}\]. In this case even the effect of the pH on these processes can be evaluated.When the unloaded delivery system is put in contact with the biomembrane model dispersion, we can evaluate the eventual effect of the system on the thermotropic parameters of the biomembrane model. In case we obtain no effects, we can ascertain that when we put the loaded delivery system in contact with the biomembrane model, the effect on the transition temperature is exerted just by the drug. In this process even the effect of the pH can be monitored. ::: {#F0003 .fig} Figure 3 ::: {.caption} ###### Release of the BA as a free form (a) or from the delivery system (b) and transfer to the biomembrane model :::  ::: Depending on the use of MLV and LUV, different information can be obtained. Let us consider the use of MLV first. The permeation of a multi-lamellar vesicle by a BA, which is in the proximity of the biomembrane model, determines interesting effects on the thermotropic behavior of the membrane. The time variation of the specific heat profile can be used to get information on the BA diffusivity within the multi-lamellar vesicle, as shown in [Figure 4](#F0004){ref-type="fig"}. At an early stage of the process the calorimetric curve is similar to that of the pure lipid, with a sharp maximum near the T~m~ of the pure phospholipid; at intermediate times a splitting of the calorimetric peak is observed due to the local inhomogeneities of the composite membrane; finally, in the late stages of the process, the two peaks merge forming a single peak that has shifted to a lower temperature than that of the pure phospholipid phase transition.\[[@CIT27]\] ::: {#F0004 .fig} Figure 4 ::: {.caption} ###### Upper part: the concentration (Ф) profile of a BA inside a multi-lamellar vesicle at different times. *r* is the multi-lamellar vesicle radius. At *t* = 0 all the BA molecules are in the external aqueous medium (A). At intermediate times (B) one observes a nonlinear concentration gradient throughout the membrane which becomes linear in the late stages (steady-state C). At *t* = ∞ the BA concentrations in the external and internal aqueous medium become identical (D). Lower part: the corresponding variation of the specific heat (*Cp*) with temperature (*T*) at different times. *T*~m~^0^ is the melting temperature of the unperturbed multi-lamellar vesicle. Adapted from Raudino *et al*.\[[@CIT27]\] :::  ::: Both, when the BA is in a free form or it is contained in the delivery system, three different cases can take place. The BA does not transfer or transfers in a very low amount, through the aqueous medium, to the biomembrane model.After the first incubation period, the BA transfers through the aqueous medium into the outer MLV layers, and successively, it transfers to the inner layers.The BA quickly transfers, partially or completely, to all the MLV layers. Then, the examination of the calorimetric curves permit us to determine which of the three cases occurs \[[Figure 5](#F0005){ref-type="fig"}\]. Let us consider the case (a) the condition of non-transfer of the drug to the MLV is evidenced by the unchanged calorimetric curve during the entire incubation time. ::: {#F0005 .fig} Figure 5 ::: {.caption} ###### Panel A. Calorimetric curves during the permeation kinetic of a drug released from a delivery system through multi-lamellar vesicles. The arrows mark the position of the multi-lamellar vesicles at t = 0. Adapted from Raudino *et al*.\[[@CIT27]\] Panel B. Schematic representation of what happens in the calorimetric pan. In (a) the drug is not released from the delivery system and the calorimetric curve remains unchanged. In (b), initially, the drug is released from the delivery system and transferred to the external bilayers of MLV and a double peak is observed in the calorimetric curve; as the time passes the drug localized also in the internal bilayers and a unique peak is observed in the calorimetric curve. In (c) the drug released from the delivery system quickly localizes uniformly in all the bilayers and a unique peak is observed from the first times of contact :::  ::: In the case (b), the calorimetric curve associated to the early time of contact between the delivery system and the MLV, is very similar to that of the MLV; as the time passes the main peak splits into two peaks, caused by the inhomogeneous distribution of the BA in the MLV; in particular the BA mostly localizes in the outer bilayers; the successive scans allow the BA to pass into the inner layers by a flip-flop mechanism, in a series of mobile equilibria, among the BA loaded layers, the BA unloaded layers, and the aqueous medium. Such equilibrium brings to the final state and the two peaks merge in an unique peak due to the homogenous distribution of the BA in the multi-lamellar bilayers. The occurrence of case (c) is evidenced by the presence of an unique main peak which, quickly or slowly, moves to a lower temperature than that of the pure phospholipid. As stated before, by using the MLV we were able to detect the uptake process of a BA (free or delivery system released) by the biomembrane model, by considering the T~m~ shift. If the T~m~ moved toward the value observed when the same amount of BA was placed in the MLV during their preparation, we considered it a complete transfer of the BA into the inner lipid layers. To be sure that such a process was complete and to better define if the limiting step of the transfer process could be attributed to the dissolution and migration in the aqueous medium, rather than the permeation and partition into the lipid multilayers, LUV were employed. The LUV, having only one bilayer, do not show the problem of the BA transfer from the aqueous medium to the outer bilayers and then to the inner bilayers; then we can consider, by studying their behavior, the transfer limitation due to the medium surrounding the vesicles. [Figure 6](#F0006){ref-type="fig"} shows the three possibilities that can occur when a BA (free or loaded in a delivery system) is left in contact with the LUV. ::: {#F0006 .fig} Figure 6 ::: {.caption} ###### Calorimetric curves during the permeation kinetic of a drug thought unilamellar vesicles. The arrows mark the position of the unilamellar vesicles at t = 0. In (a) the drug is not released from the delivery system and the calorimetric curve remains unchanged. In (b) the drug is slowly released from the delivery system and transferred to LUV and the calorimetric peak slowly moves to lower temperature. In (c) the drug is quickly transferred from the delivery system and transferred to LUV and the calorimetric peak quickly moves to lower temperature. Adapted from Raudino *et al*.\[[@CIT27]\] :::  ::: As stated before for MLV, in the case (a) the BA is not able to reach the LUV surface even after several incubation periods at a temperature higher than the T~m~ . In this case, processes such as low dissolution rate and unfavorable partition between delivery system and water avoid the BA-LUV interaction. If the BA dissolves in the medium surrounding it or it is released by the delivery system and migrates to the LUV surface interacting with it, we can differentiate if the limiting step is the process of dissolution or release with respect to the interaction with the LUV bilayer by comparing the curves (b) and (c). In b, it is evidenced that the slow interaction BA-LUV can be ascribed to a lower availability of the drug near the LUV surface; while the curves (c) indicate a fast interaction that can be ascribed to the fast absorption caused by the high availability of the BA. The interaction observed when a fixed amount of BA is put in contact with an MLV or LUV, compared with that observed when the BA is loaded in the vesicles during their preparation, gives useful information about the processes of dissolution, release, and uptake. The comparison between the effect exerted by a BA on MLV or LUV (curves (b) and (c) in Figures [5](#F0005){ref-type="fig"} and [6](#F0006){ref-type="fig"}) allows to determine if the limiting step is the uptake or the transfer process and also gives information about the ability of the BA to cross the lipid layers penetrating inside the MLV. In fact, depending on the rate of interaction, if the transfer of BA from the outer to the inner layers of the MLV is complete, beside a uniform distribution, we will obtain a calorimetric curve similar to that of LUV and to those of MLV, prepared in the presence of the same amount of BA. This protocol was used for instance to follow the release of model drugs from an inulin-based hydrogel (intended to release the drug in the colon), α,β-polyasparthydrazide (PAHy) hydrogels, α,β-poly(N-hydroxyethyl)-[DL]{.smallcaps}-aspartamide (PHEA) hydrogels, Eudragit microparticles, and poly(lactide-co-glycolide) micropheres. With regard to the release of the drug (diflunisal) from the inulin-based hydrogel, the influence of the drug loading of the hydrogel swelling degree and of the pH on the release was evidenced.\[[@CIT28]\] In that study, DMPC unilamellar vesicles were used as the biomembrane model. The results obtained at pH 7.4 suggested that with a drug loading of 10.4% (w/w) the hydrogel swelled, but diflunisal, that was not so abundant inside the polymeric network, it would take time to dissolve and migrate through the network. When hydration and swelling occurred, the release would become faster. For the hydrogel with the highest drug loading (24%, w/w) we observed an initial fast release, due to the free drug molecularly dissolved inside the polymer; the following release was also fast, but not complete, perhaps due to the high amount of drug loaded in the hydrogel that could affect the swelling and so the release; or it could be also possible that during dissolution the drug started to crystallize due to the high concentration, forming large crystals that obviously dissolved slowly. The best profile of release was obtained with a drug loading of 17% (w/w); wherein besides a fast initial release, a complete release of the drug occurred, probably due to a good balance between drug dispersion in the polymeric network and swelling of the hydrogel. At pH 4, the diflunisal release from the hydrogel with a drug loading of 10.4% (w/w) was low for all the experimental period, due to the low amount of drug present in the hydrogel, the lower swelling at such a pH, and the low-aqueous solubility of the drug; all these factors contributed to reducing the release of diflunisal. The release of the drug from the hydrogel with a drug loading of 24% (w/w) was slower than that with a drug loading of 17% (w/w), and incomplete, thus indicating that the formation of drug aggregates together with migration through the polymeric network, made it difficult for the dissolution process to occur. Seventeen percent (w/w) represented the best loading, allowing a complete release and transfer of the diflunisal to the biomembrane model. Dipalmitoylphosphatidylcholine (DPPC) LUV and DSC were employed to study the suitability of the hydrogels obtained by chemical crosslinking of α,β-polyasparthydrazide (PAHy) by glutaraldehyde, as carriers for prolonged release of poorly soluble drugs, and the modulating effects exerted by polymer crosslinking.\[[@CIT23]\] It emerged that by increasing the polymer crosslinking degree the total amount of transferred drug and the release velocity were decreased. This behavior may be caused by the increase in the number of cruciate bonds in the hydrogels, which cause a free volume reduction, obstructing the passing drug. Successively, PAHy hydrogels were crosslinked with different agents (ethyleneglycoldiglycidylether (EGDGE) polyvinylalcohol (PVA) and glutaraldehyde (GLU)) and different degree of crosslinking and the release of diflunisal by these hydrogels was carried out with the aim to evaluate the effect of the crosslinking agent and the degree of crosslinking on the release. It appeared evident that the total amount of drug transferred and the release rate were affected by the polymer crosslinking degree (it increased with the crosslinking degree) as well as on the nature of the crosslinking agent.\[[@CIT29]\] The study of the ketoprofen release from a derivative of the α, β-polyapartylhydrazide polymer containing hexadecylamine to biomembrane models made of DMPC / DMPA vesicles permitted (i) to demonstrate that the polymer was able to release the drug and (ii) to explain how the release could happen; it was in fact, proposed that the hexadecylamine moiety penetrated into the lipidic bilayers, followed by the delivery of the drug. In this manner the micelles could improve the localized release close to the biological target.\[[@CIT30]\] The analysis of the release of ketoprofen from polymeric micelles made of PHEA-C16 and PHEA-PEG2000C16 to DMPC / DMPA vesicles permitted to hypothesize a likely mechanism of drug migration from the micelles to the vesicles: We can suppose that the C16 chains interact with the lipidic bilayer, thus causing the opening of the micellar structure and facilitating drug penetration in the lipidic vesicle.\[[@CIT31]\] Biomembrane models made of DPPC MLV were used\[[@CIT32]\] to study the release of moxifloxacin from uncrosslinked and crosslinked (glutaraldehyde as crosslinking agent) chitosan microspheres intended for pulmonary administration. The results showed that uncrosslinked microspheres swelled rapidly and dissolved, releasing free chitosan that was able to interact with the liposomes (increase of ∆H value), probably altering the biomembrane permeability to the drug. Crosslinked microspheres did not show this property. The release of diclofenac from Eudragit RS100^®^ microparticles and the effect of the pH on such a release was also investigated.\[[@CIT33]\] In such a study of DMPC MLV (as biomembrane models), microparticles containing two different amounts of diclofenac (14.26 and 25.0%) and two different pHs (7.4 and 4.0) were used. At pH 7.4, the release process for the microparticles loaded with a higher amount of drug appeared to be faster with respect to the lower loaded microparticles, both being slower than the free drug. In fact, the drug dissolved easily in the external pH 7.4 medium and was readily absorbed by the biomembrane model. Drug release from the microparticles was hindered by the acidic pH, which prevented its dissolution and migration through the aqueous medium to reach the model membrane. The results obtained from this study suggested that the process of drug dissolution through a polymeric matrix could be affected by the amount of drug loaded, but mainly by the pH of the dissolution medium. The release of flurbiprofen, an acidic drug, from Eudragit RS100^®^ and Eudragit RL100^®^ nanosuspensions, was studied at pH 7.4 and the results obtained from DSC, where a biomembrane model (DMPC MLV) was used, and the dialysis experiments were compared.\[[@CIT34]\] The results showed a plateau in the dissolution profile of flurbiprofen from the nanoparticles, which was related to an equilibrium among the drug release, the drug ionization in the dissolution medium, and the saturation of the binding sites on the surface of polymer particles. This behavior was ascribed to the fact that the dissolved drug that got ionized in the neutral dissolution medium was readsorbed onto the polymer particles because of the presence of opposite electrical charges. In the dialysis experiments the driving force leading to flurbiprofen release from the nanoparticles was the volume and light alkaline pH of the dissolution buffer; the absence of an uptaking system in the external medium made the release profile more strictly dependent on the nature of the polymers. In fact the two polymers showed similar time-release curves, but with a higher amount of drug released from the more permeable RL matrix. Although, when the biomembrane model was used, the drug release profile was conditioned by the MLV bilayers to capture and retain the drug molecules after their release from the polymeric system. In these experiments the volume of the dissolution medium was much smaller with respect to that of the dialysis experiments. The equilibrium among the drug bound to the nanoparticles surface, the fraction dissolved in the medium, and the amount captured by MLV were strongly affected by the affinity of the drug for the RS or RL polymers. As a consequence, the behavior observed for RL and RS nanosuspensions was quite different. In fact, although the maximum released amount was higher for the RL system due to its higher water permeability, the MLV incubated with RS nanoparticles showed a quicker uptake of the dissolved drug. The comparison of the results obtained from the two different kinds of experiments permitted us to conclude that although in the dialysis tests the permeability of RS and RL polymers was the limiting step of the release of an acidic drug such as flurbiprofen, in the smaller space of the DSC pan, the affinity equilibrium for a nanoparticle surface or MLV bilayers played a big role in determining the overall drug release profile. Similar results were obtained in a study where the release of ibuprofen from Eudragit RS100^®^ and Eudragit RL100^®^ nanosuspensions was studied.\[[@CIT35]\] Transfer of a BA from a Lipophilic Drug Delivery System to Biomembrane Models {#sec1-6} ============================================================================= The previous model can also represent a useful approach in studying the transfer of a drug from a lipophilic carrier to the biomembrane model, taking advantage of the use of a drug-loaded MLV, using the following protocol: In the calorimetric pan, equimolar amounts MLV, loaded with a known molar fraction (X = n) of the BA, which mimics a lipophilic carrier, and of the drug-unloaded MLV (X = 0), which mimics the biomembrane model, are put in contact and submitted to calorimetric analysis, in which a heating scan is followed by an isothermal period of one hour at a temperature higher than that of the phospholipid and then by a cooling scan, several times. Three different cases can occur: (a) the drug does not transfer from the loaded to the unloaded MLV, (b) the drug slowly transfers from the loaded to the unloaded MLV, and (c) the drug quickly transfers from the loaded to the unloaded MLV. The analysis of the calorimetric curves and their comparison with the calorimetric curve (reference curve) obtained from the MLV prepared in the presence of BA at X = n/2 gives us information on which of the cases occurs \[[Figure 7](#F0007){ref-type="fig"}\]. The curve X = n/2 should be obtained if 50% of BA is contained in the loaded MLV transferred to the unloaded MLV. ::: {#F0007 .fig} Figure 7 ::: {.caption} ###### Left part. Calorimetric curves of unloaded MLV (X=0) (which represent the biomembrane model) put in contact with an equimolar amount of loaded MLV (X=n) (which represent the lipophilic carrier), at increasing time of incubation. Right part. Schematic representation of what happens in the calorimetric pan. The calorimetric curves identified as X=n/2 is used as reference and it should be obtained in the case the drug transfers from loaded to unloaded MLV till to have the same concentration in all the MLV which has a mean value among the concentrations of the MLV which were put in contact. (a) The drug does not transfer from loaded MLV to unloaded MLV and the reference curve is never reached; (b) At the beginning the drug does not transfer from loaded to unloaded MLV; as the incubation time passes the drug transfers till its concentration will be the same if all the MLV and the reference curve is reached. (c) The drug quickly transfers from loaded to unloaded MLV and the reference curve is reached :::  ::: If case (a) occurs the calorimetric traces are characterized by two peaks, simply resulting from the sum related to the unloaded and loaded MLV, which were put in contact. This curve remains unaltered for all the calorimetric scans without reaching the reference curve (shape or position). If event (b) takes place, the calorimetric curves exhibit two peaks related to the unloaded and loaded MLV that were put in contact, in which, as the incubation time passes, the drug transfers from the loaded to the unloaded MLV and the two peaks approach each other (the peak related to the loaded MLV loses the BA and moves toward a higher temperature; whereas, the peak related to the unloaded MLV is enriched with BA and moves toward a lower temperature) and merge in a unique peak, which overlaps the reference curve. The case (c) is characterized by a single calorimetric peak, due to the fast transfer of the BA from the loaded to the unloaded MLV and with a fast transfer from the outer to the inner bilayers, which gradually moves toward lower temperature. If the transfer is complete the calorimetric curve reaches the reference curve. This protocol was used with gemcitabine and acyclovir prodrugs obtained by the conjugation of the drug with Squalene.\[[@CIT36][@CIT37]\] These prodrugs were obtained with the aim of increasing the lipophilic character of the drug. It was found that the affinity of the prodrugs to the biomembrane models was much stronger with respect to the free drug and the prodrugs were released from the loaded MLV to the unloaded MLV very gradually, suggesting that the liposomes could be used as a carrier for the sustained release of the prodrugs. In a recent article of our research group it was shown that trimethylresveratrol transfers slowly from the lipophilic carrier to the biomembrane models; whereas resveratrol exhibits a quicker kinetics of transfer.\[[@CIT38]\] These behaviors can be explained by the different lipophilicity of the two compounds. Trimethylresveratrol being more lipophilic than resveratrol has a bigger affinity for the lipophilic environment, with respect to resveratrol; hence, it leaves the lipophilic carrier slower than resveratrol. In a recent research\[[@CIT39]\] the transfer of omega-3 fatty acids from a lipophilic carrier to biomembrane model was studied. From that study it has emerged that docosapentaenoic acid (DPA) and docosahexaenoic acid (DHA) possessing the same carbon atoms, but differing in the unsaturation (DPA with five double bonds and DHA with six double bonds) show different behaviors. In fact, the transfer from the loaded to the unloaded MLV is slow and not complete for DPA, but quick and complete for DHA. The smaller and slower transfer of DPA was attributed to its stronger affinity for the lipophilic environment of MLV. Toward Modelization of the Combined Release-uptake Kinetics {#sec1-7} =========================================================== The great deal of experimental study described in the previous sections can be exploited in order to gain quantitative information about a number of useful parameters related to the release-uptake process in biological systems. This goal can be accomplished by developing a non-trivial mathematical modelization of the involved coupled processes. Here we sketch the minimal physical effects to be considered in a minimal but realistic picture. The starting point for any release kinetics is the transport equation $$\frac{\partial C_{i}}{\partial t}\ = \nabla J_{i}\left( C_{i} \right) + F\left( C_{i} \right)$$ which states that the time variation of the concentration *C~i~* of a generic i-th species depends upon the difference between the ingoing and outgoing flux *J~i~(C~i~)* inside a generic volume element of the system (the term ∇*J~i~(C~i~)*, the symbol ∇ meaning spatial difference) augmented by a term that describes the concentration variation related to eventual chemical reactions (the term *F(C~i~)*. The flux J~i~(C~i~) can originate from different causes. In the simplest case it originates from the concentration gradients (Fick's law: *J~i~(C~i~)* ≈ *D~i~gradC~i~*, where *D~i~* is the diffusion coefficient of the i-th species). Things are generally more intricate because the diffusion coefficient *D~i~* may depend on the swelling degree of the carrier's matrix (see, e.g., Grassi *et al*,\[[@CIT40]\]), moreover, other complex effects, such as hydrodynamic effects, may contribute. A similar equation must be used to deal with the uptake of a diffusing molecule. There is, however, a key difference, because a key contribution comes from the 'immobilization' of the processes occurring when the diffusant is trapped at a generic binding site. These immobilization effects can be described in a variety of ways, one of the simplest ones is to introduce a fictious chemical reaction, which destroys the diffusing particles when they hit the binding sites within the target. Such an effect modifies the local concentration of the diffusant, and therefore, the whole diffusant motion. In the case of a prescribed distribution of irreversible traps, the kinetic term to be added to the diffusion equation takes the simple form: *F(Ci)* ≈ *k* *(C*($\overset{\rightarrow}{r}$)-- *C~i~)C~i~*, where *C*($\overset{\rightarrow}{r}$)is the spatial distribution of the target's binding sites and *k* is the kinetic constant of the binding process (therefore the quantity *C*($\overset{\rightarrow}{r}$)-- *C~i~* describes the number of unoccupied binding sites). The above-mentioned model leads to two coupled diffusion-type equations (associated with the release and uptake processes, respectively) that can be easily solved. Once the time-varying local concentration of the diffusant has been calculated, we can easily calculate the variation of the calorimetric properties in each region of the sample. Indeed, as previously stated, the transition temperature and the associated enthalpy of a system (e.g., a multi-lamellar liposome) depend on the local concentration of a foreign compound; such a relationship is empirically determined by calibration curves. Thus the real calorimetric signal calculated at different times is proportional to the space concentration of the diffusant averaged over the whole sample. The advantage of the above sketched mathematical model is twofold: It enables one to numerically derive useful parameters such as the diffusion mobility of a drug in different matrices (e.g., a multi-lamellar liposome), the water-lipid partition coefficient, the release kinetic constant from a dissolving matrix, and so on.To test the validity of release-uptake mechanisms by the fitting of theoretical and experimental curves. Studies along these lines will be discussed in a near future. In conclusion, the use of biomembrane models and differential scanning calorimetry are useful approaches to study the release of a BA from a delivery system, as also the uptake of the BA from biomembranes models. In our researchers we have studied the release of BA from several kinds of delivery systems and we have obtained useful information, not only on the kinetics of the release, but also on the uptake of the BA by the biomembrane. The described approach can be applied to study the BA release from delivery systems other than those described in this article. For example, in some of our studies we have employed the differential scanning calorimetry to characterize solid lipid nanoparticle (SLN) and nanostructured lipid particle (NLP) colloidal carriers,\[[@CIT41][@CIT42]\] developed as alternative systems to the existing traditional carriers (emulsions, liposomes, and polymeric nanoparticles), especially for the delivery of lipophilic compounds and recently, due to its versatility, we have applied the described approach to follow the release of drugs from SLN and NLP, and their uptake by biomembrane models. **Source of Support:** Nil **Conflict of Interest:** None declared.

PubMed Central

2024-06-05T04:04:19.990714

2011-01-01

PMC3053525

In the last two decades, mucoadhesion has shown renewed interest for prolonging the residence time of mucoadhesive dosage forms through various mucosal routes in drug delivery applications. Mucoadhesive-based topical and local systems have shown enhanced bioavailability. Mucoadhesive drug delivery gives rapid absorption and good bioavailability due to its considerable surface area and high blood flow. Drug delivery across the mucosa bypasses the first-pass hepatic metabolism and avoiding the degradation of gastrointestinal enzymes. Thus mucosal drug delivery system could be of value in delivering a growing number of high-molecular-weight sensitive molecules such as peptide and oligonucleotides. In this review, the aim is to provide detailed understanding of mucoadhesion, bioadhesion of polymer, and techniques for the determination of mucoadhesion; finally most common routes of mucoadhesive administration will be presented along with examples of formulation studied. Bioadhesion and Mucoadhesion {#sec1-1} ============================ The term bioadhesion can be defined as the state in which two materials, at least one biological in nature, are held together for an extended period of time by interfacial forces.\[[@CIT1]\] In biological systems, bioadhesion can be classified into 3 types: Type 1, adhesion between two biological phases, for example, platelet aggregation and wound healing.Type 2, adhesion of a biological phase to an artificial substrate, for example, cell adhesion to culture dishes and biofilm formation on prosthetic devices and inserts.Type 3, adhesion of an artificial material to a biological substrate, for example, adhesion of synthetic hydrogels to soft tissues\[[@CIT2]\] and adhesion of sealants to dental enamel. For drug delivery purposes, the term bioadhesion implies attachment of a drug carrier system to a specified biological location. The biological surface can be epithelial tissue or the mucus coat on the surface of a tissue. If adhesive attachment is to a mucus coat, the phenomenon is referred to as mucoadhesion. Leung and Robinson\[[@CIT3]\] described mucoadhesion as the interaction between a mucin surface and a synthetic or natural polymer. Mucoadhesion should not be confused with bioadhesion; in bioadhesion, the polymer is attached to the biological membrane and if the substrate is mucus membrane the term mucoadhesion is used. Theories of Mucoadhesion {#sec1-2} ======================== Various theories exist to explain at least some of the experimental observations made during the bioadhesion process. Unfortunately, each theoretical model can only explain a limited number of the diverse range of interactions that constitute the bioadhesive bond.\[[@CIT4]\] However, four main theories can be distinguished. Wetting Theory of Mucoadhesion {#sec1-3} ============================== The wetting theory is perhaps the oldest established theory of adhesion. It is best applied to liquid or low-viscosity bioadhesives. It explains adhesion as an embedding process, whereby adhesive agents penetrate into surface irregularities of the substrate and ultimately harden, producing many adhesive anchors. Free movement of the adhesive on the surface of the substrate means that it must overcome any surface tension effects present at the interface.\[[@CIT5]\] The wetting theory calculates the contact angle and the thermodynamic work of adhesion. The work done is related to the surface tension of both the adhesive and the substrate, as given by Dupre's equation;\[[@CIT6]\] $$\omega_{A} = \ \gamma_{b} + \gamma_{t} - \gamma_{b}$$ where ω~A~ is the specific thermodynamic work of adhesion and γ~b~, γ~τ~, and γ~bt~ represent, respectively, the surface tensions of the bioadhesive polymer, the substrate, and the interfacial tension. The adhesive work done is a sum of the surface tensions of the two adherent phases, less the interfacial tensions apparent between both phases.\[[@CIT7]\] [Figure 1](#F0001){ref-type="fig"} shows a drop of liquid bioadhesive spreading over a soft-tissue surface. ::: {#F0001 .fig} Figure 1 ::: {.caption} ###### A liquid bioadhesive spreading over a typical soft tissue surface :::  ::: Horizontal resolution of the forces gives the Young equation: $$\gamma_{ta} = \ \gamma_{bt} + \gamma_{ba}\cos\theta$$ where θ is the angle of contact, γ~bt~ is the surface tension between the tissue and polymer, γ~ba~ is the surface tension between polymer and air, and γ~ta~ is the surface tension between tissue and air. [Equation 3](#FD3){ref-type="disp-formula"} states that if the angle of contact,θ, is greater than zero, the wetting will be incomplete. If the vector γ~ta~ greatly exceeds γ~bt~ + γ~ba~, that is: $$\gamma_{ta} \geq \ \gamma_{bt} + \gamma_{ba}$$ then θ will approach zero and wetting will be complete. If a bioadhesive material is to successfully adhere to a biological surface, it must first dispel barrier substances and then spontaneously spread across the underlying substrate, either tissue or mucus. The spreading coefficient, *S*~b~, can be defined as shown in [Equation 4](#FD4){ref-type="disp-formula"}: $$S_{b} = \ \gamma_{ta} - \gamma_{bt} - \gamma_{ba\ > 0}$$ which states that bioadhesion is successful if *S*~b~ is positive, thereby setting the criteria for the surface tension vectors; in other words, bioadhesion is favored by large values of γ~ta~ or by small values of γ~bt~ and γ~ba~.\[[@CIT7]\] Electrostatic Theory of Mucoadhesion {#sec1-4} ==================================== According to electrostatic theory, transfer of electrons occurs across the adhesive interface and adhering surface. This results in the establishment of the electrical double layer at the interface and a series of attractive forces responsible for maintaining contact between the two layers.\[[@CIT8]\] Diffusion Theory of Mucoadhesion {#sec1-5} ================================ Diffusion theory describes that polymeric chains from the bioadhesive interpenetrate into glycoprotein mucin chains and reach a sufficient depth within the opposite matrix to allow formation of a semipermanent bond.\[[@CIT9]\] The process can be visualized from the point of initial contact. The existence of concentration gradients will drive the polymer chains of the bioadhesive into the mucus network and the glycoprotein mucin chains into the bioadhesive matrix until an equilibrium penetration depth is achieved as shown in [Figure 2](#F0002){ref-type="fig"}. ::: {#F0002 .fig} Figure 2 ::: {.caption} ###### \(a) Schematic representation of the diffusion theory of bioadhesion. Blue polymer layer and red mucus layer before contact; (b) Upon contact; (c) The interface becomes diffuse after contact for a period of time :::  ::: The exact depth needed for good bioadhesive bonds is unclear, but is estimated to be in the range of 0.2--0.5 *μ*m.\[[@CIT10]\] The mean diffusional depth of the bioadhesive polymer segments, *s*, may be represented by [Equation 5](#FD5){ref-type="disp-formula"}: $$s = \sqrt{2tD}$$ where *D* is the diffusion coefficient and t is the contact time. Duchene\[[@CIT11]\] adapted [Equation 5](#FD5){ref-type="disp-formula"} to give [Equation 6](#FD6){ref-type="disp-formula"}, which can be used to determine the time, *t*, to bioadhesion of a particular polymer: $$t = \frac{l^{2}}{D_{b}}$$ in which *l* represents the interpenetrating depth and *D~b~* the diffusion coefficient of a bioadhesive through the substrate. Once intimate contact is achieved, the substrate and adhesive chains move along their respective concentration gradients into the opposite phases. Depth of diffusion is dependent on the diffusion coefficient of both phases. Reinhart and Peppas\[[@CIT12]\] reported that the diffusion coefficient depended on the molecular weight of the polymer strand and that it decreased with increasing cross-linking density. Adsorption Theory of Mucoadhesion {#sec1-6} ================================= According to the adsorption theory, after an initial contact between two surfaces, the materials adhere because of surface forces acting between the chemical structures at the two surfaces.\[[@CIT13]\] When polar molecules or groups are present, they reorientate at the interface.\[[@CIT7]\] Chemisorption can occur when adhesion is particularly strong. The theory maintains that adherence to tissue is due to the net result of one or more secondary forces (van der Waal's forces, hydrogen bonding, and hydrophobic bonding).\[[@CIT14]--[@CIT16]\] Fracture Theory of Adhesion {#sec1-7} =========================== This theory describes the force required for the separation of two surfaces after adhesion. The fracture strength is equivalent adhesive strength through the following equation. This theory is useful for the study of bioadhesion by tensile apparatus. $$\sigma\ = \ \left( {E\ \times \ \varepsilon/L} \right)^{1/2}$$ where σ is the fracture strength, e fracture energy, E young modulus of elasticity, and L the critical crack length.\[[@CIT17]\] Mucoadhesive Materials {#sec1-8} ====================== Mucoadhesive polymers have numerous hydrophilic groups, such as hydroxyl, carboxyl, amide, and sulfate. These groups attach to mucus or the cell membrane by various interactions such as hydrogen bonding and hydrophobic or electrostatic interactions. These hydrophilic groups also cause polymers to swell in water and, thus, expose the maximum number of adhesive sites.\[[@CIT16]\] An ideal polymer for a bioadhesive drug delivery system should have the following characteristics;\[[@CIT9][@CIT13]\] The polymer and its degradation products should be nontoxic and nonabsorbable.It should be nonirritant.It should preferably form a strong noncovalent bond with the mucus or epithelial cell surface.It should adhere quickly to moist tissue and possess some site specificity.It should allow easy incorporation of the drug and offer no hindrance to its release.The polymer must not decompose on storage or during the shelf life of the dosage form.The cost of the polymer should not be high so that the prepared dosage form remains competitive. Polymers that adhere to biological surfaces can be divided into three broad categories:\[[@CIT7][@CIT10]\] Polymers that adhere through nonspecific, noncovalent interactions which are primarily electrostatic in naturePolymers possessing hydrophilic functional groups that hydrogen bond with similar groups on biological substratesPolymers that bind to specific receptor sites on the cell or mucus surface The latter polymer category includes lectins and thiolated polymers. Lectins are generally defined as proteins or glycoprotein complexes of nonimmune origin that are able to bind sugars selectively in a noncovalent manner.\[[@CIT18]\] Lectins are capable of attaching themselves to carbohydrates on the mucus or epithelial cell surface and have been extensively studied, notably for drug-targeting applications.\[[@CIT19][@CIT20]\] These second-generation bioadhesives not only provide for cellular binding, but also for subsequent endo- and transcytosis. Thiolated polymers, also designated thiomers, are hydrophilic macromolecules exhibiting free thiol groups on the polymeric backbone. Due to these functional groups, various features of polyacrylates and cellulose derivatives were strongly improved.\[[@CIT21]\] The presence of thiol groups in the polymer allows the formation of stable covalents bonds with cysteine-rich subdomains of mucus glycoproteins leading to increased residence time and improved bioavailability.\[[@CIT22]\] Other advantageous mucoadhesive properties of thiolated polymers include improved tensile strength, rapid swelling, and water uptake behavior. [Table 1](#T0001){ref-type="fig"} shows the chemical structures of several bioadhesive polymers commonly used in modern drug delivery. ::: {#T0001 .fig} Table 1 ::: {.caption} ###### Chemical structures of some bioadhesive polymers used in drug delivery :::  ::: Factors Affecting Mucoadhesion {#sec1-9} ============================== Mucoadhesion may be affected by a number of factors, including hydrophilicity, molecular weight, cross-linking, swelling, pH, and the concentration of the active polymer.\[[@CIT9][@CIT13][@CIT23]\] Hydrophilicity {#sec1-10} ============== Bioadhesive polymers possess numerous hydrophilic functional groups, such as hydroxyl and carboxyl. These groups allow hydrogen bonding with the substrate, swelling in aqueous media, thereby allowing maximal exposure of potential anchor sites. In addition, swollen polymers have the maximum distance between their chains leading to increased chain flexibility and efficient penetration of the substrate. Molecular Weight {#sec1-11} ================ The interpenetration of polymer molecules is favored by low-molecular-weight polymers, whereas entanglements are favored at higher molecular weights. The optimum molecular weight for the maximum mucoadhesion depends on the type of polymer, with bioadhesive forces increasing with the molecular weight of the polymer up to 100,000. Beyond this level, there is no further gain.\[[@CIT24]\] Cross-linking and Swelling {#sec1-12} ========================== Cross-link density is inversely proportional to the degree of swelling.\[[@CIT25]\] The lower the cross-link density, the higher the flexibility and hydration rate; the larger the surface area of the polymer, the better the mucoadhesion. To achieve a high degree of swelling, a lightly cross-linked polymer is favored. However, if too much moisture is present and the degree of swelling is too great, a slippy mucilage results and this can be easily removed from the substrate.\[[@CIT26]\] The mucoadhesion of cross-linked polymers can be enhanced by the inclusion in the formulation of adhesion promoters, such as free polymer chains and polymers grafted onto the preformed network.\[[@CIT23]\] Spatial Conformation {#sec1-13} ==================== Besides molecular weight or chain length, spatial conformation of a polymer is also important. Despite a high molecular weight of 19,500,000 for dextrans, they have adhesive strength similar to that of polyethylene glycol (PEG), with a molecular weight of 200,000. The helical conformation of dextran may shield many adhesively active groups, primarily responsible for adhesion, unlike PEG polymers, which have a linear conformation.\[[@CIT9]\] pH {#sec1-14} == The pH at the bioadhesive to substrate interface can influence the adhesion of bioadhesives possessing ionizable groups. Many bioadhesives used in drug delivery are polyanions possessing carboxylic acid functionalities. If the local pH is above the p*K* of the polymer, it will be largely ionized; if the pH is below the p*K* of the polymer, it will be largely unionized. The approximate p*K*^a^ for the poly(acrylic acid) family of polymers is between 4 and 5. The maximum adhesive strength of these polymers is observed around pH 4--5 and decreases gradually above a pH of 6. A systematic investigation of the mechanisms of mucoadhesion clearly showed that the protonated carboxyl groups, rather than the ionized carboxyl groups, react with mucin molecules, presumably by the simultaneous formation of numerous hydrogen bonds.\[[@CIT27]\] Concentration of Active Polymer {#sec1-15} =============================== Ahuja\[[@CIT10]\] stated that there is an optimum concentration of polymer corresponding to the best mucoadhesion. In highly concentrated systems, beyond the optimum concentration the adhesive strength drops significantly. In concentrated solutions, the coiled molecules become solvent-poor and the chains available for interpenetration are not numerous. This result seems to be of interest only for more or less liquid mucoadhesive formulations. It was shown by Duchêne\[[@CIT11]\] that, for solid dosage forms such as tablets, the higher the polymer concentration, the stronger the mucoadhesion. Drug/Excipient Concentration {#sec1-16} ============================ Drug/excipient concentration may influence the mucoadhesion. BlancoFuente\[[@CIT28]\] studied the effect of propranolol hydrochloride to Carbopol^®^ (a lightly cross-linked poly(acrylic acid) polymer) hydrogels adhesion. Author demonstrated increased adhesion when water was limited in the system due to an increase in the elasticity, caused by the complex formation between drug and the polymer. While in the presence of large quantities of water, the complex precipitated out, leading to a slight decrease in the adhesive character. Increasing toluidine blue O (TBO) concentration in mucoadhesive patches based on Gantrez^®^ (poly(methylvinylether/maleic acid) significantly increased mucoadhesion to porcine cheek tissue.\[[@CIT29]\] This was attributed to increased internal cohesion within the patches due to electrostatic interactions between the cationic drug and anionic copolymer. Other Factors Affecting Mucoadhesion {#sec1-17} ==================================== Mucoadhesion may be affected by the initial force of application.\[[@CIT30]\] Higher forces lead to enhanced interpenetration and high bioadhesive strength.\[[@CIT11]\] In addition, the greater the initial contact time between bioadhesive and substrate, the greater the swelling and interpenetration of polymer chains.\[[@CIT31]\] Physiological variables can also affect mucoadhesion. The rate of mucus turnover can be affected by disease states and also by the presence of a bioadhesive device.\[[@CIT32]\] In addition, the nature of the surface presented to the bioadhesive formulation can vary significantly depending on the body site and the presence of local or systemic disease.\[[@CIT31]\] Techniques for the Determination of Mucoadhesion {#sec1-18} ================================================ The evaluation of bioadhesive properties is fundamental to the development of novel bioadhesive delivery systems. These tests are also important to screen large number of materials and their mechanisms. Numerous methods have been developed for studying mucoadhesion. Since no standard apparatus is available for testing bioadhesive strength, an inevitable lack of uniformity between test methods has arisen. Nevertheless, three main testing modes are recognized -- tensile test, shear strength, and peel strength. The most popular technique used for the determination of force of separation in bioadhesive testing is the application of force perpendicularly to the tissue/adhesive interface, during which a state of tensile stress is set up. But during the shear stress, the direction of the forces is reoriented so that it acts along the joint interface. In both tensile and shear modes, an equal pressure is distributed over the contact area.\[[@CIT33]\] The peel test is based on the calculation of energy required to detach the patch from the substrate. The peel test is of limited use in most bioadhesive systems. However, it is of value when the bioadhesive system is formulated as a patch.\[[@CIT34]\] In tensile and shear experiments, the stress is uniformly distributed over the adhesive joint, whereas in the peel strength stress is focused at the edge of the joint. Thus tensile and shear measure the mechanical properties of the system, whereas peel strength measures the resistant of the peeling force. Review of the literature confirmed that the most common technique used for the measurement of bioadhesion test is tensile strength method. McCarron *et al*.\[[@CIT26][@CIT34][@CIT35]\] and Donnelly\[[@CIT36]\] have reported extensively on the use of a commercial apparatus, in the form of a texture profile analyzer \[[Figure 3](#F0004){ref-type="fig"}\] operating in bioadhesive test mode, to measure the force required to remove bioadhesive films from excised tissue *in vitro*. ::: {#F0004 .fig} Figure 3 ::: {.caption} ###### Texture profile analyzer in bioadhesion test mode :::  ::: The texture analyzer, operating in tensile test mode and coupled with a sliding lower platform, was also used to determine peel strength of similar formulations \[[Figure 4](#F0005){ref-type="fig"}\].\[[@CIT34]\] ::: {#F0005 .fig} Figure 4 ::: {.caption} ###### Simplified representation of a typical test set-up used to determine peel strength of bioadhesive films :::  ::: Rheological techniques that study the flow and deformation of materials may be useful in predicting the mucoadhesive ability of a polymeric formulation. A simple rheological approach for polymer solutions and gels was first suggested by Hassan and Gallo.\[[@CIT37]\] In this method, rheological interaction between a polymer gel and mucin solution was determined. It was shown that a polymer gel and mucin solution mixture exhibited larger rheological responsethan the sum of the values of polymer and mucin. However, a wide variation in results is found in the literature that utilize rheological methods for mucoadhesion determination, which may be attributable to differences in mucin type and concentration,\[[@CIT38][@CIT39]\] as well as polymer concentrations.\[[@CIT40][@CIT39]\] Therefore, Hagerstrom\[[@CIT41]\] recommend that the rheological method should not be used as a stand-alone method for studying the mucoadhesive properties of the polymer gels. *In vivo* aspects of mucoadhesive testing have recently been reported to monitor the mucoadhesion on tissue surface such as the GIT or the buccal cavity. However, there are only a limited number of *in vivo* studies reported in the literature *in vitro* work because of the time, cost, and ethical constrains. The most common *in vivo* techniques to monitor mucoadhesion include GI transit times of bioadhesive-coated particles and drug release from *in situ* bioadhesive devices. Ch'ng\[[@CIT42]\] studied the *in vivo* transit time for bioadhesive beads in the rat GIT. A 51Cr-labeled bioadhesive was inserted at selected time intervals; the GITs were removed. The GIT of the rat was then cut into 20 equal segments and the radioactivity was measured. Davis\[[@CIT43]\] investigated the noninvasive *in vivo* technique to determine the transit of mucoadhesive agent. Therefore, in this study a formulation was used containing a gamma-emitting radionuclide. The release characteristics and the position polymer could be examined by gamma scintigraphy. In recent times, magnetic resonance imaging (MRI) is another noninvasive technique that is widely used. Christian Kremser\[[@CIT44]\] used MRI to detect the time and location of release of mucoadhesive formulation using dry Gd-DOTA powder. Routes of Administration for Mucoadhesive-based Drug Delivery Systems {#sec1-19} ===================================================================== Mucosa or the mucus membrane is the moist tissue that lines organs and body cavities such as mouth, gut, rectum, genital area, nose, and eye lid. Anatomical differences of the mucus membrane at varying body locations are given in [Table 2](#T0002){ref-type="table"}. Mucoadhesive drug delivery systems in the past have been formulated as powders, compacts, sprays, semisolids, or films. For example, compacts have been used for drug delivery to the oral cavity,\[[@CIT51]\] and powders and nanoparticles have been used to facilitate drug administration to the nasal mucosa.\[[@CIT52][@CIT53]\] Recently oral strips\[[@CIT54]\] were developed for tongue or buccal cavity. Details of the mucoadhesive dosage forms are given in [Table 3](#T0003){ref-type="table"}. Recently, there has been a growing interest in alternative delivery system designs. Buccal films have been suggested as a means of offering greater flexibility and comfort than adhesive tablets. In addition, films may circumvent the problem of the relatively short residence time of oral gels.\[[@CIT77]\] Film-forming bioadhesive polymers used in the production of bioadhesive films include the cellulose derivatives,\[[@CIT77]\] poly(acrylic acids) such as Carbopol,^®^[@CIT78]\] and Gantrez^®^ copolymers such as poly(methylvinylether/maleic anhydride).\[[@CIT45]\] ::: {#T0002 .table-wrap} Table 2 ::: {.caption} ###### Anatomical differences of the mucus membrane ::: Mucus membrane Relevant anatomical features --------------------- -------------------------------------------------------------------------------------------------------------------------------------------------- Buccal\[[@CIT45]\] Buccal mucosa surface area approximately 30 cm^2^ Comprised of three distinct layers -- epithelium, basement membrane, and connective tissues Buccal mucosa, sublingual are soft palate nonkeratinized tissue, and gingival are hard palate keratinized tissue Thickness of buccal epithelium is in the range of 500--800 *μ*m, 40--50 cells thick Mucus secreted by salivary glands, as a component of saliva, forming a 0.1--0.7 mm thick layer Turnover time for buccal epithelium 5--6 days Permeability barrier property of oral mucosa due to intercellular materials derived from membrane-coating granules Nasal\[[@CIT46]\] Nasal cavity surface area 160 cm^2^ Lined with mucous membrane containing columnar cells, goblet cells, and basal cells Columnar cells are covered with cilia, apart from the anterior part of the nasal cavity Both keratinized and nonkeratinized epithelial cells present depending upon location within nasal cavity Cilia responsible for mucociliary clearance Mucus secreted by the submucosal glands and the goblet cells, forming a mucus layer approximately 5--20 *μ*m thick Nasal cavity length approximately 60 mm Nasal cavity volume approximately 20 mL Turn-over time for mucus is usually 10--15 min Ocular\[[@CIT47]\] Cornea is composed of five layers -- epithelium, Bowman's layer, stroma, Descemet's membrane, and endothelium Epithelium consists of 5--6 layers of cells, with the cells of the basal layer being columnar, and the outermost cells flattened polygonal cells Tight junctions present between the basal cells of the corneal epithelium At the corneal margin, the conjunctiva is structurally continuous with the corneal epithelium The conjunctival tissue is permeable to molecules up to 20,000 Da, whereas the cornea is impermeable to molecules greater than 5000 Da The conjunctiva contains around 1.5 million goblet cells, which synthesize secretory mucins and peptides A volume of about 2--3 *μ*L of mucus os secreted daily A turnover of the mucus layer occurs in approximately 15--20 h Exposed part of the eye is covered by a thin fluid layer -- percorneal tear film Mucus Membrane Relevant Anatomical Features Tear film thickness is approximately 3--10 *μ*m Vaginal\[[@CIT48]\] Length of vagina varies from 6 to 10 cm The epithelial layer consists of the lamina propia and stratified squamous epithelium A cell turnover of about 10--15 layers is estimated to be in the order of 7 days Although there are no glands in the vagina mucosa, the surface is usually covered with vaginal fluid Major components of vaginal fluid are cervical mucus and vaginal fluid from the well-vascularized mucosa The volume, viscosity, and pH of the cervical mucus vary with age and during the menstrual cycle Rectal\[[@CIT49]\] Length approximately 15--20 cm Surface area of approximately 300 cm^2^ Epithelium consists of a single layer of cylindrical cells and goblet cells secreting mucus Flat surface, without villi, and with three major fold, the rectal valves Approximately 3 mL of mucus with a neutral pH spread over the surface ::: ::: {#T0003 .table-wrap} Table 3 ::: {.caption} ###### Different types of mucoadhesive dosage forms ::: Delivery routes Dosage forms ----------------- --------------------------------------------- -------------------------------------------------- ----------------------------------------------- ------------------------------------ --------------------------------------------- Buccal Theophylline, multiple polymers\[[@CIT55]\] Benzyl nicotinate, multiple polymers\[[@CIT56]\] Benzydamine, chitosan derivatives\[[@CIT57]\] Miconazole, PVA/PVP\[[@CIT58]\] Fentanyl, PVP\[[@CIT59]\] Nasal N/A Mupirocin, glycerin ester\[[@CIT60]\] Insulin, starch\[[@CIT61]\] Insulin, chitosan/PEG\[[@CIT62]\] Chlorpromazine, chitosan/pectin\[[@CIT63]\] Ocular Diclofenac, poly(acrylic) acid\[[@CIT64]\] Sulphadicramide, multiple polymers\[[@CIT65]\] Puerarin, poloxamer/carbopol\[[@CIT66]\] Ciprofloxacin, PVA/CMC\[[@CIT67]\] Fluorescein, HPMC\[[@CIT68]\] Vaginal Metronidazole, chitosan\[[@CIT69]\] Terameprocol, white petroleum\[[@CIT70]\] Amphotericin, pluronic\[[@CIT71]\] ALA, PMVE/MA\[[@CIT34]\] SDS, multiple polymers\[[@CIT72]\] Rectal Ramosetron, carbopol\[[@CIT73]\] Zinc oxide, petroleum\[[@CIT74]\] Quinine, HPMC\[[@CIT75]\] N/A Theophylline, pHEMA\[[@CIT76]\] ::: Oral Mucoadhesive Drug Delivery Systems {#sec1-20} ======================================= Drug delivery through the oral mucosa has gained significant attention due to its convenient accessibility. The buccal and sublingual routes are considered as the most commonly used rotes. The nonkeratinized epithelium in the oral cavity, such as the soft palate, the mouth floor, the ventral side of the tongue, and the buccal mucosa, offers a relatively permeable barrier for drug transport.\[[@CIT79]\] Hydrophilic compounds and large or highly polar molecules follow paracellular transport, whereas transcellular transport through the lipid bilayer is followed by lipophilic drugs.\[[@CIT80]\] Drug delivery through the oral mucosa has proven particularly useful and offers several advantages over other drug delivery systems including bypassing hepatic first-pass metabolism, increasing the bioavailability of drugs, improved patient compliance, excellent accessibility, unidirectional drug flux, and improved barrier permeability compared, for example, to intact skin.\[[@CIT81][@CIT82]\] The oral cavity has been used as a site for local and systemic drug delivery. Local drug therapy is used to treat disease states like aphthous ulceration gingivitis, periodontal disease, and xerostoma. Different dosage designs include adhesive gels, tablets, films, patches, ointments, mouth washes, and pastes. Until now adhesive tablets have been the most commonly used dosage forms for buccal drug delivery. Tablets can be applied to different regions of oral cavity, such as cheeks, lips, gums, and palate. Unlike conventional tablets, buccal tablets allow drinking, eating, and speaking without any major discomfort. Perioli\[[@CIT83]\] studied the influence of compression force on tablet behavior and drug release rate for mucoadhesive buccal tablets. Tablets were prepared by using hydroxyethyl cellulose (HEC) and carbopol 940 in a 1:1 ratio as matrix-forming polymers at varying compression forces. Compression forces did not significantly affect the water penetration and polymer chain stretching; however, mucoadhesion performance and drug release were influenced by compression force. Increase in compression force resulted in a decreased *in vitro* and *in vivo* drug release while giving the best *in vivo* mucoadhesive and hydration time. Moreover, it was observed that tablets prepared with the lowest force gave the best results, in comparison with tablets prepared with highest forces causing pain during *in vivo* application, needing to be detached by human volunteers. Oral mucosal ulceration is a common condition with up to 50% of healthy adults suffering from recurrent minor mouth ulcers (aphthous stomatitis). Shermer\[[@CIT84]\] evaluated the efficacy and tolerability of a mucoadhesive patch compared with a pain-relieving oral solution for the treatment of aphthous stomatitis. The mucoadhesive patch was found to be more effective than the oral solution in terms of healing time and pain intensity after 12 and 24 h. Local adverse effects 1 h after the treatment were significantly less frequent among the mucoadhesive patch patients compared with the oral solution patients. Donnelly\[[@CIT29]\] reported on a mucoadhesive patch containing TBO as a potential delivery system for use in photodynamic antimicrobial chemotherapy (PACT) of oropharyngeal candidiasis. Patches are prepared from aqueous blends of poly(methyl vinyl ether/maleic anhydride) and tripropyleneglycol methyl ether. The authors concluded that short application times of TBO-containing mucoadhesive patches should allow the treatment of recently acquired oropharyngeal candidiasis, caused solely by planktonic cells. Longer patch application times may be required for persistent disease where biofilms are implicated. Periodontitis is an inflammatory disease of the oral cavity, which results in the destruction of the supporting structures of the teeth.\[[@CIT85]\] Inflammatory periodontitis disease can be treated by the combination of mechanical and intraperiondontal pocket chemotherapeutic agents.\[[@CIT86]\] Jones and Andrews\[[@CIT87][@CIT88]\] described the formulation and physicochemical characterization of syringeable semisolid, bioadhesive networks (containing tetracycline, metronidazole, or model protein drugs). Such systems may be formulated to exhibit requisitory flow properties (and hence may be easily administered into the periodontal pocket using a syringe), mucoadhesive properties (ensuring prolonged retention within the pocket), and sustained release of therapeutic agent within this environment. Mucosal delivery of drugs *via* the buccal route is still very challenging in spite of extensive clinical studies. Here, we are underlining several formulations which are in clinical trials or commercial products. The 3M company has developed a buccal patch system which consists of a matrix patch containing drug, mucoadhesive polymers, and polymeric elastomers surrounded by a backing material. Their buprenorphine patch is capable of delivering the drug for a period up to 12 h, with good patient comfort reported.\[[@CIT89]\] Oralin, a novel liquid aerosol formulation (Generex Biotechnology), has been developed and it is now in clinical phase II trials.\[[@CIT90]\] Oralin allows precise insulin dose delivery *via* a metered dose inhaler in the form of fine aerosolized droplets directed into the mouth. Levels of drug in the mouth are noticeably increased compared with conventional formulations. This oral aerosol formulation is rapidly absorbed through the buccal mucosal epithelium, and it provides the plasma insulin levels necessary to control postprandial glucose rise in diabetic patients. This novel, pain-free, oral insulin formulation has a number of advantages, including rapid absorption, user-friendly administration technique, precise dosing control (comparable to injection within one unit), and bolus delivery of drug. Furthermore, BioAlliance Pharma's miconazole tablet (Lauriad^®^) formulation is now in clinical phase III trials, and Aphtach^®^ (triamcinolone acetonide buccal tablets from Teijin Ltd.) are now commercially available.\[[@CIT90]\] Nasal Mucoadhesive Drug Delivery Systems {#sec1-21} ======================================== The area of the normal human nasal mucosa is approximately 150 cm^2^, a highly dense vascular network and relatively permeable membrane structure; all these factors make nasal cavity interesting.\[[@CIT91]\] Drawbacks are local toxicity/irritation mucociliary clearance of 5 min, presence of proteolytic enzymes, and influence of pathological conditions (cold and allergies). Among the advantages are rapid uptake and avoiding first-pass hepatic metabolism. In addition, bioadhesive application of liquids, semisolids, and solids can significantly increase retention time. Nasal delivery of protein and peptide therapeutics can be compromised by the brief residence time at this mucosal surface. Some bioadhesive polymers have been suggested to extend residence time and improve protein uptake across the nasal mucosa. McInnes\[[@CIT92]\] quantified nasal residence of bioadhesive formulations using gamma scintigraphy and investigated absorption of insulin. A four-way crossover study was conducted in six healthy male volunteers, comparing a conventional nasal spray solution with three lyophilized nasal insert formulations (1--3% w/w hydroxypropylmethyl cellulose, HPMC). The authors concluded that the 2% w/w HPMC lyophilized insert formulation achieved extended nasal residence, demonstrating an optimum combination of rapid adhesion without overhydration. Coucke\[[@CIT93]\] studied viscosity-enhancing mucosal delivery systems for the induction of an adaptive immune response against viral antigen. Powder formulations based on spray-dried mixtures of starch (Amioca^®^) and poly(acrylic acid) (Carbopol^®^ 974P) in different ratios were used as carriers of the viral antigen. A comparison of these formulations for intranasal delivery of heat-inactivated influenza virus combined with LTR 192G adjuvant was made *in vivo* in a rabbit model. The authors concluded that the use of bioadhesive carriers based on starch and poly(acrylic acid) facilitates the induction of a systemic anti-HA antibody response after intranasal vaccination with a whole virus influenza vaccine. Functionalized mucoadhesive polymers, such as polycarbophil, hyaluronan, and amberlite resin, have been developed and the characterization and safety aspects of nasal drug products extensively studied. Recently, mucosal vaccines have been introduced in immunization to induce a systemic immune response. Addition of mucoadhesive polymer to the vaccine formulation increases the affinity for mucus membranes and may enhance the stability of the preparation. Examples of these include intranasal vaccines against influenza, diphtheria, and tetanus.\[[@CIT94]\] Pilot studies involving the use of a nasal morphine--chitosan formulation for the treatment of breakthrough pain in 14 cancer patients suggested that this system was acceptable, well-tolerated, and may lead to rapid onset of pain relief.\[[@CIT95]\] Tzachev\[[@CIT96]\] has compared a mucoadhesive solution (formulation of xylometazoline) with commercially available decongestatnt solution in 20 human subjects with perennial allergic rhinitis. The author concluded that the mucoadhesive formulation exhibited a significantly more prolonged clinical effect than the nonmucoadhesive product. Ocular Mucoadhesive Drug Delivery Systems {#sec1-22} ========================================= Drug administration to the eye is a challenge because there are several mechanisms (tear production, tear flow, and blinking) that protect the eye from the harmful agents. Conventional delivery methods are not ideal. Solutions and suspensions are readily washed from the cornea and ointments alter the tear refractive index and blur vision; so it is a target to prolong the residence time by mucoadhesion. Sensoy\[[@CIT97]\] aimed to prepare bioadhesive sulfacetamide sodium microspheres to increase residence time on the ocular surface and to enhance treatment efficacy of ocular keratitis. Microspheres were fabricated by a spray-drying method using a mixture of polymers, such as pectin, polycarbophil, and HPMC at different ratios. Author concluded that a sulfacetamide sodium--loaded polycarbophil microsphere formulation with a polymer:drug ratio of 2:1 was found to be the most suitable for ocular application and used in *in vivo* studies on New Zealand male rabbit eyes with keratitis caused by *Pseudomonas aeruginosa* and *Staphylococcus aureus*. Gene transfer is considered to be a promising alternative for the treatment of several chronic diseases that affect the ocular surface. De la Fuente\[[@CIT98]\] investigated the efficacy and mechanism of action of a bioadhesive DNA nanocarrier made of hyaluronan (HA) and chitosan (CS), specifically designed for topical ophthalmic gene therapy. The author concluded that on topical administration to rabbits, the nanoparticles entered the corneal and conjunctival epithelial cells and got assimilated by the cells. More importantly, the nanoparticles provided an efficient delivery of the associated plasmid DNA inside the cells, reaching significant transfection levels. Many clinical studies have been performed on mucoadhesive ocular dosage forms. Ocular films applied behind the eye lid were found to prolong retention time and precision of dosing. However, films were found to have a tendency to move across the surface of the eye, thus resulting in irritation, for example, from Ocusert^®^ (Alza). It has been shown that the addition of mucoadhesive polymers to ocular films reduced film movement across the eye, minimizing ocular irritation and burning sensations.\[[@CIT94]\] Baeyens\[[@CIT99]\] conducted a clinical study in dogs presenting with external ophthalmic diseases (conjunctivitis, superficial corneal ulcer, or keratoconjuctivitissicca) using soluble bioadhesive ophthalmic drug inserts (BODI^®^) in comparison with classical Tiacil^®^ eye drops from Virbac Laboratories. The results of the clinical study showed that BODI^®^ demonstrated an advantage over the Tiacil^®^ by reducing the treatment to a single application and, therefore, improving patient compliance. Mucoadhesive polymers have been incorporated into ophthalmic gels to increase gel efficacy, such as NyoGel ^®^ (timolol, Novartis) and Pilogel^®^ (pilocarpine hydrochloride, Alcon Laborataries).\[[@CIT100]\] Vaginal Mucoadhesive Drug Delivery Systems {#sec1-23} ========================================== The vagina is a fibrovascular tube connecting the uterus to the outer surface of the body. The vaginal epithelium consists of a stratified squamous epithelium and lamina propia. Dosage forms used for vaginal route are solutions, gels, suspensions, suppositories, creams, and tablets and all have short residence time.\[[@CIT101][@CIT102][@CIT103]\] Bioadhesives can control the rate of drug release from, and extend the residence time of, vaginal formulations. These formulations may contain drug or, quite simply, act in conjunction with moisturizing agents as a control for vaginal dryness. Alam\[[@CIT104]\] developed an acid-buffering bioadhesive vaginal clotrimazole (antifungal) and metronidazole (antiprotozoal and antibacterial) tablets for the treatment of genitourinary tract infections. From bioadhesion experiment and release studies, it was found that polycarbophil and sodium carboxymethyl cellulose was a good combination for an acid-buffering bioadhesive vaginal tablet. From *ex vivo* retention studies, it was found that the bioadhesive polymers held the tablet for more than 24 h inside the vagina. The cumulative release profile of the developed tablet was matched with a marketed conventional tablet (Infa-V^®^). The *in vitro* spreadability of the swelled tablet was comparable to the marketed gel. In the *in vitro* antimicrobial study, it was found that the acid-buffering bioadhesive tablet produced better antimicrobial action than marketed intravaginal drug delivery systems (Infa-V^®^, Candid-V^®^, and Canesten^®^ 1). Cevher\[[@CIT105]\] aimed to prepare clomiphene citrate (CLM) gel formulations for the local treatment of human papilloma virus infections. In this respect, 1% w/w CLM gels including polyacrylic acid (PAA) polymers such as Carbopol^®^ 934P (C934P), Carbopol^®^ 971P (C971P), Carbopol^®^ 974P (C974P) in various concentrations, and their conjugates containing thiol groups, were prepared. Author concluded that gels containing C934P-Cys showed the highest adhesiveness and mucoadhesion. A significant decrease was observed in drug release from gel formulations as the polymer concentration increased. Recent advances in polymeric technology have increased the potential of vaginal gels. Vaginal gels are semisolid polymeric matrices comprising small amounts of solid, dispersed in relatively large amounts of liquid and have been used in systems for microbicides, contraceptives, labor inducers, and other substances. Several clinical trials are in underway on microbicidal gels. Microbicidal gels are intended to improve mucosal permeation rate of microbicides for the prevention of sexually transmitted diseases. A 1% tenofovir gel is being investigated in phase II clinical trials for determining the safety and acceptability of vaginal microbicides.\[[@CIT106]\] Various clinical trials of contraceptive gels are also ongoing, with a view to determine their effectiveness. BufferGel^®^ is in phases II and III clinical trial comparing itto the Gynol II ^®^ marketed product.\[[@CIT106]\] Pharmacia conducted clinical trials of the Prostin E2^®^ suppository containing dinoprostone, and found that administration of prostaglandin E2 gel showed to be more effective in inducing labor.\[[@CIT106]\] Janssen Pharmaceutica conducted phase III clinical trial of mucoadhesive systems based on itraconazole vaginal cream containing cyclodextrins and other ingredients. Clinical investigation indicated that application of 5 g of 2% cream was well tolerated and was found to be an effective delivery system for selective vaginal delivery.\[[@CIT107]\] Rectal Mucoadhesive Drug Delivery Systems {#sec1-24} ========================================= The rectum is part of the colon, it is 10 cm in length, and has surface area 300 cm^2^. The function of the rectum is mostly concerned with removing water. Surface area without villi gives it a relatively small surface area for drug absorption.\[[@CIT54]\] Most rectal absorption of drugs is achieved by a simple diffusion process through the lipid membrane. Drugs that are liable to extensive first-pass metabolism can benefit greatly if delivered to the rectal area, especially if they are targeted to areas close to the anus. Furthermore, addition of bioadhesive polymer the migration distance in the rectum decreased. Kim\[[@CIT108]\] aimed to develop a thermoreversible flurbiprofen liquid suppository base composed of poloxamer and sodium alginate for the improvement of rectal bioavailability of flurbiprofen. Cyclodextrin derivatives, such as alpha-, beta-, gamma-cyclodextrin, and hydroxypropyl-beta-cyclodextrin (HP-beta-CD), were used to enhance the aqueous solubility of flurbiprofen. Pharmacokinetic studies were performed after rectal administration of flurbiprofen liquid suppositories with and without HP-beta-CD or after intravenous administration of a commercially available product (Lipfen^®^, flurbiprofen axetil-loaded emulsion) to rats. Flurbiprofen liquid suppository containing HP-beta-CD showed an excellent bioavailability in that the AUC of flurbiprofen after its rectal administration was not significantly different from that after intravenous administration of Lipfen^®^. The authors concluded that HP-beta-CD could be a preferable solubility enhancer for the development of liquid suppositories containing poorly water-soluble drugs. Cervical and Vulval Drug Delivery Systems {#sec1-25} ========================================= A novel bioadhesive cervical patch containing 5-fluorouracil for the treatment of cervical intraepithelial neoplasia (CIN) was described by Woolfson.\[[@CIT109]\] This patch was a bilaminar design, with a drug-loaded bioadhesive film cast from a gel containing 2% w/w Carbopol^®^ 981 plasticized with 1%w/w glycerine; the casting solvent was ethanol:water 30:70. The film, which was mechanically stable on storage under ambient conditions, was bonded directly to a backing layer formed from thermally cured poly(vinyl chloride) emulsion. Release of 5-fluorouracil from the bioadhesive layer into an aqueous sink was rapid but was controlled down to an undetectable level through the backing layer. Despite the relatively hydrophilic nature of 5-fluorouracil, substantial drug release through human cervical tissue samples was observed over approximately 20 h.\[[@CIT110]\] Donnelly\[[@CIT111]\] described the design, physicochemical characterization, and clinical evaluation of bioadhesive drug delivery systems for photodynamic therapy of difficult-to-manage vulval neoplasias and dysplasias. Aminolevulic acid (ALA) is commonly delivered to the vulva using creams or solutions, which are covered with an occlusive dressing. Such dressings are poor at staying in place at the vulva, where shear forces are high in mobile patients. To overcome the problems, the authors produced a bioadhesive patch by a novel laminating procedure. The ALA loading was 38 mg cm ^−2^ . Patches were shown to release more ALA over 6 h than the proprietary cream (Porphin^®^, 20% w/w ALA). Clinically, the patch was extensively used in successful PDT of vulval intraepithelial neoplasia, lichen sclerosus, squamous hyperplasia, Paget's disease, and vulvodynia. Gastrointestinal Mucoadhesive Drug Delivery Systems {#sec1-26} =================================================== Oral route is undoubtedly most favored route of administration, but hepatic first-pass metabolism, degradation of drug during absorption, mucus covering GI epithilia, and high turnover of mucus are serious concerns of oral route. In recent years, the gastrointestinal tract (GIT) delivery emerged as a most important route of administration. Bioadhesive retentive system involves the use of bioadhesive polymers, which can adhere to the epithelial surface in the GIT. Using bioadhesive would be achieved increase GI transit time and increase in bioavailability. Ahmed\[[@CIT112]\] studied gastric retention formulations (GRFs) made of naturally occurring carbohydrate polymers and containing riboflavin *in vitro* for swelling and dissolution characteristics as well as in fasting dogs for gastric retention. The bioavailability of riboflavin, from the GRFs was studied in fasted healthy humans and compared to an immediate release formulation. It was found that when the GRFs were dried and immersed in gastric juice, they swelled rapidly and released their drug payload in a zero-order fashion for a period of 24 h. *In vivo* studies in dogs showed that a rectangular shaped GRF stayed in the stomach of fasted dogs for more than 9 h, then disintegrated and reached the colon in 24 h. Considering pharmacokinetic parameters of human subjects under fasting conditions, bioavailability of riboflavin from a large size GRF was more than triple of that measured after administration of an immediate release formulation. Salman\[[@CIT113]\] aimed to develop polymeric nanoparticulate carriers with bioadhesive properties and to evaluate their adjuvant potential for oral vaccination. Thiamine was used as a specific ligand--nanoparticle conjugate (TNP) to target specific sites within the gastrointestinal tract, namely enterocytes and Peyer's patches. The affinity of nanoparticles to the gut mucosa was studied in orally inoculated rats. The authors concluded that thiamine-coated nanoparticles showed promise as particulate vectors for oral vaccination and immunotherapy. Conclusion {#sec1-27} ========== The mucoadhesive dosage forms offer prolonged contact at the site of administration, low enzymatic activity, and patient compliance. The formulation of mucoadhesive drug delivery system depends on the selection of suitable polymer with excellent mucosal adhesive properties and biocompatibility. Now researchers are looking beyond traditional polymers, in particular next-generation mucoadhesive polymers (lectins, thiols, etc.); these polymers offer greater attachment and retention of dosage forms. However, these novel mucoadhesive formulations require much more work, to deliver clinically for the treatment of both topical and systemic diseases. **Source of Support:** Nil **Conflict of Interest:** None declared.

PubMed Central

2024-06-05T04:04:19.995953

2011-01-01

PMC3053586

Background ========== Although bone possesses great intrinsic potential for regeneration and repair, impaired healing response (delayed union/non-union) following a fracture has been reported to range between 5-10% \[[@B1]\]. Several factors have been associated with non-union of fractures including poor mechanical stability, the presence of a gap at the fracture site, extensive soft tissue damage and open fractures, administration of pharmacological agents, such as NSAIDs, and smoking \[[@B2],[@B3]\]. However, the possible role of genetic variations on the fracture healing response among individuals and a potential genetic predisposition of atrophic non-union of fractures remain unknown. Recently, with the completion of the human genome project, the importance of genes as causes of diseases or as predisposing factors has become indisputable \[[@B4]-[@B8]\]. The observed polymorphisms demonstrated for a specific disease process are of different nature. Some of them are mutations located within endonuclease restriction sites, others are single nucleotide polymorphisms (SNPs: DNA variations at a single nucleotide) or consist of insertions or deletions of larger fragments as detected by polymerase chain reaction technique (PCR) \[[@B9]\]. During fracture healing and bone repair, a number of molecules present on the extracellular matrix regulate the cascade of events at the molecular and cellular level. Among other molecules, the group of bone morphogenetic proteins (BMPs), which are members of the transforming growth factor-beta (TGF-β) superfamily, are being extensively studied, as they exhibit powerful osteoinductive properties by inducing both proliferation and differentiation of mesenchymal stem cells (MSCs) and osteoprogenitor cells \[[@B10],[@B11]\]. Currently, a number of different human BMPs have been identified, according to their primary amino acid sequence. BMP-2 and BMP-7 are two widely studied members and are already in clinical use as osteoinductive molecules. BMP signal transduction is induced via serine/threonine kinase receptors, initiating the intracellular Smad signalling pathway. The Smad family includes three groups: the signal-transducing receptor regulated (R-Smads: 1, 2, 3, 5, 8), the common mediator (co-Smad or Smad4), and the inhibitory ones (I-Smads: 6 and 7) \[[@B12],[@B13]\]. Recently, a number of molecules displaying inhibitory properties and regulating the BMP pathway, as well as other pathways during bone regeneration, have been also identified \[[@B14]\]. A well-known extracellular inhibitor of BMPs is noggin which antagonises their actions by preventing their binding with the BMP receptors \[[@B15]\]. The purpose of this pilot study was to investigate whether genetic variants within genes of the fracture healing cascade, can be correlated with an impaired fracture healing response, by performing a preliminary SNP analysis of the BMP pathway. The primary hypothesis was that specific SNPs may be associated with the development of atrophic fracture non-unions. Other parameters known to predispose to non-union were also evaluated. Methods ======= After approval by the Local Research Ethics Committee of Leeds, (East) Research Ethics Committee (Project No: 03/220), we retrospectively studied 109 patients with long bone fractures admitted and treated in the author\'s institution from 2005 to 2007. The patients were selected randomly from the hospital database and those who met the inclusion criteria of the study were invited to participate. Only British born Caucasians were included, in an effort to have a genetically homogenous cohort of patients. All patients had initially sustained a long bone fracture as a result of a road traffic accident, a fall from height or a direct blow. All fractures were long bone fractures (open or closed, and diaphyseal or diaphyseal+metaphyseal fractures). Non-union was defined as the cessation of all healing processes and failure to achieve union after the expected period of time, as seen clinically and radiologically. Union was defined as painless, without movement fracture site or painless full weight bearing in case of fractures of the lower extremity; with the presence of bridging callus in three out of four cortices in two radiological planes \[[@B2]\]. During the selection of union patients, a quota sampling regarding open vs closed fractures (approximately 50% of each) has been performed in an effort to match the high incidence of open fractures seen in non-union patients. Exclusion criteria included children and patients with a known systemic inflammatory disease process (i.e. rheumatoid arthritis), osteoporosis and other metabolic bone diseases, pathological fractures and subsequent non-unions, hypertrophic and infected non-unions. Pregnant women and patients younger than 18 years old and older than 65 years old (for the non-union group) were also excluded from the study. Dedicated clinics for patients\' recruitment and evaluation were set up specifically for this study. The hospital notes and radiographs of all patients recruited were reviewed and such details were documented in a computerized database as patients\' demographics, initial fracture pattern, initial treatment received in terms of osteosynthesis, the presence or absence of fracture gap, whether the fracture was closed or open, intake of pharmacological agents, smoking habits, co-morbid conditions and mode of mobilisation. Blood was withdrawn and stored after informed signed consent followed by an interview and a clinical examination. DNA isolation ------------- DNA was extracted from peripheral venous blood sample using the QIAamp^®^DNA Mini Kit (Qiagen, West Sussex, UK). An aliquot of each blood sample was stored at -70°C allowing further DNA extraction, if needed. Genes and SNPs selection ------------------------ Two known BMPs: BMP-2 and BMP-7 and two inhibitory molecules of the BMP pathway: noggin and Smad6 have been selected. BMP-2 gene, located on chromosome 20p12, encompasses 2 exons with a coding region of 1191 nucleotides, produces a protein molecule of 396 amino acids that belongs to the TGF-β superfamily and induces bone and cartilage formation. It has been demonstrated that it is a crucial component for normal fracture healing. Total loss of BMP-2 is lethal; however transgenic mice, in which BMP-2 was inactivated in a limb-specific manner prior to the onset of skeletal development, had spontaneous fractures which did not resolve with time \[[@B16]\]. In particular, it is the earliest steps of fracture healing that seem to be blocked, in the absence of BMP-2, and MSCs at the repair site do not differentiate, leading to a failed healing response. The main role of BMP-2 in fracture healing is highlighted, since its absence could not be compensated efficiently by all the other osteogenic stimuli that were present in the skeleton of these animals \[[@B17]\]. BMP-7 gene (also known as osteogenic protein-1, OP-1), located on human chromosome 20q13, has a coding region of 1296 nucleotides, containing 7 exons and encodes a protein molecule of 431 amino acids that belongs to the TGF-β superfamily and also induces bone and cartilage formation. Its role on the skeleton is suggested from in vivo studies, where BMP-7-deficient mice exhibit skeletal alterations during development. This is restricted to a limited subset of skeletal elements: the rib cage (such as asymmetric pairing of ribs, fusion of ribs, and malformation of the xiphoid process), the skull, and the hind limbs (polydactyly) \[[@B18]\]. On the other hand, BMP-7 null homozygosity in mice is a postnatal lethal condition, associated with various developmental defects, including retarded ossification of bones, fused ribs and vertebrae, and polydactyly \[[@B19]\]. Additionally, in vitro data indicate that BMP-7 possesses different chondrogenic potentials and is more potent than BMP-2 in inducing chondrogenic differentiation of MSCs \[[@B20]\]. NOGGIN gene (NOG), located on human chromosome 17q21-q22, has only one exon of 699 nucleotides, which encodes a protein of 232 amino acids that binds and inactivates BMP signalling. Various animal studies highlight noggin\'s important role in skeletal physiology. Transgenic mice that over express noggin in osteoblasts exhibit reduced bone mineral densities and bone formation rate, suffer from long bone fractures and osteopenia \[[@B21],[@B22]\]. It has also been shown that exogenous noggin modifies bone formation in adult rats by inhibiting the extent of membranous ossification \[[@B23]\]. SMAD6 gene, located on chromosome 15q21-q22, encompasses 4 exons with a coding region of 1491 nucleotides, producing a protein molecule of 496 amino acids that belongs to the SMAD family of proteins and negatively regulates BMP signalling pathway. Although the SMAD6 in vivo functions are largely unknown, transgenic mice over-expressing SMAD6 showed postnatal dwarfism with osteopenia, impaired bone growth and formation with thin trabecular bone. This is thought to be caused by delayed chondrocyte hypertrophy during endochondral ossification and a reduced population of hypertrophic chondrocytes after birth \[[@B24]\]. Fifteen SNPs of the aforementioned genes of the BMP pathway have been selected to be evaluated. These SNPs had previously been identified and reported in the database of the National Centre for Biotechnology Information <http://www.ncbi.nlm.nih.gov/SNP/>, with minor allele frequencies greater than 0.2. The SNPs were randomly selected, as there were no previous studies undertaken on this topic to guide selection. For BMP-7, BMP-2 and SMAD6 genes, most SNPs that have been investigated were located in intronic regions. Also, a missense mutation located in exon 3 of BMP-2 was included in the study. Regarding the NOG gene, 3 SNPs were investigated, all of which are located in intragenic regions. Details on the exact position of each SNP within the gene and their nature are summarised in Table [1](#T1){ref-type="table"}. ::: {#T1 .table-wrap} Table 1 ::: {.caption} ###### The selected SNPs (position/function), designed primers and amplicon sizes. ::: ------------------------------------------------------------------------------------------------------------- Gene SNP SNP position\* Function Primers\ Amplicon size Forward/Reverse ------------- ----------- ------------------- ------------ ---------------------------------- --------------- ***BMP-2*** rs1005464 intron 2\ intron 5\'-TGAGCGTATATTCCCTAACC-3\'\ 378 bp c.347-2744\ 5\'-TAACCTCCCAAAAAATTAAATGAC-3\' G \> A rs235768 exon 3\ missense\ 5\'-GCAGAGCTTCAGGTTTTCCG-3\'\ 269 bp c.570\ mutation\ 5\'-TGTTTCTCCTCCAAGTGGGC-3\' A \> T (p.R190S) rs235764 intron 2\ intron 5\'-ACTGACATTTTCCGTTCCACCT-3\'\ 305 bp c.346+3126\ 5\'-TAACAGACAACTGATCAAGGAG-3\' G \> A ***BMP-7*** rs4811822 intron 2\ intron 5\'-CCCAGGGCAACAACAGTCTC-3\'\ 260 bp c.612-1290\ 5\'-CCTGGGCACACAACTTGACC-3\' C \> T rs1475000 intron 2\ intron 5\'-TGCAGATGCTGGGTCCTTAA-3\'\ 281 bp c.611+10288\ 5\'-CGGGTCAGATGCCCATGAAG-3\' G \> A rs186659 intron 1\ intron 5\'-CTGCAGGGCCTCATACACTA-3\'\ 292 bp c.419-2863\ 5\'-GAGAACAGCTTCCAGGGTGA-3\' G \> A ***NOG*** rs1442828 intragenic\ intragenic 5\'-TCCTCTTCGGTCATCCAGTG-3\'\ 199 bp g.13328730\ 5\'-TGGTGGAAACCTTGCCATTC-3\' A \> G rs1372857 intragenic\ intragenic 5\'-CTGGGAGGGTTCTTGATTGG-3\'\ 170 bp g.13334320\ 5\'-ACATGTGAAATGCAGGGCAG-3\' A \> G rs9915822 intragenic\ intragenic 5\'-TTAGGCGTCACCCACAGTTG-3\'\ 190 bp g.13320012 G \> T 5\'-TGGGCAAGGTAAATGGAAGC-3\' ***SMAD6*** rs2053423 intron 3\ intron 5\'-CATGGCTTGGATGCTTGGTGT-3\'\ 398 bp c.13320013\ 5\'-TTCCCAGTCCAAATCAGGGT-3\' C \> T rs2119261 intron 3\ intron 5\'-GCCACTACTGGACAAACCTT-3\'\ 414 bp c.952+3144\ 5\'-TCCAACAACTACTCGGCAGA-3\' C \> T rs3934908 intron 3\ intron 5\'-GAATTGGATGGAGACACGTACC-3\'\ 542 bp c.953-11868\ 5\'-GATCTGGAATGCTTCCTGAG-3\' C \> T ------------------------------------------------------------------------------------------------------------- (\* SNP position according to NCBI SNP: Geneview) ::: PCR amplification ----------------- Primers were designed in close proximity to the selected SNPs and are summarised in Table [1](#T1){ref-type="table"}. Amplification of 100 ng of genomic DNA was performed in a 50 μl reaction containing 10 mM Tris HCl pH 9.6, 50 mM KCl, 0.1% v/v Triton-X, 1.5 mM MgCl2, 200 μM dNTPs, 50pmol of each primer and 2.5 units of Taq DNA Polymerase (Promega, Madison, USA). PCR reactions were heated on a PTC-225 Thermal Cycler (MJ Research Inc, USA) at 96°C for 2 minutes, followed by 40 cycles of denaturation at 94°C for 45 seconds, annealing at 55°C for 45 seconds and extension at 72°C for 60 seconds. A final extension step was performed at 72°C for 7 minutes. PCR products were purified using the Exonuclease I/Shrimp Alkaline Phosphatase Method (ExoSAP-IT^®^, USB, Staufen, Germany). Sequence Analysis ----------------- Automated cycle sequencing for both strands was performed with the BigDye Terminator Cycle Sequencing kit (Applied Biosystems, Warrington, UK). DNA template of 0.2 pmol was mixed with 8 μl sequencing reagent premix and 5 pmol primer and was initially denaturated at 96 °C for 2 minutes, followed by 50 cycles of denaturation at 94 °C for 45 seconds, annealing at 50-55 °C for 45 seconds and extension at 72 °C for 2 minutes. PCR products were then electrophorized in an ABI Prism^®^3100 Genetic Analyzer (Applied Biosystems, Warrington, UK). Sequences obtained were aligned using the Sequencher^®^PC software (Gene Codes, USA) with normal sequences taken from Genbank and examined for the presence of polymorphisms (Figure [1](#F1){ref-type="fig"}). ::: {#F1 .fig} Figure 1 ::: {.caption} ###### **Electropherograms of six patients showing two of the selected SNPs**. Electropherograms of six patients showing the three possible genotypes of the two SNPs found to be statistically significant: the rs1372857 of the NOG gene (1a, b and c) and the rs2053423 of the SMAD6 gene (2a, b and c). :::  ::: Statistical Analysis -------------------- Initial exploratory statistical analysis of the various parameters independently, comparing atrophic non-union patients (Group A) with union patients (Group B), was performed using Student\'s t-test and Pearson\'s Chi-square test. The overall frequencies of the 15 SNPs were computed for all cases with respect to patient\'s age, gender, smoking habits, and the use of NSAIDs. Since the outcome measure of interest, namely the non-union or union of the fracture, is dichotomous, a binary logistic regression was used to permit the exploration of many covariates simultaneously \[[@B25]\]. Statistical analysis was performed using STATA 11.1 (StataCorp, Texas USA). A *p*value of 0.05 or less was considered as statistically significant. Odds ratios were calculated in order to evaluate the size of the effect of the tested covariates and describe the strength of their association or non-independence to non-union \[[@B26]\]. Results ======= There were sixty-two patients (45 men and 17 women) with atrophic long bone non-unions (Group A/non-union group) with a mean age of 43.9 years (range: 19-65). There were 18 femoral non-unions, 41 tibial, 2 humeral and 1 ulnar non-unions. All these cases required further intervention to achieve union. Control group (Group B/union group) consisted of forty-seven patients (33 men and 14 women) who had uneventful fracture union, with a mean age of 38.4 years (range: 19-78). There was a total of 54 long bone fractures (22 femoral, 26 tibial, 5 humeral and 1 ulnar). The main documented parameters in both groups of patients are summarised in Table [2](#T2){ref-type="table"}. The frequencies of all the SNPs genotypes for both groups are summarised in Table [3](#T3){ref-type="table"}. ::: {#T2 .table-wrap} Table 2 ::: {.caption} ###### Individual parameters for patients in Groups A and B and *p*values. ::: -------------------------------------------------------------------------------------------- Parameters Group A\ Group B\ *p\** *p^§^*\[OR, (95% CI)\] (Atrophic\ (Union) non-union) ---------------------- ------------ ---------- ----------- --------------------------------- Mean age 43.9 yrs 38.4 yrs **0.025** **0.01 \[1.05, (1.01, 1.08)\]** Sex M/F 45/17 33/14 0.83 0.83 \[1.11, (0.44, 2.76)\] Open fractures 45.9% 51.1% n/a \- Smoking 46.8% 35.6% 0.32 0.12 \[1.99, (0.83, 4.74)\] NSAIDs 38.3% 22.2% 0.09 0.17 \[1.92, (0.76, 4.82)\] Fracture comminution 32.2% 27.6% 0.67 \- Segmental fracture 9.7% 2.1% 0.14 \- Bone gap 14.5% 6.4% 0.22 \- Implant failure 6.4% 0 0.13 \- Total number 62 47 n/a \- -------------------------------------------------------------------------------------------- \[n/a: non applicable, ***p***\*: simple statistical analysis using Student\'s t-test for the parameter of age and Pearson\'s Chi-square test for the other parameters, ***p^§^***: logistic regression results for individual parameters, odds ratios (OR) and 95% confidence interval (CI)\] ::: ::: {#T3 .table-wrap} Table 3 ::: {.caption} ###### SNPs frequencies and logistic regression results for the non-union status. ::: -------------------------------------------------------------------------------------------------------------------------- Gene SNP Genotypes Group A\ Group B\ Age unadjusted Age unadjusted Age adjusted Age adjusted (non-union) (union) --------- ----------- ----------- ------------- ---------- ---------------- ---------------- -------------- -------------- **OR** ***p*** **OR** ***p*** C/C 20 13 1.00 1.00 rs4811822 C/T 26 27 0.63 0.30 0.57 0.23 T/T 16 7 1.49 0.49 1.29 0.67 C/C 17 12 1.00 1.00 rs4811823 C/T 27 27 0.71 0.45 0.61 0.31 *BMP-7* T/T 18 8 1.59 0.42 1.36 0.60 A/A 25 12 1.00 1.00 rs1475000 G/A 28 28 0.48 0.10 0.48 0.10 G/G 9 7 0.62 0.43 0.70 0.57 A/A 27 15 1.00 1.00 rs186659 G/A 27 26 0.58 0.19 0.58 0.21 G/G 8 6 0.74 0.63 0.80 0.73 A/A 5 6 1.00 1.00 rs1005464 G/A 21 11 2.29 0.24 1.90 0.38 G/G 36 30 1.44 0.58 1.13 0.85 A/A 11 9 1.00 1.00 *BMP-2* rs235768 T/A 31 24 1.06 0.92 1.10 0.86 T/T 20 14 1.17 0.78 1.34 0.62 A/A 8 7 1.00 1.00 rs235764 G/A 24 13 1.62 0.44 1.67 0.42 G/G 30 27 0.97 0.96 1.21 0.76 A/A 11 4 1.00 1.00 rs1442828 G/A 33 21 0.57 0.39 0.60 0.45 G/G 18 22 0.30 0.07 0.28 0.06 A/A 16 19 1.00 1.00 *NOG* rs1372857 G/A 30 24 1.48 0.37 1.63 0.27 G/G 16 4 4.75 **0.02** 4.56 **0.02** G/G 10 10 1.00 1.00 rs9915822 G/T 29 27 1.07 0.89 1.28 0.65 T/T 23 10 2.30 0.16 2.26 0.17 C/C 1 5 1.00 1.00 rs2053423 C/T 25 20 6.25 0.11 7.74 0.09 T/T 36 22 8.18 **0.05** 10.27 **0.04** C/C 23 20 1.00 1.00 rs2119261 C/T 34 21 1.41 0.41 1.41 0.41 T/T 5 6 0.72 0.64 1.00 0.99 C/C 36 22 1.00 1.00 *SMAD6* rs2119260 C/T 25 21 0.73 0.43 0.73 0.45 T/T 1 4 0.15 0.10 0.11 0.07 A/A 2 1 1.00 1.00 rs3934907 C/A 18 11 0.82 0.88 0.85 0.90 C/C 42 35 0.60 0.68 0.62 0.71 C/C 22 16 1.00 1.00 rs3934908 C/T 30 24 0.91 0.82 0.95 0.92 T/T 10 7 1.04 0.95 1.05 0.93 -------------------------------------------------------------------------------------------------------------------------- \[Logistic regression results of genotypes for non-union status: ***p***: *p-*value, **OR:**odds ratios\] ::: Simple statistical analysis using Student\'s t-test for the parameter of age and Pearson\'s Chi-square test for the other parameters, including patient\'s related parameters (gender, smoking, use of NSAIDs), fracture pattern\'s related factors (comminution, segmental fracture, bone gap) and implant failure, revealed that age was the only statistically significant parameter for the development of fracture non-union with a *p*value of 0.025 (Table [2](#T2){ref-type="table"}). In undertaking these separate univariable tests, multiple testing has occurred. As this is only an exploratory study, no adjustment (such as Bonferroni) has been employed. As the univariate analyses did not allow any adjustment for other factors and in order to assess factors concurrently, a multiple logistic regression of individual parameters in non-union was performed. This yielded adjusted odds ratios (OR) for individual parameters, including age, gender, use of NSAIDs amd smoking (Table [2](#T2){ref-type="table"}). Age was found to be an important covariate (*p*= 0.01, OR 1.05 \[per year\], 95% CI \[1.01, 1.08\]), whereas the other parameters were found no statistical significant (*p*= 0.83 for gender, *p*= 0.17 for NSAIDs and *p*= 0.12 for smoking). These findings were consistent between unadjusted and adjusted analyses. Therefore, for the consideration of SNPs only age was used as an adjusting covariate, thus simplifying the analysis. The coefficients from binary logistic regression of the non-union status on the SNP base-pair combinations with and without the inclusion of age as a covariate, based on the previously reported statistical finding are summarised in Table [3](#T3){ref-type="table"}. The G/G genotype compared to the A/A genotype of the SNP: rs1372857, located on NOG gene was found to be statistically significant (*p*= 0.02, OR 4.56, age adjusted 95% CI \[1.24,16.79\]), when union to non-union patients were compared. Specifically, 25.8% from the non-union group were found to be carriers of the G/G genotype compared to the A/A genotype (25.8%), whereas only 8.5% from the union group were found to be carriers of the G/G genotype compared with 40.4% A/A. The same effect (*p*= 0.02, OR 4.75, 95% CI \[1.32,17.11\]) was observed when analysis was performed without adjusting for age. The T/T genotype of the SNP: rs2053423 located on SMAD6 gene was noted also to be statistically significant (*p*= 0.04, OR 10.27, age adjusted 95% CI \[0.98,107.81\]) when compared to the C/C genotype, when union to non-union patients was compared with the age parameter adjusted. Specifically, 58.1% from the non-union group were found to be carriers of the T/T genotype compared to the C/C genotype (1.6%), whereas 46.8% from the union group were found to be carriers of the same genotype compared to 10.7% with the C/C genotype. A similar effect was found to be on the boarders of statistical significance, when analysis was performed without the age parameter adjusted (*p*= 0.05, OR 8.18, 95% CI \[0.90,74.69\]). Discussion ========== Despite the multi-factorial nature of fracture non-unions, it has been our observation that patients, with comparable fracture patterns and risk factors, may or may not develop non-union. This phenomenon may reflect the presence of a genetic component to impaired bone regeneration and fracture healing. Differences seen in fracture healing response and final outcome therefore may be attributed to biological variations among patients resulting in a \'disturbed\' signalling pathway. Genetic variability was found to significantly contribute to the process of bone regeneration \[[@B27]\] and genetic differences between mice strains were shown to affect the length of each stage of fracture healing and the overall healing rate \[[@B28]\]. Therefore, the genetic contribution with or without the interaction of other exogenous factors in cases of impaired fracture healing, is yet to be elucidated. The characterisation of important mediators regulating the fracture healing process, the advances made in diagnostic techniques and the completion of the human genome facilitate the design of studies to explore the potential of disturbed or inhibited physiological processes to be considered in terms of genetic susceptibility. It can therefore be speculated that the \"inert or deficient local biology\" at the fracture site seen in these cases may represent a genetically predisposed environment with reduced potentials for bone regeneration. Specific genetic variations within the genes involved in fracture healing, and in particular in the BMP signalling pathway therefore, may contribute to the development of atrophic non-union. In the present study we evaluated a total of fifteen SNPs of specific genes implicated in the BMP pathway, as well as other patient and fracture related factors in the development of atrophic non-unions. Our analysis showed that two specific SNPs and age are statistically significantly associated with atrophic non-union. To the best of our knowledge, this is the first study to investigate the potential of genetic susceptibility to fracture non-union and to suggest a possible genetic association to the impaired bone healing response occurring in these patients. In particular, two genetic polymorphisms in genes involved in the BMP pathway, one found close (\~1 Kb) to NOG gene (G/G genotype of the SNP rs1372857) and the other within intron 3 of SMAD6 gene (T/T genotype of the SNP rs2053423) have been associated to the non-union phenotype. Hence, patients with these two particular genotypes may have an increased risk for the development of atrophic non-union. Additionally, an initial exploration of the data revealed that age is an important co-factor and combining the genetic analysis with age as a covariate, a higher impact of our findings was noted. Noggin, a major antagonist of BMPs, has important functions in respect to bone healing and bone formation. Several mutations of its gene (*NOG*) have been previously reported to be implicated in skeletal anomalies, such as proximal symphalangism, tarsal/carpal coalition syndrome, and brachydactyly type B (BDB) \[[@B29],[@B30]\]. Noggin mutations have also been reported in fibrodysplasia ossificans progressiva (FOP) \[[@B31]\], which is a rare autosomal dominant disorder of skeletal malformations and progressive extraskeletal ossification caused mainly by mutations in the BMP type I receptor ACVR1 \[[@B32]\]. Noggin mutations seem to alter its binding ability to BMPs and growth-differentiation factors (GDFs), thus interfering with the canonical BMP signalling pathway. Animal studies have shown that transgenic mice overexpressing noggin suffered from long bone fractures and osteopenia \[[@B21]\], whereas the knockout noggin mouse was lethal with severe skeletal defects, such as multiple joint fusions \[[@B33]\]. Additionally, exogenous noggin was found to modify bone formation by inhibiting the extent of membranous ossification \[[@B23]\]. The important role of noggin in the BMP cascade is demonstrated by experimental data coming from recent in vitro studies, showing that noggin\'s reduction promotes osteogenesis by the enhancement of BMP signalling and that addition of noggin successfully blocks in vitro osteogenic differentiation by 50%, resulting in the lowering of BMPs endogenous levels \[[@B6],[@B34]\]. A remarkable finding, highlighting noggin\'s potential important role in bone healing, was that the bone regenerated closely resembled to normal bone, when the muscle stem cells engineered to express noggin were co-implanted with transduced muscle stem cells producing BMP-4. When the muscle stem cells producing BMP-4 were implanted alone, bone overgrowth was observed. In addition, the co-localisation observed between noggin and BMP-4 during fracture healing, suggests that the balance between them may be an important factor in the regulation of callus formation \[[@B35]\]. Taking into account all reported findings from human, animal and in vitro studies that underline the importance of noggin in skeletal physiology, and in an effort to interpret our results, it can be speculated that specific genetic variations within the NOG gene may be associated with impaired bone healing. In our study, the G/G genotype of the SNP rs1372857 of NOG gene was found to be statistically significant (*p*= 0.02) suggesting a possible association with the defective fracture healing process, seen in atrophic non-unions. A patient with this genotype was found to be of the order of 4 times more likely to develop non-union than a patient with the A/A genotype (OR = 4.75, unadjusted for age 95% CI \[1.32,17.11\], and OD 4.56, age adjusted 95% CI \[1.24,16.79\]). Smad6 is an intracellular inhibitor of the BMP pathway. Although its in vivo functions are largely unknown, transgenic mice overexpressing Smad6 in chondrocytes showed postnatal dwarfism with osteopenia and impaired bone growth and formation, caused by delayed chondrocyte hypertrophy during endochondral ossification \[[@B24]\]. These findings may indicate a distinct role for Smad6 in bone regeneration. However, there are no human genetic data correlating SMAD6 mutations with syndromes with skeletal involvement, except from a recently published study assessing the role of the SMAD6 in the regulation of bone mass. Association analysis between bone mineral density and SMAD6 SNPs in 721 Japanese postmenopausal women identified a specific SNP (rs755451), located on intron 3 of SMAD6, to be associated with lower bone mineral density in postmenopausal women, and thus increasing the risk of osteoporosis. This finding marks the regulatory role of SMAD6 in bone homeostasis \[[@B36]\]. In the present study, logistic regression showed that the T/T genotype of the rs2053423 SNP within the same intron (intron 3) of SMAD6 was more frequent compared to the C/C genotype within non-union patients, with an odds ratio of ≈ 8 and 10 (after age adjustment) for fracture patients to develop non-union if they have the T/T genotype of this particular SNP compared to the C/C genotype. In the herein study, statistical analysis of other factors known to predispose to atrophic non-union did not reveal any association except the covariate of age, with an odds ratio of 1.05 for one year of age. This can be attributed to the effect of aging at the cellular level. Although, MSCs were reported to maintain their differentiation potential during aging \[[@B37]\], researchers found aging to be associated with decreased proliferative capacity of osteoprogenitor cells and therefore with a decreased osteoblastic cell number and osteoblastogenesis \[[@B38]\]. Study limitations ----------------- The sample size of patients that were included is relatively small to perform a fully powered statistical analysis and to make firm conclusions about the likelihood of genetic predisposition to fracture non-union. Also, the heterogeneity of the two groups regarding the mechanism and pattern of the fractures, the use of NSAIDs, smoking, the different treatment options, as well as the small number of SNPs and genes involved in fracture healing that were evaluated, could be considered as additional limitations of this case control association study. Thus, it may be possible that some associations have not been detected and the role of other genes has not been evaluated. In addition, due to the high cost of genetic profiling, it was not possible to perform a full cohort study or to examine a larger number of SNPs either for the selected genes or for a number of other genes known to be expressed during fracture healing. However, it should be noted that this study was undertaken as a pilot study based on observational data, aiming to explore a potential impact of genetic variations on the development of fracture non-union and to stimulate further research into a number of candidate genes of the bone healing cascade. A second limitation of the present study is that tag SNPs have not been considered and most SNPs that have been investigated were located in intronic or intragenic regions. This selection was made, because the majority of the published exonic sequence variants for the selected genes were either synonymous or frame shift or nonsense mutations, suggesting that they were pathogenic mutations. However, these intronic or intragenic SNPs, especially for the relatively small genes that have been investigated in this study, may be linked with genetic variants within the coding and regulatory regions of these genes due to the phenomenon of linkage disequilibrium. In general, the actual association between genes and disease, and especially in polygenic and exogenously affected traits, can be performed using intronic or intragenic variants, commonly observed within the general population, as tags for the identification of the actual predisposing variants \[[@B39],[@B40]\]. In general, the BMP signalling pathway is implicated in a variety of processes during development and adult life, and is expressed in different tissues besides bone. Therefore, no major genetic alterations, if any, within the BMP signalling genes are expected to be found, even if genetic predisposition to atrophic non-union does exist. Important mutations of these genes are linked to major phenotype alterations and defects, and even lethal conditions. However, multiple low penetrance alleles (each with a small effect) may interact and be associated with the defective bone regeneration, seen in atrophic non-unions. Furthermore, since no previous work has been published assessing the genetic predisposition for the development of atrophic non-union (using SNPs or any other genetic markers), this pilot study aimed to explore a possible genetic impact by selecting a small initial number of SNPs located on specific genes implicated in fracture healing. With the herein preliminary study, the analysis can only be exploratory, aiming to suggest candidate covariates for further studies and stimulate future research. Larger trials may determine that a number of genetic variants in combination with other known factors may be influential in the healing of fractures and more subtle effects may be revealed. The genetic input to the impaired fracture healing may be determined by the genotyping of exonic polymorphisms not only within the already short-listed genes, but also within other genes involved in the complex cascade of bone healing. Advanced technology, such as custom-built microarrays can be very helpful and can simultaneously examine a large number of SNPs within the numerous genes expressed during fracture healing, in order to identify functional polymorphisms as well as influential combinations of SNPs with a higher predictive power. Clinical relevance ------------------ From the clinical perspective, analysis of SNPs linked to aberrant bone healing can be used as a potential powerful tool to rapidly identify patients at risk of developing atrophic fracture non-union. As most fractures unite uneventfully, one may argue that it may not be cost effective to subject all fracture patients to genetic testing. However, such analysis may be valuable in patients that demonstrate slow progression to union or no progression at all, especially in the absence of other known risk factors or in particular in the older patient with a long bone fracture non-union. In these cases, early intervention to augment the local biology for bone regeneration, could facilitate the union of the fracture and even accelerate the time to union. Consequently, greater knowledge of the genes involved in fracture repair may provide new approaches at the molecular level in the treatment of these patients and the on-time intervention in the biologic aspects of bone healing. Currently, biological response modifiers are already in clinical use or under extensive investigation as alternatives or adjuvants for the management of defective bone healing. Specifically, with the use of recombinant technology, BMP-2 and BMP-7 are available for implantation for acceleration or stimulation of bone regeneration in cases of open fractures and atrophic fracture non-unions, respectively \[[@B41],[@B42]\]. However, these currently available treatment modalities do not address the issue of possible isolated gene disturbances. There are novel methods such as gene therapy (with local or systemic administration) and tissue engineering, which aim to address such issues, but are still under investigation. If genetic predisposition to atrophic non-union does exist, such expensive modalities may be used to selected (after genetic testing) patients. Conclusion ========== This pilot study investigated the possible impact of genetic predisposition to atrophic fracture non-union, by assessing a number of candidate genes of the different signalling pathways of fracture healing, and suggests the potential existence of a genetically predetermined impairment within the BMP signaling cascade, initiated after a fracture and when combined with other risk factors could synergistically increase the susceptibility of a patient to develop non-union. The genetic component and its role and interaction with other risk factors in the development of atrophic fracture non-unions merit further investigation. Competing interests =================== The authors declare that they have no competing interests. Authors\' contributions ======================= RD is the main author of this paper, carried out the laboratory work (DNA extraction, PCR, sequencing) as well as the recruitment and clinical and radiological assessment of the patients and the analysis and interpretation of the data. IMC supervised and also contributed in the laboratory work, participated in the design of the study and helped to draft the manuscript. His contribution was also in the analysis and interpretation of the results. RBW made substantial contributions to the analysis, interpretation and presentation of data; by performing the statistical analysis and helping to write the results of the study. AFM made substantial contributions to conception and design of the study, and participated in its coordination. PVG made substantial contributions to conception and design of the study, acquisition and interpretation of data, and coordinated the study. He helped to draft the manuscript and revised it critically for important intellectual content. All authors read and have given final approval of the final manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2474/12/44/prepub> Acknowledgements ================ This project was supported by an outside funding from the AO foundation, Switzerland, grant No: 03-G43, for the purchase of laboratory materials.

PubMed Central

2024-06-05T04:04:20.001272

2011-2-10